terraform/internal/terraform/context_validate_test.go

2034 lines
52 KiB
Go
Raw Normal View History

package terraform
import (
"errors"
"fmt"
"strings"
"testing"
"github.com/zclconf/go-cty/cty"
"github.com/hashicorp/terraform/internal/addrs"
"github.com/hashicorp/terraform/internal/configs/configschema"
"github.com/hashicorp/terraform/internal/providers"
"github.com/hashicorp/terraform/internal/provisioners"
"github.com/hashicorp/terraform/internal/states"
"github.com/hashicorp/terraform/internal/tfdiags"
)
func TestContext2Validate_badCount(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = getProviderSchemaResponseFromProviderSchema(&ProviderSchema{
ResourceTypes: map[string]*configschema.Block{
"aws_instance": {
Attributes: map[string]*configschema.Attribute{},
},
},
})
m := testModule(t, "validate-bad-count")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_badResource_reference(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = getProviderSchemaResponseFromProviderSchema(&ProviderSchema{
ResourceTypes: map[string]*configschema.Block{
"aws_instance": {
Attributes: map[string]*configschema.Attribute{},
},
},
})
m := testModule(t, "validate-bad-resource-count")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_badVar(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = getProviderSchemaResponseFromProviderSchema(&ProviderSchema{
ResourceTypes: map[string]*configschema.Block{
"aws_instance": {
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
"num": {Type: cty.String, Optional: true},
},
},
},
})
m := testModule(t, "validate-bad-var")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_varNoDefaultExplicitType(t *testing.T) {
m := testModule(t, "validate-var-no-default-explicit-type")
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
c, diags := NewContext(&ContextOpts{})
if diags.HasErrors() {
t.Fatalf("unexpected NewContext errors: %s", diags.Err())
}
// NOTE: This test has grown idiosyncratic because originally Terraform
// would (optionally) check variables during validation, and then in
// Terraform v0.12 we switched to checking variables during NewContext,
// and now most recently we've switched to checking variables only during
// planning because root variables are a plan option. Therefore this has
// grown into a plan test rather than a validate test, but it lives on
// here in order to make it easier to navigate through that history in
// version control.
_, diags = c.Plan(m, states.NewState(), DefaultPlanOpts)
if !diags.HasErrors() {
// Error should be: The input variable "maybe_a_map" has not been assigned a value.
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_computedVar(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"value": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
pt := testProvider("test")
pt.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"test_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"id": {Type: cty.String, Computed: true},
"value": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "validate-computed-var")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
addrs.NewDefaultProvider("test"): testProviderFuncFixed(pt),
},
})
p.ValidateProviderConfigFn = func(req providers.ValidateProviderConfigRequest) (resp providers.ValidateProviderConfigResponse) {
2020-10-08 18:43:46 +02:00
val := req.Config.GetAttr("value")
if val.IsKnown() {
resp.Diagnostics = resp.Diagnostics.Append(fmt.Errorf("value isn't computed"))
}
2020-10-08 18:43:46 +02:00
return
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
if p.ConfigureProviderCalled {
2020-10-08 18:26:12 +02:00
t.Fatal("Configure should not be called for provider")
}
}
func TestContext2Validate_computedInFunction(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"attr": {Type: cty.Number, Optional: true},
},
},
},
},
DataSources: map[string]providers.Schema{
"aws_data_source": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"optional_attr": {Type: cty.String, Optional: true},
"computed": {Type: cty.String, Computed: true},
},
},
},
},
}
m := testModule(t, "validate-computed-in-function")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
// Test that validate allows through computed counts. We do this and allow
// them to fail during "plan" since we can't know if the computed values
// can be realized during a plan.
func TestContext2Validate_countComputed(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
DataSources: map[string]providers.Schema{
"aws_data_source": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"compute": {Type: cty.String, Optional: true},
"value": {Type: cty.String, Computed: true},
},
},
},
},
}
m := testModule(t, "validate-count-computed")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_countNegative(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
m := testModule(t, "validate-count-negative")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_countVariable(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "apply-count-variable")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_countVariableNoDefault(t *testing.T) {
p := testProvider("aws")
m := testModule(t, "validate-count-variable")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
c, diags := NewContext(&ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
assertNoDiagnostics(t, diags)
_, diags = c.Plan(m, nil, &PlanOpts{})
if !diags.HasErrors() {
// Error should be: The input variable "foo" has not been assigned a value.
