terraform/internal/addrs/module_source.go

496 lines
20 KiB
Go
Raw Permalink Normal View History

package addrs
import (
"fmt"
"path"
"regexp"
"strings"
svchost "github.com/hashicorp/terraform-svchost"
"github.com/hashicorp/terraform/internal/getmodules"
)
// ModuleSource is the general type for all three of the possible module source
// address types. The concrete implementations of this are ModuleSourceLocal,
// ModuleSourceRegistry, and ModuleSourceRemote.
type ModuleSource interface {
// String returns a full representation of the address, including any
// additional components that are typically implied by omission in
// user-written addresses.
//
// We typically use this longer representation in error message, in case
// the inclusion of normally-omitted components is helpful in debugging
// unexpected behavior.
String() string
// ForDisplay is similar to String but instead returns a representation of
// the idiomatic way to write the address in configuration, omitting
// components that are commonly just implied in addresses written by
// users.
//
// We typically use this shorter representation in informational messages,
// such as the note that we're about to start downloading a package.
ForDisplay() string
moduleSource()
}
var _ ModuleSource = ModuleSourceLocal("")
var _ ModuleSource = ModuleSourceRegistry{}
var _ ModuleSource = ModuleSourceRemote{}
var moduleSourceLocalPrefixes = []string{
"./",
"../",
".\\",
"..\\",
}
// ParseModuleSource parses a module source address as given in the "source"
// argument inside a "module" block in the configuration.
//
// For historical reasons this syntax is a bit overloaded, supporting three
// different address types:
// - Local paths starting with either ./ or ../, which are special because
// Terraform considers them to belong to the same "package" as the caller.
// - Module registry addresses, given as either NAMESPACE/NAME/SYSTEM or
// HOST/NAMESPACE/NAME/SYSTEM, in which case the remote registry serves
// as an indirection over the third address type that follows.
// - Various URL-like and other heuristically-recognized strings which
// we currently delegate to the external library go-getter.
//
// There is some ambiguity between the module registry addresses and go-getter's
// very liberal heuristics and so this particular function will typically treat
// an invalid registry address as some other sort of remote source address
// rather than returning an error. If you know that you're expecting a
// registry address in particular, use ParseModuleSourceRegistry instead, which
// can therefore expose more detailed error messages about registry address
// parsing in particular.
func ParseModuleSource(raw string) (ModuleSource, error) {
if isModuleSourceLocal(raw) {
localAddr, err := parseModuleSourceLocal(raw)
if err != nil {
// This is to make sure we really return a nil ModuleSource in
// this case, rather than an interface containing the zero
// value of ModuleSourceLocal.
return nil, err
}
return localAddr, nil
}
// For historical reasons, whether an address is a registry
// address is defined only by whether it can be successfully
// parsed as one, and anything else must fall through to be
// parsed as a direct remote source, where go-getter might
// then recognize it as a filesystem path. This is odd
// but matches behavior we've had since Terraform v0.10 which
// existing modules may be relying on.
// (Notice that this means that there's never any path where
// the registry source parse error gets returned to the caller,
// which is annoying but has been true for many releases
// without it posing a serious problem in practice.)
if ret, err := ParseModuleSourceRegistry(raw); err == nil {
return ret, nil
}
// If we get down here then we treat everything else as a
// remote address. In practice there's very little that
// go-getter doesn't consider invalid input, so even invalid
// nonsense will probably interpreted as _something_ here
// and then fail during installation instead. We can't
// really improve this situation for historical reasons.
Refactoring of module source addresses and module installation It's been a long while since we gave close attention to the codepaths for module source address parsing and external module package installation. Due to their age, these codepaths often diverged from our modern practices such as representing address types in the addrs package, and encapsulating package installation details only in a particular location. In particular, this refactor makes source address parsing a separate step from module installation, which therefore makes the result of that parsing available to other Terraform subsystems which work with the configuration representation objects. This also presented the opportunity to better encapsulate our use of go-getter into a new package "getmodules" (echoing "getproviders"), which is intended to be the only part of Terraform that directly interacts with go-getter. This is largely just a refactor of the existing functionality into a new code organization, but there is one notable change in behavior here: the source address parsing now happens during configuration loading rather than module installation, which may cause errors about invalid addresses to be returned in different situations than before. That counts as backward compatible because we only promise to remain compatible with configurations that are _valid_, which means that they can be initialized, planned, and applied without any errors. This doesn't introduce any new error cases, and instead just makes a pre-existing error case be detected earlier. Our module registry client is still using its own special module address type from registry/regsrc for now, with a small shim from the new addrs.ModuleSourceRegistry type. Hopefully in a later commit we'll also rework the registry client to work with the new address type, but this commit is already big enough as it is.
