terraform/internal/command/format/diff.go

1966 lines
60 KiB
Go
Raw Normal View History

package format
import (
"bufio"
"bytes"
"fmt"
backend/local: treat output changes as side-effects to be applied This is a baby-step towards an intended future where all Terraform actions which have side-effects in either remote objects or the Terraform state can go through the plan+apply workflow. This initial change is focused only on allowing plan+apply for changes to root module output values, so that these can be written into a new state snapshot (for consumption by terraform_remote_state elsewhere) without having to go outside of the primary workflow by running "terraform refresh". This is also better than "terraform refresh" because it gives an opportunity to review the proposed changes before applying them, as we're accustomed to with resource changes. The downside here is that Terraform Core was not designed to produce accurate changesets for root module outputs. Although we added a place for it in the plan model in Terraform 0.12, Terraform Core currently produces inaccurate changesets there which don't properly track the prior values. We're planning to rework Terraform Core's evaluation approach in a forthcoming release so it would itself be able to distinguish between the prior state and the planned new state to produce an accurate changeset, but this commit introduces a temporary stop-gap solution of implementing the logic up in the local backend code, where we can freeze a snapshot of the prior state before we take any other actions and then use that to produce an accurate output changeset to decide whether the plan has externally-visible side-effects and render any changes to output values. This temporary approach should be replaced by a more appropriately-placed solution in Terraform Core in a release, which should then allow further behaviors in similar vein, such as user-visible drift detection for resource instances.
2020-05-27 01:59:06 +02:00
"log"
"sort"
"strings"
"github.com/mitchellh/colorstring"
"github.com/zclconf/go-cty/cty"
ctyjson "github.com/zclconf/go-cty/cty/json"
"github.com/hashicorp/terraform/internal/addrs"
"github.com/hashicorp/terraform/internal/configs/configschema"
"github.com/hashicorp/terraform/internal/lang/marks"
"github.com/hashicorp/terraform/internal/plans"
"github.com/hashicorp/terraform/internal/plans/objchange"
"github.com/hashicorp/terraform/internal/states"
)
2021-09-13 23:30:16 +02:00
// DiffLanguage controls the description of the resource change reasons.
type DiffLanguage rune
//go:generate go run golang.org/x/tools/cmd/stringer -type=DiffLanguage diff.go
const (
// DiffLanguageProposedChange indicates that the change is one which is
// planned to be applied.
DiffLanguageProposedChange DiffLanguage = 'P'
// DiffLanguageDetectedDrift indicates that the change is detected drift
// from the configuration.
DiffLanguageDetectedDrift DiffLanguage = 'D'
)
// ResourceChange returns a string representation of a change to a particular
// resource, for inclusion in user-facing plan output.
//
// The resource schema must be provided along with the change so that the
// formatted change can reflect the configuration structure for the associated
// resource.
//
// If "color" is non-nil, it will be used to color the result. Otherwise,
// no color codes will be included.
func ResourceChange(
change *plans.ResourceInstanceChangeSrc,
schema *configschema.Block,
color *colorstring.Colorize,
2021-09-13 23:30:16 +02:00
language DiffLanguage,
) string {
addr := change.Addr
var buf bytes.Buffer
if color == nil {
color = &colorstring.Colorize{
Colors: colorstring.DefaultColors,
Disable: true,
Reset: false,
}
}
dispAddr := addr.String()
if change.DeposedKey != states.NotDeposed {
dispAddr = fmt.Sprintf("%s (deposed object %s)", dispAddr, change.DeposedKey)
}
switch change.Action {
case plans.Create:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] will be created"), dispAddr))
case plans.Read:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] will be read during apply\n # (config refers to values not yet known)"), dispAddr))
case plans.Update:
2021-09-13 23:30:16 +02:00
switch language {
case DiffLanguageProposedChange:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] will be updated in-place"), dispAddr))
2021-09-13 23:30:16 +02:00
case DiffLanguageDetectedDrift:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] has changed"), dispAddr))
2021-09-13 23:30:16 +02:00
default:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] update (unknown reason %s)"), dispAddr, language))
2021-09-13 23:30:16 +02:00
}
case plans.CreateThenDelete, plans.DeleteThenCreate:
switch change.ActionReason {
case plans.ResourceInstanceReplaceBecauseTainted:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] is tainted, so must be [bold][red]replaced"), dispAddr))
case plans.ResourceInstanceReplaceByRequest:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] will be [bold][red]replaced[reset], as requested"), dispAddr))
default:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] must be [bold][red]replaced"), dispAddr))
}
case plans.Delete:
2021-09-13 23:30:16 +02:00
switch language {
case DiffLanguageProposedChange:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] will be [bold][red]destroyed"), dispAddr))
2021-09-13 23:30:16 +02:00
case DiffLanguageDetectedDrift:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] has been deleted"), dispAddr))
2021-09-13 23:30:16 +02:00
default:
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] delete (unknown reason %s)"), dispAddr, language))
2021-09-13 23:30:16 +02:00
}
if change.DeposedKey != states.NotDeposed {
// Some extra context about this unusual situation.
2021-08-31 23:33:26 +02:00
buf.WriteString(color.Color("\n # (left over from a partially-failed replacement of this instance)"))
}
case plans.NoOp:
if change.Moved() {
buf.WriteString(fmt.Sprintf(color.Color("[bold] # %s[reset] has moved to [bold]%s[reset]"), change.PrevRunAddr.String(), dispAddr))
break
}
fallthrough
default:
// should never happen, since the above is exhaustive
buf.WriteString(fmt.Sprintf("%s has an action the plan renderer doesn't support (this is a bug)", dispAddr))
}
buf.WriteString(color.Color("[reset]\n"))
if change.Moved() && change.Action != plans.NoOp {
buf.WriteString(fmt.Sprintf(color.Color("[bold] # [reset]([bold]%s[reset] has moved to [bold]%s[reset])\n"), change.PrevRunAddr.String(), dispAddr))
}
if change.Moved() && change.Action == plans.NoOp {
buf.WriteString(" ")
} else {
buf.WriteString(color.Color(DiffActionSymbol(change.Action)) + " ")
}
switch addr.Resource.Resource.Mode {
case addrs.ManagedResourceMode:
buf.WriteString(fmt.Sprintf(
"resource %q %q",
addr.Resource.Resource.Type,
addr.Resource.Resource.Name,
))
case addrs.DataResourceMode:
buf.WriteString(fmt.Sprintf(
"data %q %q ",
addr.Resource.Resource.Type,
addr.Resource.Resource.Name,
))
default:
// should never happen, since the above is exhaustive
buf.WriteString(addr.String())
}
buf.WriteString(" {")
p := blockBodyDiffPrinter{
buf: &buf,
color: color,
action: change.Action,
requiredReplace: change.RequiredReplace,
}
// Most commonly-used resources have nested blocks that result in us
// going at least three traversals deep while we recurse here, so we'll
// start with that much capacity and then grow as needed for deeper
// structures.
path := make(cty.Path, 0, 3)
changeV, err := change.Decode(schema.ImpliedType())
if err != nil {
// Should never happen in here, since we've already been through
// loads of layers of encode/decode of the planned changes before now.
panic(fmt.Sprintf("failed to decode plan for %s while rendering diff: %s", addr, err))
}
// We currently have an opt-out that permits the legacy SDK to return values
// that defy our usual conventions around handling of nesting blocks. To
// avoid the rendering code from needing to handle all of these, we'll
// normalize first.