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_moduleBadOutput(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "validate-bad-module-output")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_moduleGood(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "validate-good-module")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_moduleBadResource(t *testing.T) {
m := testModule(t, "validate-module-bad-rc")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateResourceConfigResponse = &providers.ValidateResourceConfigResponse{
Diagnostics: tfdiags.Diagnostics{}.Append(fmt.Errorf("bad")),
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_moduleDepsShouldNotCycle(t *testing.T) {
m := testModule(t, "validate-module-deps-cycle")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"id": {Type: cty.String, Optional: true},
},
},
},
},
}
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_moduleProviderVar(t *testing.T) {
m := testModule(t, "validate-module-pc-vars")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateProviderConfigFn = func(req providers.ValidateProviderConfigRequest) (resp providers.ValidateProviderConfigResponse) {
2020-10-08 18:43:46 +02:00
if req.Config.GetAttr("foo").IsNull() {
resp.Diagnostics = resp.Diagnostics.Append(errors.New("foo is null"))
}
return
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_moduleProviderInheritUnused(t *testing.T) {
m := testModule(t, "validate-module-pc-inherit-unused")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateProviderConfigFn = func(req providers.ValidateProviderConfigRequest) (resp providers.ValidateProviderConfigResponse) {
2020-10-08 18:43:46 +02:00
if req.Config.GetAttr("foo").IsNull() {
resp.Diagnostics = resp.Diagnostics.Append(errors.New("foo is null"))
}
return
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_orphans(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
"num": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "validate-good")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateResourceConfigFn = func(req providers.ValidateResourceConfigRequest) providers.ValidateResourceConfigResponse {
var diags tfdiags.Diagnostics
if req.Config.GetAttr("foo").IsNull() {
diags = diags.Append(errors.New("foo is not set"))
}
return providers.ValidateResourceConfigResponse{
Diagnostics: diags,
}
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_providerConfig_bad(t *testing.T) {
m := testModule(t, "validate-bad-pc")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateProviderConfigResponse = &providers.ValidateProviderConfigResponse{
Diagnostics: tfdiags.Diagnostics{}.Append(fmt.Errorf("bad")),
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if len(diags) != 1 {
t.Fatalf("wrong number of diagnostics %d; want %d", len(diags), 1)
}
if !strings.Contains(diags.Err().Error(), "bad") {
t.Fatalf("bad: %s", diags.Err().Error())
}
}
2020-05-08 07:26:21 +02:00
func TestContext2Validate_providerConfig_skippedEmpty(t *testing.T) {
2020-12-06 18:35:35 +01:00
m := testModule(t, "validate-skipped-pc-empty")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateProviderConfigResponse = &providers.ValidateProviderConfigResponse{
Diagnostics: tfdiags.Diagnostics{}.Append(fmt.Errorf("should not be called")),
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
2020-11-10 01:26:17 +01:00
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_providerConfig_good(t *testing.T) {
m := testModule(t, "validate-bad-pc")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
// In this test there is a mismatch between the provider's fqn (hashicorp/test)
// and it's local name set in required_providers (arbitrary).
func TestContext2Validate_requiredProviderConfig(t *testing.T) {
m := testModule(t, "validate-required-provider-config")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"required_attribute": {Type: cty.String, Required: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_provisionerConfig_bad(t *testing.T) {
m := testModule(t, "validate-bad-prov-conf")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
pr := simpleMockProvisioner()
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-11-13 18:43:28 +01:00
Provisioners: map[string]provisioners.Factory{
"shell": testProvisionerFuncFixed(pr),
},
})
p.ValidateProviderConfigResponse = &providers.ValidateProviderConfigResponse{
Diagnostics: tfdiags.Diagnostics{}.Append(fmt.Errorf("bad")),
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_badResourceConnection(t *testing.T) {
m := testModule(t, "validate-bad-resource-connection")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
pr := simpleMockProvisioner()
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-11-13 18:43:28 +01:00
Provisioners: map[string]provisioners.Factory{
"shell": testProvisionerFuncFixed(pr),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
t.Log(diags.Err())
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_badProvisionerConnection(t *testing.T) {
m := testModule(t, "validate-bad-prov-connection")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
pr := simpleMockProvisioner()
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-11-13 18:43:28 +01:00
Provisioners: map[string]provisioners.Factory{
"shell": testProvisionerFuncFixed(pr),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
t.Log(diags.Err())
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_provisionerConfig_good(t *testing.T) {
m := testModule(t, "validate-bad-prov-conf")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
Provider: providers.Schema{
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
pr := simpleMockProvisioner()
pr.ValidateProvisionerConfigFn = func(req provisioners.ValidateProvisionerConfigRequest) provisioners.ValidateProvisionerConfigResponse {
var diags tfdiags.Diagnostics
if req.Config.GetAttr("test_string").IsNull() {
diags = diags.Append(errors.New("test_string is not set"))
}
return provisioners.ValidateProvisionerConfigResponse{
Diagnostics: diags,
}
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-11-13 18:43:28 +01:00
Provisioners: map[string]provisioners.Factory{
"shell": testProvisionerFuncFixed(pr),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_requiredVar(t *testing.T) {
m := testModule(t, "validate-required-var")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"ami": {Type: cty.String, Optional: true},
},
},
},
},
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
c, diags := NewContext(&ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
assertNoDiagnostics(t, diags)
// NOTE: This test has grown idiosyncratic because originally Terraform
// would (optionally) check variables during validation, and then in
// Terraform v0.12 we switched to checking variables during NewContext,
// and now most recently we've switched to checking variables only during
// planning because root variables are a plan option. Therefore this has
// grown into a plan test rather than a validate test, but it lives on
// here in order to make it easier to navigate through that history in
// version control.