2021-05-28 04:24:59 +02:00
remoteAddr, err := parseModuleSourceRemote(raw)
if err != nil {
// This is to make sure we really return a nil ModuleSource in
// this case, rather than an interface containing the zero
// value of ModuleSourceRemote.
return nil, err
}
return remoteAddr, nil
}
// ModuleSourceLocal is a ModuleSource representing a local path reference
// from the caller's directory to the callee's directory within the same
// module package.
//
// A "module package" here means a set of modules distributed together in
// the same archive, repository, or similar. That's a significant distinction
// because we always download and cache entire module packages at once,
// and then create relative references within the same directory in order
// to ensure all modules in the package are looking at a consistent filesystem
// layout. We also assume that modules within a package are maintained together,
// which means that cross-cutting maintenence across all of them would be
// possible.
//
// The actual value of a ModuleSourceLocal is a normalized relative path using
// forward slashes, even on operating systems that have other conventions,
// because we're representing traversal within the logical filesystem
// represented by the containing package, not actually within the physical
// filesystem we unpacked the package into. We should typically not construct
// ModuleSourceLocal values directly, except in tests where we can ensure
// the value meets our assumptions. Use ParseModuleSource instead if the
// input string is not hard-coded in the program.
type ModuleSourceLocal string
func parseModuleSourceLocal(raw string) (ModuleSourceLocal, error) {
// As long as we have a suitable prefix (detected by ParseModuleSource)
// there is no failure case for local paths: we just use the "path"
// package's cleaning logic to remove any redundant "./" and "../"
// sequences and any duplicate slashes and accept whatever that
// produces.
// Although using backslashes (Windows-style) is non-idiomatic, we do
// allow it and just normalize it away, so the rest of Terraform will
// only see the forward-slash form.
if strings.Contains(raw, `\`) {
// Note: We use string replacement rather than filepath.ToSlash
// here because the filepath package behavior varies by current
// platform, but we want to interpret configured paths the same
// across all platforms: these are virtual paths within a module
// package, not physical filesystem paths.
raw = strings.ReplaceAll(raw, `\`, "/")
}
// Note that we could've historically blocked using "//" in a path here
// in order to avoid confusion with the subdir syntax in remote addresses,
// but we historically just treated that as the same as a single slash
// and so we continue to do that now for compatibility. Clean strips those
// out and reduces them to just a single slash.
clean := path.Clean(raw)
// However, we do need to keep a single "./" on the front if it isn't
// a "../" path, or else it would be ambigous with the registry address
// syntax.
if !strings.HasPrefix(clean, "../") {
clean = "./" + clean
}
return ModuleSourceLocal(clean), nil
}
func isModuleSourceLocal(raw string) bool {
for _, prefix := range moduleSourceLocalPrefixes {
if strings.HasPrefix(raw, prefix) {
return true
}
}
return false
}
func (s ModuleSourceLocal) moduleSource() {}
func (s ModuleSourceLocal) String() string {
// We assume that our underlying string was already normalized at
// construction, so we just return it verbatim.
return string(s)
}
func (s ModuleSourceLocal) ForDisplay() string {
Refactoring of module source addresses and module installation It's been a long while since we gave close attention to the codepaths for module source address parsing and external module package installation. Due to their age, these codepaths often diverged from our modern practices such as representing address types in the addrs package, and encapsulating package installation details only in a particular location. In particular, this refactor makes source address parsing a separate step from module installation, which therefore makes the result of that parsing available to other Terraform subsystems which work with the configuration representation objects. This also presented the opportunity to better encapsulate our use of go-getter into a new package "getmodules" (echoing "getproviders"), which is intended to be the only part of Terraform that directly interacts with go-getter. This is largely just a refactor of the existing functionality into a new code organization, but there is one notable change in behavior here: the source address parsing now happens during configuration loading rather than module installation, which may cause errors about invalid addresses to be returned in different situations than before. That counts as backward compatible because we only promise to remain compatible with configurations that are _valid_, which means that they can be initialized, planned, and applied without any errors. This doesn't introduce any new error cases, and instead just makes a pre-existing error case be detected earlier. Our module registry client is still using its own special module address type from registry/regsrc for now, with a small shim from the new addrs.ModuleSourceRegistry type. Hopefully in a later commit we'll also rework the registry client to work with the new address type, but this commit is already big enough as it is.