// (Ideally we'd do this as part of the SDK opt-out implementation in core,
// but we've added it here for now to reduce risk of unexpected impacts
// on other code in core.)
changeV.Change.Before = objchange.NormalizeObjectFromLegacySDK(changeV.Change.Before, schema)
changeV.Change.After = objchange.NormalizeObjectFromLegacySDK(changeV.Change.After, schema)
2020-09-10 17:06:40 +02:00
result := p.writeBlockBodyDiff(schema, changeV.Before, changeV.After, 6, path)
if result.bodyWritten {
buf.WriteString("\n")
buf.WriteString(strings.Repeat(" ", 4))
}
buf.WriteString("}\n")
return buf.String()
}
backend/local: treat output changes as side-effects to be applied This is a baby-step towards an intended future where all Terraform actions which have side-effects in either remote objects or the Terraform state can go through the plan+apply workflow. This initial change is focused only on allowing plan+apply for changes to root module output values, so that these can be written into a new state snapshot (for consumption by terraform_remote_state elsewhere) without having to go outside of the primary workflow by running "terraform refresh". This is also better than "terraform refresh" because it gives an opportunity to review the proposed changes before applying them, as we're accustomed to with resource changes. The downside here is that Terraform Core was not designed to produce accurate changesets for root module outputs. Although we added a place for it in the plan model in Terraform 0.12, Terraform Core currently produces inaccurate changesets there which don't properly track the prior values. We're planning to rework Terraform Core's evaluation approach in a forthcoming release so it would itself be able to distinguish between the prior state and the planned new state to produce an accurate changeset, but this commit introduces a temporary stop-gap solution of implementing the logic up in the local backend code, where we can freeze a snapshot of the prior state before we take any other actions and then use that to produce an accurate output changeset to decide whether the plan has externally-visible side-effects and render any changes to output values. This temporary approach should be replaced by a more appropriately-placed solution in Terraform Core in a release, which should then allow further behaviors in similar vein, such as user-visible drift detection for resource instances.
2020-05-27 01:59:06 +02:00
// OutputChanges returns a string representation of a set of changes to output
// values for inclusion in user-facing plan output.
//
// If "color" is non-nil, it will be used to color the result. Otherwise,
// no color codes will be included.
func OutputChanges(
changes []*plans.OutputChangeSrc,
color *colorstring.Colorize,
) string {
var buf bytes.Buffer
p := blockBodyDiffPrinter{
buf: &buf,
color: color,
action: plans.Update, // not actually used in this case, because we're not printing a containing block
backend/local: treat output changes as side-effects to be applied This is a baby-step towards an intended future where all Terraform actions which have side-effects in either remote objects or the Terraform state can go through the plan+apply workflow. This initial change is focused only on allowing plan+apply for changes to root module output values, so that these can be written into a new state snapshot (for consumption by terraform_remote_state elsewhere) without having to go outside of the primary workflow by running "terraform refresh". This is also better than "terraform refresh" because it gives an opportunity to review the proposed changes before applying them, as we're accustomed to with resource changes. The downside here is that Terraform Core was not designed to produce accurate changesets for root module outputs. Although we added a place for it in the plan model in Terraform 0.12, Terraform Core currently produces inaccurate changesets there which don't properly track the prior values. We're planning to rework Terraform Core's evaluation approach in a forthcoming release so it would itself be able to distinguish between the prior state and the planned new state to produce an accurate changeset, but this commit introduces a temporary stop-gap solution of implementing the logic up in the local backend code, where we can freeze a snapshot of the prior state before we take any other actions and then use that to produce an accurate output changeset to decide whether the plan has externally-visible side-effects and render any changes to output values. This temporary approach should be replaced by a more appropriately-placed solution in Terraform Core in a release, which should then allow further behaviors in similar vein, such as user-visible drift detection for resource instances.
2020-05-27 01:59:06 +02:00
}
// We're going to reuse the codepath we used for printing resource block
// diffs, by pretending that the set of defined outputs are the attributes
// of some resource. It's a little forced to do this, but it gives us all
// the same formatting heuristics as we normally use for resource
// attributes.
oldVals := make(map[string]cty.Value, len(changes))
newVals := make(map[string]cty.Value, len(changes))
synthSchema := &configschema.Block{
Attributes: make(map[string]*configschema.Attribute, len(changes)),
}
for _, changeSrc := range changes {
name := changeSrc.Addr.OutputValue.Name
change, err := changeSrc.Decode()
if err != nil {
// It'd be weird to get a decoding error here because that would
// suggest that Terraform itself just produced an invalid plan, and
// we don't have any good way to ignore it in this codepath, so
// we'll just log it and ignore it.
log.Printf("[ERROR] format.OutputChanges: Failed to decode planned change for output %q: %s", name, err)
continue
}
synthSchema.Attributes[name] = &configschema.Attribute{
Type: cty.DynamicPseudoType, // output types are decided dynamically based on the given value
Optional: true,
Sensitive: change.Sensitive,
}
oldVals[name] = change.Before
newVals[name] = change.After
}
p.writeBlockBodyDiff(synthSchema, cty.ObjectVal(oldVals), cty.ObjectVal(newVals), 2, nil)
return buf.String()
}
type blockBodyDiffPrinter struct {
buf *bytes.Buffer
color *colorstring.Colorize
action plans.Action
requiredReplace cty.PathSet
// verbose is set to true when using the "diff" printer to format state
verbose bool
}
type blockBodyDiffResult struct {
bodyWritten bool
skippedAttributes int
skippedBlocks int
}
const forcesNewResourceCaption = " [red]# forces replacement[reset]"
// writeBlockBodyDiff writes attribute or block differences
// and returns true if any differences were found and written
func (p *blockBodyDiffPrinter) writeBlockBodyDiff(schema *configschema.Block, old, new cty.Value, indent int, path cty.Path) blockBodyDiffResult {
path = ctyEnsurePathCapacity(path, 1)
result := blockBodyDiffResult{}
// write the attributes diff
blankBeforeBlocks := p.writeAttrsDiff(schema.Attributes, old, new, indent, path, &result)
p.writeSkippedAttr(result.skippedAttributes, indent+2)
{
blockTypeNames := make([]string, 0, len(schema.BlockTypes))
for name := range schema.BlockTypes {
blockTypeNames = append(blockTypeNames, name)
}
sort.Strings(blockTypeNames)
for _, name := range blockTypeNames {
blockS := schema.BlockTypes[name]
oldVal := ctyGetAttrMaybeNull(old, name)
newVal := ctyGetAttrMaybeNull(new, name)
result.bodyWritten = true
skippedBlocks := p.writeNestedBlockDiffs(name, blockS, oldVal, newVal, blankBeforeBlocks, indent, path)
if skippedBlocks > 0 {
result.skippedBlocks += skippedBlocks
}
// Always include a blank for any subsequent block types.
blankBeforeBlocks = true
}
if result.skippedBlocks > 0 {
noun := "blocks"
if result.skippedBlocks == 1 {
noun = "block"
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), result.skippedBlocks, noun))
}
}
return result
}
func (p *blockBodyDiffPrinter) writeAttrsDiff(
attrsS map[string]*configschema.Attribute,
old, new cty.Value,
indent int,
path cty.Path,
result *blockBodyDiffResult) bool {
blankBeforeBlocks := false
attrNames := make([]string, 0, len(attrsS))
attrNameLen := 0
for name := range attrsS {
oldVal := ctyGetAttrMaybeNull(old, name)
newVal := ctyGetAttrMaybeNull(new, name)
if oldVal.IsNull() && newVal.IsNull() {
// Skip attributes where both old and new values are null
// (we do this early here so that we'll do our value alignment
// based on the longest attribute name that has a change, rather
// than the longest attribute name in the full set.)
continue
}
attrNames = append(attrNames, name)
if len(name) > attrNameLen {
attrNameLen = len(name)
}
}
sort.Strings(attrNames)
if len(attrNames) > 0 {
blankBeforeBlocks = true
}
for _, name := range attrNames {
attrS := attrsS[name]
oldVal := ctyGetAttrMaybeNull(old, name)
newVal := ctyGetAttrMaybeNull(new, name)
result.bodyWritten = true
skipped := p.writeAttrDiff(name, attrS, oldVal, newVal, attrNameLen, indent, path)
if skipped {
result.skippedAttributes++
}
}
return blankBeforeBlocks
}
// getPlanActionAndShow returns the action value
// and a boolean for showJustNew. In this function we
// modify the old and new values to remove any possible marks
func getPlanActionAndShow(old cty.Value, new cty.Value) (plans.Action, bool) {
var action plans.Action
showJustNew := false
switch {
case old.IsNull():
action = plans.Create
showJustNew = true
case new.IsNull():
action = plans.Delete
case ctyEqualWithUnknown(old, new):
action = plans.NoOp
showJustNew = true
default:
action = plans.Update
}
return action, showJustNew
}
func (p *blockBodyDiffPrinter) writeAttrDiff(name string, attrS *configschema.Attribute, old, new cty.Value, nameLen, indent int, path cty.Path) bool {
path = append(path, cty.GetAttrStep{Name: name})
action, showJustNew := getPlanActionAndShow(old, new)
if action == plans.NoOp && !p.verbose && !identifyingAttribute(name, attrS) {
return true
}
if attrS.NestedType != nil {
p.writeNestedAttrDiff(name, attrS.NestedType, old, new, nameLen, indent, path, action, showJustNew)
return false
}
p.buf.WriteString("\n")
p.writeSensitivityWarning(old, new, indent, action, false)
p.buf.WriteString(strings.Repeat(" ", indent))
p.writeActionSymbol(action)
p.buf.WriteString(p.color.Color("[bold]"))
p.buf.WriteString(name)
p.buf.WriteString(p.color.Color("[reset]"))
p.buf.WriteString(strings.Repeat(" ", nameLen-len(name)))
p.buf.WriteString(" = ")
if attrS.Sensitive {
p.buf.WriteString("(sensitive value)")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
} else {
switch {
case showJustNew:
p.writeValue(new, action, indent+2)
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
default:
// We show new even if it is null to emphasize the fact
// that it is being unset, since otherwise it is easy to
// misunderstand that the value is still set to the old value.