_, diags = c.Plan(m, states.NewState(), DefaultPlanOpts)
if !diags.HasErrors() {
// Error should be: The input variable "foo" has not been assigned a value.
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_resourceConfig_bad(t *testing.T) {
m := testModule(t, "validate-bad-rc")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateResourceConfigResponse = &providers.ValidateResourceConfigResponse{
Diagnostics: tfdiags.Diagnostics{}.Append(fmt.Errorf("bad")),
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if !diags.HasErrors() {
t.Fatalf("succeeded; want error")
}
}
func TestContext2Validate_resourceConfig_good(t *testing.T) {
m := testModule(t, "validate-bad-rc")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_tainted(t *testing.T) {
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
"num": {Type: cty.String, Optional: true},
},
},
},
},
}
m := testModule(t, "validate-good")
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
p.ValidateResourceConfigFn = func(req providers.ValidateResourceConfigRequest) providers.ValidateResourceConfigResponse {
var diags tfdiags.Diagnostics
if req.Config.GetAttr("foo").IsNull() {
diags = diags.Append(errors.New("foo is not set"))
}
return providers.ValidateResourceConfigResponse{
Diagnostics: diags,
}
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_targetedDestroy(t *testing.T) {
m := testModule(t, "validate-targeted")
p := testProvider("aws")
pr := simpleMockProvisioner()
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
"num": {Type: cty.String, Optional: true},
},
},
},
},
}
state := states.NewState()
root := state.EnsureModule(addrs.RootModuleInstance)
testSetResourceInstanceCurrent(root, "aws_instance.foo", `{"id":"i-bcd345"}`, `provider["registry.terraform.io/hashicorp/aws"]`)
testSetResourceInstanceCurrent(root, "aws_instance.bar", `{"id":"i-abc123"}`, `provider["registry.terraform.io/hashicorp/aws"]`)
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-11-13 18:43:28 +01:00
Provisioners: map[string]provisioners.Factory{
"shell": testProvisionerFuncFixed(pr),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_varRefUnknown(t *testing.T) {
m := testModule(t, "validate-variable-ref")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
var value cty.Value
p.ValidateResourceConfigFn = func(req providers.ValidateResourceConfigRequest) providers.ValidateResourceConfigResponse {
value = req.Config.GetAttr("foo")
return providers.ValidateResourceConfigResponse{}
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
c.Validate(m)
// Input variables are always unknown during the validate walk, because
// we're checking for validity of all possible input values. Validity
// against specific input values is checked during the plan walk.
if !value.RawEquals(cty.UnknownVal(cty.String)) {
t.Fatalf("bad: %#v", value)
}
}
// Module variables weren't being interpolated during Validate phase.
// related to https://github.com/hashicorp/terraform/issues/5322
func TestContext2Validate_interpolateVar(t *testing.T) {
input := new(MockUIInput)
m := testModule(t, "input-interpolate-var")
p := testProvider("null")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"template_file": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"template": {Type: cty.String, Optional: true},
},
},
},
},
}
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("template"): testProviderFuncFixed(p),
},
UIInput: input,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
// When module vars reference something that is actually computed, this
// shouldn't cause validation to fail.
func TestContext2Validate_interpolateComputedModuleVarDef(t *testing.T) {
input := new(MockUIInput)
m := testModule(t, "validate-computed-module-var-ref")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"attr": {Type: cty.String, Optional: true},
},
},
},
},
}
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
UIInput: input,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
// Computed values are lost when a map is output from a module
func TestContext2Validate_interpolateMap(t *testing.T) {
input := new(MockUIInput)
m := testModule(t, "issue-9549")
p := testProvider("template")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("template"): testProviderFuncFixed(p),
},
UIInput: input,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
}
func TestContext2Validate_varSensitive(t *testing.T) {
// Smoke test through validate where a variable has sensitive applied
m := testModuleInline(t, map[string]string{
"main.tf": `
variable "foo" {
default = "xyz"
sensitive = true
}
variable "bar" {
sensitive = true
}
data "aws_data_source" "bar" {
foo = var.bar
}
resource "aws_instance" "foo" {
foo = var.foo
}
`,
})
p := testProvider("aws")
p.ValidateResourceConfigFn = func(req providers.ValidateResourceConfigRequest) providers.ValidateResourceConfigResponse {
// Providers receive unmarked values
if got, want := req.Config.GetAttr("foo"), cty.UnknownVal(cty.String); !got.RawEquals(want) {
t.Fatalf("wrong value for foo\ngot: %#v\nwant: %#v", got, want)
}
return providers.ValidateResourceConfigResponse{}
}
p.ValidateDataResourceConfigFn = func(req providers.ValidateDataResourceConfigRequest) (resp providers.ValidateDataResourceConfigResponse) {
if got, want := req.Config.GetAttr("foo"), cty.UnknownVal(cty.String); !got.RawEquals(want) {
t.Fatalf("wrong value for foo\ngot: %#v\nwant: %#v", got, want)
}
return providers.ValidateDataResourceConfigResponse{}
}
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatal(diags.Err())
}
if !p.ValidateResourceConfigCalled {
t.Fatal("expected ValidateResourceConfigFn to be called")
}
if !p.ValidateDataResourceConfigCalled {
t.Fatal("expected ValidateDataSourceConfigFn to be called")
}
}
// Manually validate using the new PlanGraphBuilder
func TestContext2Validate_PlanGraphBuilder(t *testing.T) {
fixture := contextFixtureApplyVars(t)
opts := fixture.ContextOpts()
c := testContext2(t, opts)
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
graph, diags := ValidateGraphBuilder(&PlanGraphBuilder{
Config: fixture.Config,
State: states.NewState(),
Plugins: c.plugins,
}).Build(addrs.RootModuleInstance)
if diags.HasErrors() {
t.Fatalf("errors from PlanGraphBuilder: %s", diags.Err())
}
defer c.acquireRun("validate-test")()
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
walker, diags := c.walk(graph, walkValidate, &graphWalkOpts{
core: Graph walk loads plugin schemas opportunistically Previously our graph walker expected to recieve a data structure containing schemas for all of the provider and provisioner plugins used in the configuration and state. That made sense back when terraform.NewContext was responsible for loading all of the schemas before taking any other action, but it no longer has that responsiblity. Instead, we'll now make sure that the "contextPlugins" object reaches all of the locations where we need schema -- many of which already had access to that object anyway -- and then load the needed schemas just in time. The contextPlugins object memoizes schema lookups, so we can safely call it many times with the same provider address or provisioner type name and know that it'll still only load each distinct plugin once per Context object. As of this commit, the Context.Schemas method is now a public interface only and not used by logic in the "terraform" package at all. However, that does leave us in a rather tenuous situation of relying on the fact that all practical users of terraform.Context end up calling "Schemas" at some point in order to verify that we have all of the expected versions of plugins. That's a non-obvious implicit dependency, and so in subsequent commits we'll gradually move all responsibility for verifying plugin versions into the caller of terraform.NewContext, which'll heal a long-standing architectural wart whereby the caller is responsible for installing and locating the plugin executables but not for verifying that what's installed is conforming to the current configuration and dependency lock file.