p.writeValueDiff(old, new, indent+2, path)
}
}
return false
}
// writeNestedAttrDiff is responsible for formatting Attributes with NestedTypes
// in the diff.
func (p *blockBodyDiffPrinter) writeNestedAttrDiff(
name string, objS *configschema.Object, old, new cty.Value,
nameLen, indent int, path cty.Path, action plans.Action, showJustNew bool) {
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
p.writeActionSymbol(action)
p.buf.WriteString(p.color.Color("[bold]"))
p.buf.WriteString(name)
p.buf.WriteString(p.color.Color("[reset]"))
p.buf.WriteString(strings.Repeat(" ", nameLen-len(name)))
result := &blockBodyDiffResult{}
switch objS.Nesting {
case configschema.NestingSingle:
p.buf.WriteString(" = {")
if action != plans.NoOp && (p.pathForcesNewResource(path) || p.pathForcesNewResource(path[:len(path)-1])) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.writeAttrsDiff(objS.Attributes, old, new, indent+2, path, result)
p.writeSkippedAttr(result.skippedAttributes, indent+4)
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString("}")
case configschema.NestingList:
p.buf.WriteString(" = [")
if action != plans.NoOp && (p.pathForcesNewResource(path) || p.pathForcesNewResource(path[:len(path)-1])) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeActionSymbol(action)
p.buf.WriteString("{")
oldItems := ctyCollectionValues(old)
newItems := ctyCollectionValues(new)
// Here we intentionally preserve the index-based correspondance
// between old and new, rather than trying to detect insertions
// and removals in the list, because this more accurately reflects
// how Terraform Core and providers will understand the change,
// particularly when the nested block contains computed attributes
// that will themselves maintain correspondance by index.
// commonLen is number of elements that exist in both lists, which
// will be presented as updates (~). Any additional items in one
// of the lists will be presented as either creates (+) or deletes (-)
// depending on which list they belong to.
var commonLen int
// unchanged is the number of unchanged elements
var unchanged int
switch {
case len(oldItems) < len(newItems):
commonLen = len(oldItems)
default:
commonLen = len(newItems)
}
for i := 0; i < commonLen; i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
oldItem := oldItems[i]
newItem := newItems[i]
if oldItem.RawEquals(newItem) {
action = plans.NoOp
unchanged++
}
if action != plans.NoOp {
p.writeAttrsDiff(objS.Attributes, oldItem, newItem, indent+6, path, result)
p.writeSkippedAttr(result.skippedAttributes, indent+8)
p.buf.WriteString("\n")
}
}
for i := commonLen; i < len(oldItems); i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
oldItem := oldItems[i]
newItem := cty.NullVal(oldItem.Type())
p.writeAttrsDiff(objS.Attributes, oldItem, newItem, indent+6, path, result)
p.buf.WriteString("\n")
}
for i := commonLen; i < len(newItems); i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
newItem := newItems[i]
oldItem := cty.NullVal(newItem.Type())
p.writeAttrsDiff(objS.Attributes, oldItem, newItem, indent+6, path, result)
p.buf.WriteString("\n")
}
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString("},\n")
p.writeSkippedElems(unchanged, indent+4)
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.buf.WriteString("]")
if !new.IsKnown() {
p.buf.WriteString(" -> (known after apply)")
}
case configschema.NestingSet:
oldItems := ctyCollectionValues(old)
newItems := ctyCollectionValues(new)
var all cty.Value
if len(oldItems)+len(newItems) > 0 {
allItems := make([]cty.Value, 0, len(oldItems)+len(newItems))
allItems = append(allItems, oldItems...)
allItems = append(allItems, newItems...)
all = cty.SetVal(allItems)
} else {
all = cty.SetValEmpty(old.Type().ElementType())
}
p.buf.WriteString(" = [")
for it := all.ElementIterator(); it.Next(); {
_, val := it.Element()
var action plans.Action
var oldValue, newValue cty.Value
switch {
case !val.IsKnown():
action = plans.Update
newValue = val
case !new.IsKnown():
action = plans.Delete
// the value must have come from the old set
oldValue = val
// Mark the new val as null, but the entire set will be
// displayed as "(unknown after apply)"
newValue = cty.NullVal(val.Type())
2021-08-17 00:25:16 +02:00
case old.IsNull() || !old.HasElement(val).True():
action = plans.Create
oldValue = cty.NullVal(val.Type())
newValue = val
2021-08-17 00:25:16 +02:00
case new.IsNull() || !new.HasElement(val).True():
action = plans.Delete
oldValue = val
newValue = cty.NullVal(val.Type())
default:
action = plans.NoOp
oldValue = val
newValue = val
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeActionSymbol(action)
p.buf.WriteString("{")
if action != plans.NoOp && (p.pathForcesNewResource(path) || p.pathForcesNewResource(path[:len(path)-1])) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
path := append(path, cty.IndexStep{Key: val})
p.writeAttrsDiff(objS.Attributes, oldValue, newValue, indent+6, path, result)
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString("},")
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.buf.WriteString("]")
if !new.IsKnown() {
p.buf.WriteString(" -> (known after apply)")
}
case configschema.NestingMap:
// For the sake of handling nested blocks, we'll treat a null map
// the same as an empty map since the config language doesn't
// distinguish these anyway.
old = ctyNullBlockMapAsEmpty(old)
new = ctyNullBlockMapAsEmpty(new)
oldItems := old.AsValueMap()
newItems := map[string]cty.Value{}
if new.IsKnown() {
newItems = new.AsValueMap()
}
allKeys := make(map[string]bool)
for k := range oldItems {
allKeys[k] = true
}
for k := range newItems {
allKeys[k] = true
}
allKeysOrder := make([]string, 0, len(allKeys))
for k := range allKeys {
allKeysOrder = append(allKeysOrder, k)
}
sort.Strings(allKeysOrder)
p.buf.WriteString(" = {\n")
// unchanged tracks the number of unchanged elements
unchanged := 0
for _, k := range allKeysOrder {
var action plans.Action
oldValue := oldItems[k]
newValue := newItems[k]
switch {
case oldValue == cty.NilVal:
oldValue = cty.NullVal(newValue.Type())
action = plans.Create
case newValue == cty.NilVal:
newValue = cty.NullVal(oldValue.Type())
action = plans.Delete
case !newValue.RawEquals(oldValue):
action = plans.Update
default:
action = plans.NoOp
unchanged++
}
if action != plans.NoOp {
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeActionSymbol(action)
fmt.Fprintf(p.buf, "%q = {", k)
if p.pathForcesNewResource(path) || p.pathForcesNewResource(path[:len(path)-1]) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
path := append(path, cty.IndexStep{Key: cty.StringVal(k)})
p.writeAttrsDiff(objS.Attributes, oldValue, newValue, indent+6, path, result)
p.writeSkippedAttr(result.skippedAttributes, indent+8)
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString("},\n")
}
}
p.writeSkippedElems(unchanged, indent+4)
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.buf.WriteString("}")
if !new.IsKnown() {
p.buf.WriteString(" -> (known after apply)")
}
}
}
func (p *blockBodyDiffPrinter) writeNestedBlockDiffs(name string, blockS *configschema.NestedBlock, old, new cty.Value, blankBefore bool, indent int, path cty.Path) int {
skippedBlocks := 0
path = append(path, cty.GetAttrStep{Name: name})
if old.IsNull() && new.IsNull() {
// Nothing to do if both old and new is null
return skippedBlocks
}
// If either the old or the new value is marked,
// Display a special diff because it is irrelevant
// to list all obfuscated attributes as (sensitive)
if old.HasMark(marks.Sensitive) || new.HasMark(marks.Sensitive) {
p.writeSensitiveNestedBlockDiff(name, old, new, indent, blankBefore, path)
return 0
}
// Where old/new are collections representing a nesting mode other than
// NestingSingle, we assume the collection value can never be unknown
// since we always produce the container for the nested objects, even if
// the objects within are computed.
switch blockS.Nesting {
configs/configschema: Introduce the NestingGroup mode for blocks In study of existing providers we've found a pattern we werent previously accounting for of using a nested block type to represent a group of arguments that relate to a particular feature that is always enabled but where it improves configuration readability to group all of its settings together in a nested block. The existing NestingSingle was not a good fit for this because it is designed under the assumption that the presence or absence of the block has some significance in enabling or disabling the relevant feature, and so for these always-active cases we'd generate a misleading plan where the settings for the feature appear totally absent, rather than showing the default values that will be selected. NestingGroup is, therefore, a slight variation of NestingSingle where presence vs. absence of the block is not distinguishable (it's never null) and instead its contents are treated as unset when the block is absent. This then in turn causes any default values associated with the nested arguments to be honored and displayed in the plan whenever the block is not explicitly configured. The current SDK cannot activate this mode, but that's okay because its "legacy type system" opt-out flag allows it to force a block to be processed in this way anyway. We're adding this now so that we can introduce the feature in a future SDK without causing a breaking change to the protocol, since the set of possible block nesting modes is not extensible.
2019-04-09 00:32:53 +02:00
case configschema.NestingSingle, configschema.NestingGroup:
var action plans.Action
configs/configschema: Introduce the NestingGroup mode for blocks In study of existing providers we've found a pattern we werent previously accounting for of using a nested block type to represent a group of arguments that relate to a particular feature that is always enabled but where it improves configuration readability to group all of its settings together in a nested block. The existing NestingSingle was not a good fit for this because it is designed under the assumption that the presence or absence of the block has some significance in enabling or disabling the relevant feature, and so for these always-active cases we'd generate a misleading plan where the settings for the feature appear totally absent, rather than showing the default values that will be selected. NestingGroup is, therefore, a slight variation of NestingSingle where presence vs. absence of the block is not distinguishable (it's never null) and instead its contents are treated as unset when the block is absent. This then in turn causes any default values associated with the nested arguments to be honored and displayed in the plan whenever the block is not explicitly configured. The current SDK cannot activate this mode, but that's okay because its "legacy type system" opt-out flag allows it to force a block to be processed in this way anyway. We're adding this now so that we can introduce the feature in a future SDK without causing a breaking change to the protocol, since the set of possible block nesting modes is not extensible.
2019-04-09 00:32:53 +02:00
eqV := new.Equals(old)
switch {
case old.IsNull():
action = plans.Create
case new.IsNull():
action = plans.Delete
configs/configschema: Introduce the NestingGroup mode for blocks In study of existing providers we've found a pattern we werent previously accounting for of using a nested block type to represent a group of arguments that relate to a particular feature that is always enabled but where it improves configuration readability to group all of its settings together in a nested block. The existing NestingSingle was not a good fit for this because it is designed under the assumption that the presence or absence of the block has some significance in enabling or disabling the relevant feature, and so for these always-active cases we'd generate a misleading plan where the settings for the feature appear totally absent, rather than showing the default values that will be selected. NestingGroup is, therefore, a slight variation of NestingSingle where presence vs. absence of the block is not distinguishable (it's never null) and instead its contents are treated as unset when the block is absent. This then in turn causes any default values associated with the nested arguments to be honored and displayed in the plan whenever the block is not explicitly configured. The current SDK cannot activate this mode, but that's okay because its "legacy type system" opt-out flag allows it to force a block to be processed in this way anyway. We're adding this now so that we can introduce the feature in a future SDK without causing a breaking change to the protocol, since the set of possible block nesting modes is not extensible.
2019-04-09 00:32:53 +02:00
case !new.IsWhollyKnown() || !old.IsWhollyKnown():
// "old" should actually always be known due to our contract
// that old values must never be unknown, but we'll allow it
// anyway to be robust.
action = plans.Update
configs/configschema: Introduce the NestingGroup mode for blocks In study of existing providers we've found a pattern we werent previously accounting for of using a nested block type to represent a group of arguments that relate to a particular feature that is always enabled but where it improves configuration readability to group all of its settings together in a nested block. The existing NestingSingle was not a good fit for this because it is designed under the assumption that the presence or absence of the block has some significance in enabling or disabling the relevant feature, and so for these always-active cases we'd generate a misleading plan where the settings for the feature appear totally absent, rather than showing the default values that will be selected. NestingGroup is, therefore, a slight variation of NestingSingle where presence vs. absence of the block is not distinguishable (it's never null) and instead its contents are treated as unset when the block is absent. This then in turn causes any default values associated with the nested arguments to be honored and displayed in the plan whenever the block is not explicitly configured. The current SDK cannot activate this mode, but that's okay because its "legacy type system" opt-out flag allows it to force a block to be processed in this way anyway. We're adding this now so that we can introduce the feature in a future SDK without causing a breaking change to the protocol, since the set of possible block nesting modes is not extensible.
2019-04-09 00:32:53 +02:00
case !eqV.IsKnown() || !eqV.True():
action = plans.Update
}
if blankBefore {
p.buf.WriteRune('\n')
}
skipped := p.writeNestedBlockDiff(name, nil, &blockS.Block, action, old, new, indent, path)
if skipped {
return 1
}
case configschema.NestingList:
// For the sake of handling nested blocks, we'll treat a null list
// the same as an empty list since the config language doesn't
// distinguish these anyway.
old = ctyNullBlockListAsEmpty(old)
new = ctyNullBlockListAsEmpty(new)
oldItems := ctyCollectionValues(old)
newItems := ctyCollectionValues(new)
// Here we intentionally preserve the index-based correspondance
// between old and new, rather than trying to detect insertions
// and removals in the list, because this more accurately reflects
// how Terraform Core and providers will understand the change,
// particularly when the nested block contains computed attributes
// that will themselves maintain correspondance by index.
// commonLen is number of elements that exist in both lists, which
// will be presented as updates (~). Any additional items in one
// of the lists will be presented as either creates (+) or deletes (-)
// depending on which list they belong to.
var commonLen int
switch {
case len(oldItems) < len(newItems):
commonLen = len(oldItems)
default:
commonLen = len(newItems)
}
if blankBefore && (len(oldItems) > 0 || len(newItems) > 0) {
p.buf.WriteRune('\n')
}
for i := 0; i < commonLen; i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
oldItem := oldItems[i]
newItem := newItems[i]
action := plans.Update
if oldItem.RawEquals(newItem) {
action = plans.NoOp
}
skipped := p.writeNestedBlockDiff(name, nil, &blockS.Block, action, oldItem, newItem, indent, path)
if skipped {
skippedBlocks++
}
}
for i := commonLen; i < len(oldItems); i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
oldItem := oldItems[i]
newItem := cty.NullVal(oldItem.Type())
skipped := p.writeNestedBlockDiff(name, nil, &blockS.Block, plans.Delete, oldItem, newItem, indent, path)
if skipped {
skippedBlocks++
}
}
for i := commonLen; i < len(newItems); i++ {
path := append(path, cty.IndexStep{Key: cty.NumberIntVal(int64(i))})
newItem := newItems[i]
oldItem := cty.NullVal(newItem.Type())
skipped := p.writeNestedBlockDiff(name, nil, &blockS.Block, plans.Create, oldItem, newItem, indent, path)
if skipped {
skippedBlocks++
}
}
case configschema.NestingSet:
// For the sake of handling nested blocks, we'll treat a null set
// the same as an empty set since the config language doesn't
// distinguish these anyway.