2021-09-01 02:53:03 +02:00
Config: fixture.Config,
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
})
if diags.HasErrors() {
t.Fatal(diags.Err())
}
if len(walker.NonFatalDiagnostics) > 0 {
t.Fatal(walker.NonFatalDiagnostics.Err())
}
}
func TestContext2Validate_invalidOutput(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
data "aws_data_source" "name" {}
output "out" {
value = "${data.aws_data_source.name.missing}"
}`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
// Should get this error:
// Unsupported attribute: This object does not have an attribute named "missing"
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Unsupported attribute"; !strings.Contains(got, want) {
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_invalidModuleOutput(t *testing.T) {
m := testModuleInline(t, map[string]string{
"child/main.tf": `
data "aws_data_source" "name" {}
output "out" {
value = "${data.aws_data_source.name.missing}"
}`,
"main.tf": `
module "child" {
source = "./child"
}
resource "aws_instance" "foo" {
foo = "${module.child.out}"
}`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
// Should get this error:
// Unsupported attribute: This object does not have an attribute named "missing"
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Unsupported attribute"; !strings.Contains(got, want) {
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_sensitiveRootModuleOutput(t *testing.T) {
m := testModuleInline(t, map[string]string{
"child/main.tf": `
variable "foo" {
default = "xyz"
sensitive = true
}
output "out" {
value = var.foo
}`,
"main.tf": `
module "child" {
source = "./child"
}
output "root" {
value = module.child.out
sensitive = true
}`,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
ctx := testContext2(t, &ContextOpts{})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatal(diags.Err())
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
}
}
func TestContext2Validate_legacyResourceCount(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
resource "aws_instance" "test" {}
output "out" {
value = aws_instance.test.count
}`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
// Should get this error:
// Invalid resource count attribute: The special "count" attribute is no longer supported after Terraform v0.12. Instead, use length(aws_instance.test) to count resource instances.
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Invalid resource count attribute:"; !strings.Contains(got, want) {
core: Static-validate resource references against schemas In the initial move to HCL2 we started relying only on full expression evaluation to catch attribute errors, but that's not sufficient for resource attributes in practice because during validation we can't know yet whether a resource reference evaluates to a single object or to a list of objects (if count is set). To address this, here we reinstate some static validation of resource references by analyzing directly the reference objects, disregarding any instance index if present, and produce errors if the remaining subsequent traversal steps do not correspond to items within the resource type schema. This also allows us to produce some more specialized error messages for certain situations. In particular, we can recognize a reference like aws_instance.foo.count, which in 0.11 and prior was a weird special case for determining the count value of a resource block, and offer a helpful error showing the new length(aws_instance.foo) usage pattern. This eventually delegates to the static traversal validation logic that was added to the configschema package in a previous commit, which also includes some specialized error messages that distinguish between attributes and block types in the schema so that the errors relate more directly to constructs the user can see in the configuration. In future we could potentially move more of the checks from the dynamic schema construction step to the static validation step, but resources are the reference type that most needs this immediately due to the ambiguity caused by the instance indexing syntax. We can safely refactor other reference types to be statically validated in later releases. This is verified by two pre-existing context validate tests which we temporarily disabled during earlier work (now re-enabled) and also by a new validate test aimed specifically at the special case for the "count" attribute.
2018-11-21 02:25:05 +01:00
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_invalidModuleRef(t *testing.T) {
// This test is verifying that we properly validate and report on references
// to modules that are not declared, since we were missing some validation
// here in early 0.12.0 alphas that led to a panic.
m := testModuleInline(t, map[string]string{
"main.tf": `
output "out" {
# Intentionally referencing undeclared module to ensure error
value = module.foo
}`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
// Should get this error:
// Reference to undeclared module: No module call named "foo" is declared in the root module.
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Reference to undeclared module:"; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_invalidModuleOutputRef(t *testing.T) {
// This test is verifying that we properly validate and report on references
// to modules that are not declared, since we were missing some validation
// here in early 0.12.0 alphas that led to a panic.
m := testModuleInline(t, map[string]string{
"main.tf": `
output "out" {
# Intentionally referencing undeclared module to ensure error
value = module.foo.bar
}`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
// Should get this error:
// Reference to undeclared module: No module call named "foo" is declared in the root module.
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Reference to undeclared module:"; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_invalidDependsOnResourceRef(t *testing.T) {
// This test is verifying that we raise an error if depends_on
// refers to something that doesn't exist in configuration.
m := testModuleInline(t, map[string]string{
"main.tf": `
resource "test_instance" "bar" {
depends_on = [test_resource.nonexistant]
}
`,
})
p := testProvider("test")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
// Should get this error:
// Reference to undeclared module: No module call named "foo" is declared in the root module.
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), "Reference to undeclared resource:"; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_invalidResourceIgnoreChanges(t *testing.T) {
// This test is verifying that we raise an error if ignore_changes
// refers to something that can be statically detected as not conforming
// to the resource type schema.
m := testModuleInline(t, map[string]string{
"main.tf": `
resource "test_instance" "bar" {
lifecycle {
ignore_changes = [does_not_exist_in_schema]
}
}
`,
})
p := testProvider("test")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
// Should get this error:
// Reference to undeclared module: No module call named "foo" is declared in the root module.
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), `no argument, nested block, or exported attribute named "does_not_exist_in_schema"`; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_variableCustomValidationsFail(t *testing.T) {
// This test is for custom validation rules associated with root module
// variables, and specifically that we handle the situation where the
// given value is invalid in a child module.
m := testModule(t, "validate-variable-custom-validations-child")
p := testProvider("test")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), `Invalid value for variable: Value must not be "nope".`; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_variableCustomValidationsRoot(t *testing.T) {
// This test is for custom validation rules associated with root module
// variables, and specifically that we handle the situation where their
// values are unknown during validation, skipping the validation check
// altogether. (Root module variables are never known during validation.)
m := testModuleInline(t, map[string]string{
"main.tf": `
variable "test" {
type = string
validation {
condition = var.test != "nope"
error_message = "Value must not be \"nope\"."
}
}
`,
})
p := testProvider("test")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error\ngot: %s", diags.Err().Error())
}
}
2020-04-05 17:17:28 +02:00
func TestContext2Validate_expandModules(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
module "mod1" {
for_each = toset(["a", "b"])
source = "./mod"
}
module "mod2" {
for_each = module.mod1
source = "./mod"
input = module.mod1["a"].out
2020-04-05 17:17:28 +02:00
}
module "mod3" {
count = length(module.mod2)
2020-04-05 17:17:28 +02:00
source = "./mod"
}
`,
"mod/main.tf": `
resource "aws_instance" "foo" {
}
output "out" {
value = 1
}
variable "input" {
type = number
default = 0
}
2020-04-05 17:17:28 +02:00
module "nested" {
count = 2
source = "./nested"
input = count.index
2020-04-05 17:17:28 +02:00
}
`,
"mod/nested/main.tf": `
variable "input" {
}
resource "aws_instance" "foo" {
count = var.input
}
`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
2020-04-05 17:17:28 +02:00
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
2020-04-05 17:17:28 +02:00
if diags.HasErrors() {
t.Fatal(diags.ErrWithWarnings())
}
}
func TestContext2Validate_expandModulesInvalidCount(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
module "mod1" {
count = -1
source = "./mod"
}
`,
"mod/main.tf": `
resource "aws_instance" "foo" {
}
`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), `Invalid count argument`; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_expandModulesInvalidForEach(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
module "mod1" {
for_each = ["a", "b"]
source = "./mod"
}
`,
"mod/main.tf": `
resource "aws_instance" "foo" {
}
`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
2020-12-01 21:51:54 +01:00
if got, want := diags.Err().Error(), `Invalid for_each argument`; !strings.Contains(got, want) {
t.Fatalf("wrong error:\ngot: %s\nwant: message containing %q", got, want)
}
}
func TestContext2Validate_expandMultipleNestedModules(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
module "modA" {
for_each = {
first = "m"
second = "n"
}
source = "./modA"
}
`,
"modA/main.tf": `
locals {
m = {
first = "m"
second = "n"
}
}
module "modB" {
for_each = local.m
source = "./modB"
y = each.value
}
module "modC" {
for_each = local.m
source = "./modC"
x = module.modB[each.key].out
y = module.modB[each.key].out
}
`,
"modA/modB/main.tf": `
variable "y" {
type = string
}
resource "aws_instance" "foo" {
foo = var.y
}
output "out" {
value = aws_instance.foo.id
}
`,
"modA/modC/main.tf": `
variable "x" {
type = string
}
variable "y" {
type = string
}
resource "aws_instance" "foo" {
foo = var.x
}
output "out" {
value = var.y
}
`,
})
p := testProvider("aws")
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatal(diags.ErrWithWarnings())
}
}
func TestContext2Validate_invalidModuleDependsOn(t *testing.T) {
// validate module and output depends_on
m := testModuleInline(t, map[string]string{
"main.tf": `
module "mod1" {
source = "./mod"
depends_on = [resource_foo.bar.baz]
}
module "mod2" {
source = "./mod"
depends_on = [resource_foo.bar.baz]
}
`,
"mod/main.tf": `
output "out" {
value = "foo"
}
`,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := testContext2(t, &ContextOpts{}).Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