old = ctyNullBlockSetAsEmpty(old)
new = ctyNullBlockSetAsEmpty(new)
oldItems := ctyCollectionValues(old)
newItems := ctyCollectionValues(new)
if (len(oldItems) + len(newItems)) == 0 {
// Nothing to do if both sets are empty
return 0
}
allItems := make([]cty.Value, 0, len(oldItems)+len(newItems))
allItems = append(allItems, oldItems...)
allItems = append(allItems, newItems...)
all := cty.SetVal(allItems)
if blankBefore {
p.buf.WriteRune('\n')
}
for it := all.ElementIterator(); it.Next(); {
_, val := it.Element()
var action plans.Action
var oldValue, newValue cty.Value
switch {
case !val.IsKnown():
action = plans.Update
newValue = val
case !old.HasElement(val).True():
action = plans.Create
oldValue = cty.NullVal(val.Type())
newValue = val
case !new.HasElement(val).True():
action = plans.Delete
oldValue = val
newValue = cty.NullVal(val.Type())
default:
action = plans.NoOp
oldValue = val
newValue = val
}
path := append(path, cty.IndexStep{Key: val})
skipped := p.writeNestedBlockDiff(name, nil, &blockS.Block, action, oldValue, newValue, indent, path)
if skipped {
skippedBlocks++
}
}
case configschema.NestingMap:
// For the sake of handling nested blocks, we'll treat a null map
// the same as an empty map since the config language doesn't
// distinguish these anyway.
old = ctyNullBlockMapAsEmpty(old)
new = ctyNullBlockMapAsEmpty(new)
oldItems := old.AsValueMap()
newItems := new.AsValueMap()
if (len(oldItems) + len(newItems)) == 0 {
// Nothing to do if both maps are empty
return 0
}
allKeys := make(map[string]bool)
for k := range oldItems {
allKeys[k] = true
}
for k := range newItems {
allKeys[k] = true
}
allKeysOrder := make([]string, 0, len(allKeys))
for k := range allKeys {
allKeysOrder = append(allKeysOrder, k)
}
sort.Strings(allKeysOrder)
if blankBefore {
p.buf.WriteRune('\n')
}
for _, k := range allKeysOrder {
var action plans.Action
oldValue := oldItems[k]
newValue := newItems[k]
switch {
case oldValue == cty.NilVal:
oldValue = cty.NullVal(newValue.Type())
action = plans.Create
case newValue == cty.NilVal:
newValue = cty.NullVal(oldValue.Type())
action = plans.Delete
case !newValue.RawEquals(oldValue):
action = plans.Update
default:
action = plans.NoOp
}
path := append(path, cty.IndexStep{Key: cty.StringVal(k)})
skipped := p.writeNestedBlockDiff(name, &k, &blockS.Block, action, oldValue, newValue, indent, path)
if skipped {
skippedBlocks++
}
}
}
return skippedBlocks
}
func (p *blockBodyDiffPrinter) writeSensitiveNestedBlockDiff(name string, old, new cty.Value, indent int, blankBefore bool, path cty.Path) {
var action plans.Action
switch {
case old.IsNull():
action = plans.Create
case new.IsNull():
action = plans.Delete
case !new.IsWhollyKnown() || !old.IsWhollyKnown():
// "old" should actually always be known due to our contract
// that old values must never be unknown, but we'll allow it
// anyway to be robust.
action = plans.Update
case !ctyEqualValueAndMarks(old, new):
action = plans.Update
}
if blankBefore {
p.buf.WriteRune('\n')
}
// New line before warning printing
p.buf.WriteRune('\n')
p.writeSensitivityWarning(old, new, indent, action, true)
p.buf.WriteString(strings.Repeat(" ", indent))
p.writeActionSymbol(action)
fmt.Fprintf(p.buf, "%s {", name)
if action != plans.NoOp && p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteRune('\n')
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString("# At least one attribute in this block is (or was) sensitive,\n")
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString("# so its contents will not be displayed.")
p.buf.WriteRune('\n')
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.buf.WriteString("}")
}
func (p *blockBodyDiffPrinter) writeNestedBlockDiff(name string, label *string, blockS *configschema.Block, action plans.Action, old, new cty.Value, indent int, path cty.Path) bool {
if action == plans.NoOp && !p.verbose {
return true
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
p.writeActionSymbol(action)
if label != nil {
fmt.Fprintf(p.buf, "%s %q {", name, *label)
} else {
fmt.Fprintf(p.buf, "%s {", name)
}
if action != plans.NoOp && (p.pathForcesNewResource(path) || p.pathForcesNewResource(path[:len(path)-1])) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
result := p.writeBlockBodyDiff(blockS, old, new, indent+4, path)
if result.bodyWritten {
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
}
p.buf.WriteString("}")
return false
}
func (p *blockBodyDiffPrinter) writeValue(val cty.Value, action plans.Action, indent int) {
// Could check specifically for the sensitivity marker
if val.HasMark(marks.Sensitive) {
p.buf.WriteString("(sensitive)")
return
}
if !val.IsKnown() {
p.buf.WriteString("(known after apply)")
return
}
if val.IsNull() {
p.buf.WriteString(p.color.Color("[dark_gray]null[reset]"))
return
}
ty := val.Type()
switch {
case ty.IsPrimitiveType():
switch ty {
case cty.String:
{
// Special behavior for JSON strings containing array or object
src := []byte(val.AsString())
ty, err := ctyjson.ImpliedType(src)
// check for the special case of "null", which decodes to nil,
// and just allow it to be printed out directly
if err == nil && !ty.IsPrimitiveType() && strings.TrimSpace(val.AsString()) != "null" {
jv, err := ctyjson.Unmarshal(src, ty)
if err == nil {
p.buf.WriteString("jsonencode(")
if jv.LengthInt() == 0 {
p.writeValue(jv, action, 0)
} else {
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeValue(jv, action, indent+4)
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent))
}
p.buf.WriteByte(')')
break // don't *also* do the normal behavior below
}
}
}
if strings.Contains(val.AsString(), "\n") {
// It's a multi-line string, so we want to use the multi-line
// rendering so it'll be readable. Rather than re-implement
// that here, we'll just re-use the multi-line string diff
// printer with no changes, which ends up producing the
// result we want here.
// The path argument is nil because we don't track path
// information into strings and we know that a string can't
// have any indices or attributes that might need to be marked
// as (requires replacement), which is what that argument is for.
p.writeValueDiff(val, val, indent, nil)
break
}
fmt.Fprintf(p.buf, "%q", val.AsString())
case cty.Bool:
if val.True() {
p.buf.WriteString("true")
} else {
p.buf.WriteString("false")
}
case cty.Number:
bf := val.AsBigFloat()
p.buf.WriteString(bf.Text('f', -1))
default:
// should never happen, since the above is exhaustive
fmt.Fprintf(p.buf, "%#v", val)
}
case ty.IsListType() || ty.IsSetType() || ty.IsTupleType():
p.buf.WriteString("[")
it := val.ElementIterator()
for it.Next() {
_, val := it.Element()
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.writeValue(val, action, indent+4)
p.buf.WriteString(",")
}
if val.LengthInt() > 0 {
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
}
p.buf.WriteString("]")
case ty.IsMapType():
p.buf.WriteString("{")
keyLen := 0
for it := val.ElementIterator(); it.Next(); {
key, _ := it.Element()
if keyStr := key.AsString(); len(keyStr) > keyLen {
keyLen = len(keyStr)
}
}
for it := val.ElementIterator(); it.Next(); {
key, val := it.Element()
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.writeValue(key, action, indent+4)
p.buf.WriteString(strings.Repeat(" ", keyLen-len(key.AsString())))
p.buf.WriteString(" = ")
p.writeValue(val, action, indent+4)
}
if val.LengthInt() > 0 {
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
}
p.buf.WriteString("}")
case ty.IsObjectType():
p.buf.WriteString("{")
atys := ty.AttributeTypes()
attrNames := make([]string, 0, len(atys))
nameLen := 0
for attrName := range atys {
attrNames = append(attrNames, attrName)
if len(attrName) > nameLen {
nameLen = len(attrName)
}
}
sort.Strings(attrNames)
for _, attrName := range attrNames {
val := val.GetAttr(attrName)
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.buf.WriteString(attrName)
p.buf.WriteString(strings.Repeat(" ", nameLen-len(attrName)))
p.buf.WriteString(" = ")
p.writeValue(val, action, indent+4)
}
if len(attrNames) > 0 {
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
}
p.buf.WriteString("}")
}
}
func (p *blockBodyDiffPrinter) writeValueDiff(old, new cty.Value, indent int, path cty.Path) {
ty := old.Type()
typesEqual := ctyTypesEqual(ty, new.Type())
// We have some specialized diff implementations for certain complex
// values where it's useful to see a visualization of the diff of
// the nested elements rather than just showing the entire old and
// new values verbatim.