if len(diags) != 2 {
t.Fatalf("wanted 2 diagnostic errors, got %q", diags)
}
for _, d := range diags {
des := d.Description().Summary
if !strings.Contains(des, "Invalid depends_on reference") {
t.Fatalf(`expected "Invalid depends_on reference", got %q`, des)
}
}
}
func TestContext2Validate_invalidOutputDependsOn(t *testing.T) {
// validate module and output depends_on
m := testModuleInline(t, map[string]string{
"main.tf": `
module "mod1" {
source = "./mod"
}
output "out" {
value = "bar"
depends_on = [resource_foo.bar.baz]
}
`,
"mod/main.tf": `
output "out" {
value = "bar"
depends_on = [resource_foo.bar.baz]
}
`,
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := testContext2(t, &ContextOpts{}).Validate(m)
if !diags.HasErrors() {
t.Fatal("succeeded; want errors")
}
if len(diags) != 2 {
t.Fatalf("wanted 2 diagnostic errors, got %q", diags)
}
for _, d := range diags {
des := d.Description().Summary
if !strings.Contains(des, "Invalid depends_on reference") {
t.Fatalf(`expected "Invalid depends_on reference", got %q`, des)
}
}
}
2020-10-28 19:51:04 +01:00
func TestContext2Validate_rpcDiagnostics(t *testing.T) {
// validate module and output depends_on
m := testModuleInline(t, map[string]string{
"main.tf": `
resource "test_instance" "a" {
}
`,
})
p := testProvider("test")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
2020-10-28 19:51:04 +01:00
"test_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"id": {Type: cty.String, Computed: true},
},
2020-10-28 19:51:04 +01:00
},
},
},
}
p.ValidateResourceConfigResponse = &providers.ValidateResourceConfigResponse{
2020-10-28 19:51:04 +01:00
Diagnostics: tfdiags.Diagnostics(nil).Append(tfdiags.SimpleWarning("don't frobble")),
}
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
2020-10-28 19:51:04 +01:00
if diags.HasErrors() {
t.Fatal(diags.Err())
}
if len(diags) == 0 {
t.Fatal("expected warnings")
}
2020-10-28 19:51:04 +01:00
for _, d := range diags {
des := d.Description().Summary
if !strings.Contains(des, "frobble") {
t.Fatalf(`expected frobble, got %q`, des)
}
}
}
func TestContext2Validate_sensitiveProvisionerConfig(t *testing.T) {
m := testModule(t, "validate-sensitive-provisioner-config")
p := testProvider("aws")
p.GetProviderSchemaResponse = &providers.GetProviderSchemaResponse{
ResourceTypes: map[string]providers.Schema{
"aws_instance": {
Block: &configschema.Block{
Attributes: map[string]*configschema.Attribute{
"foo": {Type: cty.String, Optional: true},
},
},
},
},
}
pr := simpleMockProvisioner()
c := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("aws"): testProviderFuncFixed(p),
},
Provisioners: map[string]provisioners.Factory{
"test": testProvisionerFuncFixed(pr),
},
})
pr.ValidateProvisionerConfigFn = func(r provisioners.ValidateProvisionerConfigRequest) provisioners.ValidateProvisionerConfigResponse {
if r.Config.ContainsMarked() {
t.Errorf("provisioner config contains marked values")
}
return pr.ValidateProvisionerConfigResponse
}
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := c.Validate(m)
if diags.HasErrors() {
t.Fatalf("unexpected error: %s", diags.Err())
}
if !pr.ValidateProvisionerConfigCalled {
t.Fatal("ValidateProvisionerConfig not called")
}
}
func TestContext2Plan_validateMinMaxDynamicBlock(t *testing.T) {
p := new(MockProvider)
p.GetProviderSchemaResponse = getProviderSchemaResponseFromProviderSchema(&ProviderSchema{
ResourceTypes: map[string]*configschema.Block{
"test_instance": {
Attributes: map[string]*configschema.Attribute{
"id": {
Type: cty.String,
Computed: true,
},
"things": {
Type: cty.List(cty.String),
Computed: true,
},
},
BlockTypes: map[string]*configschema.NestedBlock{
"foo": {
Block: configschema.Block{
Attributes: map[string]*configschema.Attribute{
"bar": {Type: cty.String, Optional: true},
},
},
Nesting: configschema.NestingList,
MinItems: 2,
MaxItems: 3,
},
},
},
},
})
m := testModuleInline(t, map[string]string{
"main.tf": `
resource "test_instance" "a" {
// MinItems 2
foo {
bar = "a"
}
foo {
bar = "b"
}
}
resource "test_instance" "b" {
// one dymamic block can satisfy MinItems of 2
dynamic "foo" {
for_each = test_instance.a.things
content {
bar = foo.value
}
}
}
resource "test_instance" "c" {
// we may have more than MaxItems dynamic blocks when they are unknown
foo {
bar = "b"
}
dynamic "foo" {
for_each = test_instance.a.things
content {
bar = foo.value
}
}
dynamic "foo" {
for_each = test_instance.a.things
content {
bar = "${foo.value}-2"
}
}
dynamic "foo" {
for_each = test_instance.b.things
content {
bar = foo.value
}
}
}
`})
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatal(diags.ErrWithWarnings())
}
}
func TestContext2Validate_passInheritedProvider(t *testing.T) {
m := testModuleInline(t, map[string]string{
"main.tf": `
terraform {
required_providers {
test = {
source = "hashicorp/test"
}
}
}
module "first" {
source = "./first"
providers = {
test = test
}
}
`,
// This module does not define a config for the test provider, but we
// should be able to pass whatever the implied config is to a child
// module.