// However, these specialized implementations can apply only if both
// values are known and non-null.
if old.IsKnown() && new.IsKnown() && !old.IsNull() && !new.IsNull() && typesEqual {
if old.HasMark(marks.Sensitive) || new.HasMark(marks.Sensitive) {
p.buf.WriteString("(sensitive)")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
return
}
switch {
case ty == cty.String:
// We have special behavior for both multi-line strings in general
// and for strings that can parse as JSON. For the JSON handling
// to apply, both old and new must be valid JSON.
// For single-line strings that don't parse as JSON we just fall
// out of this switch block and do the default old -> new rendering.
oldS := old.AsString()
newS := new.AsString()
{
// Special behavior for JSON strings containing object or
// list values.
oldBytes := []byte(oldS)
newBytes := []byte(newS)
oldType, oldErr := ctyjson.ImpliedType(oldBytes)
newType, newErr := ctyjson.ImpliedType(newBytes)
if oldErr == nil && newErr == nil && !(oldType.IsPrimitiveType() && newType.IsPrimitiveType()) {
oldJV, oldErr := ctyjson.Unmarshal(oldBytes, oldType)
newJV, newErr := ctyjson.Unmarshal(newBytes, newType)
if oldErr == nil && newErr == nil {
if !oldJV.RawEquals(newJV) { // two JSON values may differ only in insignificant whitespace
p.buf.WriteString("jsonencode(")
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(plans.Update)
p.writeValueDiff(oldJV, newJV, indent+4, path)
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteByte(')')
} else {
// if they differ only in insignificant whitespace
// then we'll note that but still expand out the
// effective value.
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color("jsonencode( [red]# whitespace changes force replacement[reset]"))
} else {
p.buf.WriteString(p.color.Color("jsonencode( [dim]# whitespace changes[reset]"))
}
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeValue(oldJV, plans.NoOp, indent+4)
p.buf.WriteByte('\n')
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteByte(')')
}
return
}
}
}
2020-12-01 18:42:37 +01:00
if !strings.Contains(oldS, "\n") && !strings.Contains(newS, "\n") {
break
}
p.buf.WriteString("<<-EOT")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteString("\n")
var oldLines, newLines []cty.Value
{
r := strings.NewReader(oldS)
sc := bufio.NewScanner(r)
for sc.Scan() {
oldLines = append(oldLines, cty.StringVal(sc.Text()))
}
}
{
r := strings.NewReader(newS)
sc := bufio.NewScanner(r)
for sc.Scan() {
newLines = append(newLines, cty.StringVal(sc.Text()))
}
}
// Optimization for strings which are exactly equal: just print
// directly without calculating the sequence diff. This makes a
// significant difference when this code path is reached via a
// writeValue call with a large multi-line string.
if oldS == newS {
for _, line := range newLines {
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.buf.WriteString(line.AsString())
p.buf.WriteString("\n")
}
} else {
diffLines := ctySequenceDiff(oldLines, newLines)
for _, diffLine := range diffLines {
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(diffLine.Action)
switch diffLine.Action {
case plans.NoOp, plans.Delete:
p.buf.WriteString(diffLine.Before.AsString())
case plans.Create:
p.buf.WriteString(diffLine.After.AsString())
default:
// Should never happen since the above covers all
// actions that ctySequenceDiff can return for strings
p.buf.WriteString(diffLine.After.AsString())
}
p.buf.WriteString("\n")
}
}
p.buf.WriteString(strings.Repeat(" ", indent)) // +4 here because there's no symbol
p.buf.WriteString("EOT")
return
case ty.IsSetType():
p.buf.WriteString("[")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteString("\n")
var addedVals, removedVals, allVals []cty.Value
for it := old.ElementIterator(); it.Next(); {
_, val := it.Element()
allVals = append(allVals, val)
if new.HasElement(val).False() {
removedVals = append(removedVals, val)
}
}
for it := new.ElementIterator(); it.Next(); {
_, val := it.Element()
allVals = append(allVals, val)
if val.IsKnown() && old.HasElement(val).False() {
addedVals = append(addedVals, val)
}
}
var all, added, removed cty.Value
if len(allVals) > 0 {
all = cty.SetVal(allVals)
} else {
all = cty.SetValEmpty(ty.ElementType())
}
if len(addedVals) > 0 {
added = cty.SetVal(addedVals)
} else {
added = cty.SetValEmpty(ty.ElementType())
}
if len(removedVals) > 0 {
removed = cty.SetVal(removedVals)
} else {
removed = cty.SetValEmpty(ty.ElementType())
}
suppressedElements := 0
for it := all.ElementIterator(); it.Next(); {
_, val := it.Element()
var action plans.Action
switch {
case !val.IsKnown():
action = plans.Update
case added.HasElement(val).True():
action = plans.Create
case removed.HasElement(val).True():
action = plans.Delete
default:
action = plans.NoOp
}
if action == plans.NoOp && !p.verbose {
suppressedElements++
continue
}
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.writeValue(val, action, indent+4)
p.buf.WriteString(",\n")
}
if suppressedElements > 0 {
p.writeActionSymbol(plans.NoOp)
p.buf.WriteString(strings.Repeat(" ", indent+2))
noun := "elements"
if suppressedElements == 1 {
noun = "element"
}
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), suppressedElements, noun))
p.buf.WriteString("\n")
}
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString("]")
return
case ty.IsListType() || ty.IsTupleType():
p.buf.WriteString("[")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteString("\n")
elemDiffs := ctySequenceDiff(old.AsValueSlice(), new.AsValueSlice())
// Maintain a stack of suppressed lines in the diff for later
// display or elision
var suppressedElements []*plans.Change
var changeShown bool
for i := 0; i < len(elemDiffs); i++ {
if !p.verbose {
for i < len(elemDiffs) && elemDiffs[i].Action == plans.NoOp {
suppressedElements = append(suppressedElements, elemDiffs[i])
i++
}
}
// If we have some suppressed elements on the stack…
if len(suppressedElements) > 0 {
// If we've just rendered a change, display the first
// element in the stack as context
if changeShown {
elemDiff := suppressedElements[0]
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeValue(elemDiff.After, elemDiff.Action, indent+4)
p.buf.WriteString(",\n")
suppressedElements = suppressedElements[1:]
}
hidden := len(suppressedElements)
// If we're not yet at the end of the list, capture the
// last element on the stack as context for the upcoming
// change to be rendered
var nextContextDiff *plans.Change
if hidden > 0 && i < len(elemDiffs) {
hidden--
nextContextDiff = suppressedElements[hidden]
}
// If there are still hidden elements, show an elision
// statement counting them
if hidden > 0 {
p.writeActionSymbol(plans.NoOp)
p.buf.WriteString(strings.Repeat(" ", indent+2))
noun := "elements"
if hidden == 1 {
noun = "element"
}
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), hidden, noun))
p.buf.WriteString("\n")
}
// Display the next context diff if it was captured above
if nextContextDiff != nil {
p.buf.WriteString(strings.Repeat(" ", indent+4))
p.writeValue(nextContextDiff.After, nextContextDiff.Action, indent+4)
p.buf.WriteString(",\n")
}
// Suppressed elements have now been handled so clear them again
suppressedElements = nil
}
if i >= len(elemDiffs) {
break
}
elemDiff := elemDiffs[i]
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(elemDiff.Action)
switch elemDiff.Action {
case plans.NoOp, plans.Delete:
p.writeValue(elemDiff.Before, elemDiff.Action, indent+4)
case plans.Update:
p.writeValueDiff(elemDiff.Before, elemDiff.After, indent+4, path)
case plans.Create:
p.writeValue(elemDiff.After, elemDiff.Action, indent+4)
default:
// Should never happen since the above covers all
// actions that ctySequenceDiff can return.
p.writeValue(elemDiff.After, elemDiff.Action, indent+4)
}
p.buf.WriteString(",\n")
changeShown = true
}
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString("]")
return
case ty.IsMapType():
p.buf.WriteString("{")
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
p.buf.WriteString("\n")
var allKeys []string
keyLen := 0
for it := old.ElementIterator(); it.Next(); {
k, _ := it.Element()
keyStr := k.AsString()
allKeys = append(allKeys, keyStr)
if len(keyStr) > keyLen {
keyLen = len(keyStr)
}
}
for it := new.ElementIterator(); it.Next(); {
k, _ := it.Element()
keyStr := k.AsString()
allKeys = append(allKeys, keyStr)
if len(keyStr) > keyLen {
keyLen = len(keyStr)
}
}
sort.Strings(allKeys)
suppressedElements := 0
lastK := ""
for i, k := range allKeys {
if i > 0 && lastK == k {
continue // skip duplicates (list is sorted)
}
lastK = k
kV := cty.StringVal(k)
var action plans.Action
if old.HasIndex(kV).False() {
action = plans.Create
} else if new.HasIndex(kV).False() {
action = plans.Delete
}
if old.HasIndex(kV).True() && new.HasIndex(kV).True() {
if ctyEqualValueAndMarks(old.Index(kV), new.Index(kV)) {
action = plans.NoOp
} else {
action = plans.Update
}
}
if action == plans.NoOp && !p.verbose {
suppressedElements++
continue
}
path := append(path, cty.IndexStep{Key: kV})
oldV := old.Index(kV)
newV := new.Index(kV)
p.writeSensitivityWarning(oldV, newV, indent+2, action, false)
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.writeValue(kV, action, indent+4)
p.buf.WriteString(strings.Repeat(" ", keyLen-len(k)))
p.buf.WriteString(" = ")
switch action {
case plans.Create, plans.NoOp:
v := new.Index(kV)
if v.HasMark(marks.Sensitive) {
p.buf.WriteString("(sensitive)")
} else {
p.writeValue(v, action, indent+4)
}
case plans.Delete:
oldV := old.Index(kV)
newV := cty.NullVal(oldV.Type())
p.writeValueDiff(oldV, newV, indent+4, path)
default:
if oldV.HasMark(marks.Sensitive) || newV.HasMark(marks.Sensitive) {
p.buf.WriteString("(sensitive)")
} else {
p.writeValueDiff(oldV, newV, indent+4, path)
}
}
p.buf.WriteByte('\n')
}
if suppressedElements > 0 {
p.writeActionSymbol(plans.NoOp)
p.buf.WriteString(strings.Repeat(" ", indent+2))
noun := "elements"
if suppressedElements == 1 {
noun = "element"
}
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), suppressedElements, noun))
p.buf.WriteString("\n")
}
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString("}")
return
case ty.IsObjectType():
p.buf.WriteString("{")
p.buf.WriteString("\n")
forcesNewResource := p.pathForcesNewResource(path)
var allKeys []string
keyLen := 0
for it := old.ElementIterator(); it.Next(); {
k, _ := it.Element()
keyStr := k.AsString()
allKeys = append(allKeys, keyStr)
if len(keyStr) > keyLen {
keyLen = len(keyStr)
}
}
for it := new.ElementIterator(); it.Next(); {
k, _ := it.Element()
keyStr := k.AsString()
allKeys = append(allKeys, keyStr)
if len(keyStr) > keyLen {
keyLen = len(keyStr)
}
}
sort.Strings(allKeys)
suppressedElements := 0
lastK := ""
for i, k := range allKeys {
if i > 0 && lastK == k {
continue // skip duplicates (list is sorted)
}
lastK = k
kV := k
var action plans.Action
if !old.Type().HasAttribute(kV) {
action = plans.Create
} else if !new.Type().HasAttribute(kV) {
action = plans.Delete
} else if ctyEqualValueAndMarks(old.GetAttr(kV), new.GetAttr(kV)) {
action = plans.NoOp
} else {
action = plans.Update
}
if action == plans.NoOp && !p.verbose {
suppressedElements++
continue
}
path := append(path, cty.GetAttrStep{Name: kV})
p.buf.WriteString(strings.Repeat(" ", indent+2))
p.writeActionSymbol(action)
p.buf.WriteString(k)
p.buf.WriteString(strings.Repeat(" ", keyLen-len(k)))
p.buf.WriteString(" = ")
switch action {
case plans.Create, plans.NoOp:
v := new.GetAttr(kV)
p.writeValue(v, action, indent+4)
case plans.Delete:
oldV := old.GetAttr(kV)
newV := cty.NullVal(oldV.Type())
p.writeValueDiff(oldV, newV, indent+4, path)
default:
oldV := old.GetAttr(kV)
newV := new.GetAttr(kV)
p.writeValueDiff(oldV, newV, indent+4, path)
}
p.buf.WriteString("\n")
}
if suppressedElements > 0 {
p.writeActionSymbol(plans.NoOp)
p.buf.WriteString(strings.Repeat(" ", indent+2))
noun := "elements"
if suppressedElements == 1 {
noun = "element"
}
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), suppressedElements, noun))
p.buf.WriteString("\n")
}
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString("}")
if forcesNewResource {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
return
}
}
// In all other cases, we just show the new and old values as-is
p.writeValue(old, plans.Delete, indent)
if new.IsNull() {
p.buf.WriteString(p.color.Color(" [dark_gray]->[reset] "))
} else {
p.buf.WriteString(p.color.Color(" [yellow]->[reset] "))
}
p.writeValue(new, plans.Create, indent)
if p.pathForcesNewResource(path) {
p.buf.WriteString(p.color.Color(forcesNewResourceCaption))
}
}
// writeActionSymbol writes a symbol to represent the given action, followed
// by a space.
//
// It only supports the actions that can be represented with a single character:
// Create, Delete, Update and NoAction.
func (p *blockBodyDiffPrinter) writeActionSymbol(action plans.Action) {
switch action {
case plans.Create:
p.buf.WriteString(p.color.Color("[green]+[reset] "))
case plans.Delete:
p.buf.WriteString(p.color.Color("[red]-[reset] "))
case plans.Update:
p.buf.WriteString(p.color.Color("[yellow]~[reset] "))
case plans.NoOp:
p.buf.WriteString(" ")
default:
// Should never happen
p.buf.WriteString(p.color.Color("? "))
}
}
func (p *blockBodyDiffPrinter) writeSensitivityWarning(old, new cty.Value, indent int, action plans.Action, isBlock bool) {
// Dont' show this warning for create or delete
if action == plans.Create || action == plans.Delete {
return
}
// Customize the warning based on if it is an attribute or block
diffType := "attribute value"
if isBlock {
diffType = "block"
}
// If only attribute sensitivity is changing, clarify that the value is unchanged
var valueUnchangedSuffix string
if !isBlock {
oldUnmarked, _ := old.UnmarkDeep()
newUnmarked, _ := new.UnmarkDeep()
if oldUnmarked.RawEquals(newUnmarked) {
valueUnchangedSuffix = " The value is unchanged."
}
}
if new.HasMark(marks.Sensitive) && !old.HasMark(marks.Sensitive) {
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf(p.color.Color("# [yellow]Warning:[reset] this %s will be marked as sensitive and will not\n"), diffType))
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf("# display in UI output after applying this change.%s\n", valueUnchangedSuffix))
}
// Note if changing this attribute will change its sensitivity
if old.HasMark(marks.Sensitive) && !new.HasMark(marks.Sensitive) {
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf(p.color.Color("# [yellow]Warning:[reset] this %s will no longer be marked as sensitive\n"), diffType))
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf("# after applying this change.%s\n", valueUnchangedSuffix))
}
}
func (p *blockBodyDiffPrinter) pathForcesNewResource(path cty.Path) bool {
if !p.action.IsReplace() || p.requiredReplace.Empty() {
// "requiredReplace" only applies when the instance is being replaced,
// and we should only inspect that set if it is not empty
return false
}
return p.requiredReplace.Has(path)
}
func ctyEmptyString(value cty.Value) bool {
if !value.IsNull() && value.IsKnown() {
valueType := value.Type()
if valueType == cty.String && value.AsString() == "" {
return true
}
}
return false
}
func ctyGetAttrMaybeNull(val cty.Value, name string) cty.Value {
attrType := val.Type().AttributeType(name)
if val.IsNull() {
return cty.NullVal(attrType)
}
// We treat "" as null here
// as existing SDK doesn't support null yet.