"first/main.tf": `
terraform {
required_providers {
test = {
source = "hashicorp/test"
}
}
}
module "second" {
source = "./second"
providers = {
test.alias = test
}
}`,
"first/second/main.tf": `
terraform {
required_providers {
test = {
source = "hashicorp/test"
configuration_aliases = [test.alias]
}
}
}
resource "test_object" "t" {
provider = test.alias
}
`,
})
p := simpleMockProvider()
ctx := testContext2(t, &ContextOpts{
Providers: map[addrs.Provider]providers.Factory{
addrs.NewDefaultProvider("test"): testProviderFuncFixed(p),
},
})
core: Functional-style API for terraform.Context Previously terraform.Context was built in an unfortunate way where all of the data was provided up front in terraform.NewContext and then mutated directly by subsequent operations. That made the data flow hard to follow, commonly leading to bugs, and also meant that we were forced to take various actions too early in terraform.NewContext, rather than waiting until a more appropriate time during an operation. This (enormous) commit changes terraform.Context so that its fields are broadly just unchanging data about the execution context (current workspace name, available plugins, etc) whereas the main data Terraform works with arrives via individual method arguments and is returned in return values. Specifically, this means that terraform.Context no longer "has-a" config, state, and "planned changes", instead holding on to those only temporarily during an operation. The caller is responsible for propagating the outcome of one step into the next step so that the data flow between operations is actually visible. However, since that's a change to the main entry points in the "terraform" package, this commit also touches every file in the codebase which interacted with those APIs. Most of the noise here is in updating tests to take the same actions using the new API style, but this also affects the main-code callers in the backends and in the command package. My goal here was to refactor without changing observable behavior, but in practice there are a couple externally-visible behavior variations here that seemed okay in service of the broader goal: - The "terraform graph" command is no longer hooked directly into the core graph builders, because that's no longer part of the public API. However, I did include a couple new Context functions whose contract is to produce a UI-oriented graph, and _for now_ those continue to return the physical graph we use for those operations. There's no exported API for generating the "validate" and "eval" graphs, because neither is particularly interesting in its own right, and so "terraform graph" no longer supports those graph types. - terraform.NewContext no longer has the responsibility for collecting all of the provider schemas up front. Instead, we wait until we need them. However, that means that some of our error messages now have a slightly different shape due to unwinding through a differently-shaped call stack. As of this commit we also end up reloading the schemas multiple times in some cases, which is functionally acceptable but likely represents a performance regression. I intend to rework this to use caching, but I'm saving that for a later commit because this one is big enough already. The proximal reason for this change is to resolve the chicken/egg problem whereby there was previously no single point where we could apply "moved" statements to the previous run state before creating a plan. With this change in place, we can now do that as part of Context.Plan, prior to forking the input state into the three separate state artifacts we use during planning. However, this is at least the third project in a row where the previous API design led to piling more functionality into terraform.NewContext and then working around the incorrect order of operations that produces, so I intend that by paying the cost/risk of this large diff now we can in turn reduce the cost/risk of future projects that relate to our main workflow actions.
2021-08-24 21:06:38 +02:00
diags := ctx.Validate(m)
if diags.HasErrors() {
t.Fatal(diags.ErrWithWarnings())
}
}