// This allows us to avoid spurious diffs
// until we introduce null to the SDK.
attrValue := val.GetAttr(name)
// If the value is marked, the ctyEmptyString function will fail
if !val.ContainsMarked() && ctyEmptyString(attrValue) {
return cty.NullVal(attrType)
}
return attrValue
}
func ctyCollectionValues(val cty.Value) []cty.Value {
if !val.IsKnown() || val.IsNull() {
2018-10-19 01:21:32 +02:00
return nil
}
ret := make([]cty.Value, 0, val.LengthInt())
for it := val.ElementIterator(); it.Next(); {
_, value := it.Element()
ret = append(ret, value)
}
return ret
}
// ctySequenceDiff returns differences between given sequences of cty.Value(s)
// in the form of Create, Delete, or Update actions (for objects).
func ctySequenceDiff(old, new []cty.Value) []*plans.Change {
var ret []*plans.Change
lcs := objchange.LongestCommonSubsequence(old, new)
var oldI, newI, lcsI int
for oldI < len(old) || newI < len(new) || lcsI < len(lcs) {
// We first process items in the old and new sequences which are not
// equal to the current common sequence item. Old items are marked as
// deletions, and new items are marked as additions.
//
// There is an exception for deleted & created object items, which we
// try to render as updates where that makes sense.
for oldI < len(old) && (lcsI >= len(lcs) || !old[oldI].RawEquals(lcs[lcsI])) {
// Render this as an object update if all of these are true:
//
// - the current old item is an object;
// - there's a current new item which is also an object;
// - either there are no common items left, or the current new item
// doesn't equal the current common item.
//
// Why do we need the the last clause? If we have current items in all
// three sequences, and the current new item is equal to a common item,
// then we should just need to advance the old item list and we'll
// eventually find a common item matching both old and new.
//
// This combination of conditions allows us to render an object update
// diff instead of a combination of delete old & create new.
isObjectDiff := old[oldI].Type().IsObjectType() && newI < len(new) && new[newI].Type().IsObjectType() && (lcsI >= len(lcs) || !new[newI].RawEquals(lcs[lcsI]))
if isObjectDiff {
ret = append(ret, &plans.Change{
Action: plans.Update,
Before: old[oldI],
After: new[newI],
})
oldI++
newI++ // we also consume the next "new" in this case
continue
}
// Otherwise, this item is not part of the common sequence, so
// render as a deletion.
ret = append(ret, &plans.Change{
Action: plans.Delete,
Before: old[oldI],
After: cty.NullVal(old[oldI].Type()),
})
oldI++
}
for newI < len(new) && (lcsI >= len(lcs) || !new[newI].RawEquals(lcs[lcsI])) {
ret = append(ret, &plans.Change{
Action: plans.Create,
Before: cty.NullVal(new[newI].Type()),
After: new[newI],
})
newI++
}
// When we've exhausted the old & new sequences of items which are not
// in the common subsequence, we render a common item and continue.
if lcsI < len(lcs) {
ret = append(ret, &plans.Change{
Action: plans.NoOp,
Before: lcs[lcsI],
After: lcs[lcsI],
})
// All of our indexes advance together now, since the line
// is common to all three sequences.
lcsI++
oldI++
newI++
}
}
return ret
}
// ctyEqualValueAndMarks checks equality of two possibly-marked values,
// considering partially-unknown values and equal values with different marks
// as inequal
func ctyEqualWithUnknown(old, new cty.Value) bool {
if !old.IsWhollyKnown() || !new.IsWhollyKnown() {
return false
}
return ctyEqualValueAndMarks(old, new)
}
// ctyEqualValueAndMarks checks equality of two possibly-marked values,
// considering equal values with different marks as inequal
func ctyEqualValueAndMarks(old, new cty.Value) bool {
oldUnmarked, oldMarks := old.UnmarkDeep()
newUnmarked, newMarks := new.UnmarkDeep()
sameValue := oldUnmarked.Equals(newUnmarked)
return sameValue.IsKnown() && sameValue.True() && oldMarks.Equal(newMarks)
}
// ctyTypesEqual checks equality of two types more loosely
// by avoiding checks of object/tuple elements
// as we render differences on element-by-element basis anyway
func ctyTypesEqual(oldT, newT cty.Type) bool {
if oldT.IsObjectType() && newT.IsObjectType() {
return true
}
if oldT.IsTupleType() && newT.IsTupleType() {
return true
}
return oldT.Equals(newT)
}
func ctyEnsurePathCapacity(path cty.Path, minExtra int) cty.Path {
if cap(path)-len(path) >= minExtra {
return path
}
newCap := cap(path) * 2
if newCap < (len(path) + minExtra) {
newCap = len(path) + minExtra
}
newPath := make(cty.Path, len(path), newCap)
copy(newPath, path)
return newPath
}
// ctyNullBlockListAsEmpty either returns the given value verbatim if it is non-nil
// or returns an empty value of a suitable type to serve as a placeholder for it.
//
// In particular, this function handles the special situation where a "list" is
// actually represented as a tuple type where nested blocks contain
// dynamically-typed values.
func ctyNullBlockListAsEmpty(in cty.Value) cty.Value {
if !in.IsNull() {
return in
}
if ty := in.Type(); ty.IsListType() {
return cty.ListValEmpty(ty.ElementType())
}
return cty.EmptyTupleVal // must need a tuple, then
}
// ctyNullBlockMapAsEmpty either returns the given value verbatim if it is non-nil
// or returns an empty value of a suitable type to serve as a placeholder for it.
//
// In particular, this function handles the special situation where a "map" is
// actually represented as an object type where nested blocks contain
// dynamically-typed values.
func ctyNullBlockMapAsEmpty(in cty.Value) cty.Value {
if !in.IsNull() {
return in
}
if ty := in.Type(); ty.IsMapType() {
return cty.MapValEmpty(ty.ElementType())
}
return cty.EmptyObjectVal // must need an object, then
}
// ctyNullBlockSetAsEmpty either returns the given value verbatim if it is non-nil
// or returns an empty value of a suitable type to serve as a placeholder for it.
func ctyNullBlockSetAsEmpty(in cty.Value) cty.Value {
if !in.IsNull() {
return in
}
// Dynamically-typed attributes are not supported inside blocks backed by
// sets, so our result here is always a set.
return cty.SetValEmpty(in.Type().ElementType())
}
// DiffActionSymbol returns a string that, once passed through a
// colorstring.Colorize, will produce a result that can be written
// to a terminal to produce a symbol made of three printable
// characters, possibly interspersed with VT100 color codes.
func DiffActionSymbol(action plans.Action) string {
switch action {
case plans.DeleteThenCreate:
return "[red]-[reset]/[green]+[reset]"
case plans.CreateThenDelete:
return "[green]+[reset]/[red]-[reset]"
case plans.Create:
return " [green]+[reset]"
case plans.Delete:
return " [red]-[reset]"
case plans.Read:
return " [cyan]<=[reset]"
case plans.Update:
return " [yellow]~[reset]"
default:
return " ?"
}
}
// Extremely coarse heuristic for determining whether or not a given attribute
// name is important for identifying a resource. In the future, this may be
// replaced by a flag in the schema, but for now this is likely to be good
// enough.
func identifyingAttribute(name string, attrSchema *configschema.Attribute) bool {
return name == "id" || name == "tags" || name == "name"
}
func (p *blockBodyDiffPrinter) writeSkippedAttr(skipped, indent int) {
if skipped > 0 {
noun := "attributes"
if skipped == 1 {
noun = "attribute"
}
p.buf.WriteString("\n")
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), skipped, noun))
}
}
func (p *blockBodyDiffPrinter) writeSkippedElems(skipped, indent int) {
if skipped > 0 {
noun := "elements"
if skipped == 1 {
noun = "element"
}
p.buf.WriteString(strings.Repeat(" ", indent))
p.buf.WriteString(fmt.Sprintf(p.color.Color("[dark_gray]# (%d unchanged %s hidden)[reset]"), skipped, noun))
p.buf.WriteString("\n")
}
}