2021-05-28 04:24:59 +02:00
return string(s)
}
// ModuleSourceRegistry is a ModuleSource representing a module listed in a
// Terraform module registry.
//
// A registry source isn't a direct source location but rather an indirection
// over a ModuleSourceRemote. The job of a registry is to translate the
// combination of a ModuleSourceRegistry and a module version number into
// a concrete ModuleSourceRemote that Terraform will then download and
// install.
type ModuleSourceRegistry struct {
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
// PackageAddr is the registry package that the target module belongs to.
// The module installer must translate this into a ModuleSourceRemote
// using the registry API and then take that underlying address's
// PackageAddr in order to find the actual package location.
PackageAddr ModuleRegistryPackage
// If Subdir is non-empty then it represents a sub-directory within the
// remote package that the registry address eventually resolves to.
// This will ultimately become the suffix of the Subdir of the
// ModuleSourceRemote that the registry address translates to.
//
// Subdir uses a normalized forward-slash-based path syntax within the
// virtual filesystem represented by the final package. It will never
// include `../` or `./` sequences.
Subdir string
}
// DefaultModuleRegistryHost is the hostname used for registry-based module
// source addresses that do not have an explicit hostname.
const DefaultModuleRegistryHost = svchost.Hostname("registry.terraform.io")
var moduleRegistryNamePattern = regexp.MustCompile("^[0-9A-Za-z](?:[0-9A-Za-z-_]{0,62}[0-9A-Za-z])?$")
var moduleRegistryTargetSystemPattern = regexp.MustCompile("^[0-9a-z]{1,64}$")
// ParseModuleSourceRegistry is a variant of ParseModuleSource which only
// accepts module registry addresses, and will reject any other address type.
//
// Use this instead of ParseModuleSource if you know from some other surrounding
// context that an address is intended to be a registry address rather than
// some other address type, which will then allow for better error reporting
// due to the additional information about user intent.
func ParseModuleSourceRegistry(raw string) (ModuleSource, error) {
// Before we delegate to the "real" function we'll just make sure this
// doesn't look like a local source address, so we can return a better
// error message for that situation.
if isModuleSourceLocal(raw) {
return ModuleSourceRegistry{}, fmt.Errorf("can't use local directory %q as a module registry address", raw)
}
ret, err := parseModuleSourceRegistry(raw)
if err != nil {
// This is to make sure we return a nil ModuleSource, rather than
// a non-nil ModuleSource containing a zero-value ModuleSourceRegistry.
return nil, err
}
return ret, nil
}
func parseModuleSourceRegistry(raw string) (ModuleSourceRegistry, error) {
var err error
var subDir string
raw, subDir = getmodules.SplitPackageSubdir(raw)
if strings.HasPrefix(subDir, "../") {
return ModuleSourceRegistry{}, fmt.Errorf("subdirectory path %q leads outside of the module package", subDir)
}
parts := strings.Split(raw, "/")
// A valid registry address has either three or four parts, because the
// leading hostname part is optional.
if len(parts) != 3 && len(parts) != 4 {
return ModuleSourceRegistry{}, fmt.Errorf("a module registry source address must have either three or four slash-separated components")
}
host := DefaultModuleRegistryHost
if len(parts) == 4 {
host, err = svchost.ForComparison(parts[0])
if err != nil {
// The svchost library doesn't produce very good error messages to
// return to an end-user, so we'll use some custom ones here.
switch {
case strings.Contains(parts[0], "--"):
// Looks like possibly punycode, which we don't allow here
// to ensure that source addresses are written readably.
return ModuleSourceRegistry{}, fmt.Errorf("invalid module registry hostname %q; internationalized domain names must be given as direct unicode characters, not in punycode", parts[0])
default:
return ModuleSourceRegistry{}, fmt.Errorf("invalid module registry hostname %q", parts[0])
}
}
if !strings.Contains(host.String(), ".") {
return ModuleSourceRegistry{}, fmt.Errorf("invalid module registry hostname: must contain at least one dot")
}
// Discard the hostname prefix now that we've processed it
parts = parts[1:]
}
ret := ModuleSourceRegistry{
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
PackageAddr: ModuleRegistryPackage{
Host: host,
},
Subdir: subDir,
}
if host == svchost.Hostname("github.com") || host == svchost.Hostname("bitbucket.org") {
return ret, fmt.Errorf("can't use %q as a module registry host, because it's reserved for installing directly from version control repositories", host)
}
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
if ret.PackageAddr.Namespace, err = parseModuleRegistryName(parts[0]); err != nil {
if strings.Contains(parts[0], ".") {
// Seems like the user omitted one of the latter components in
// an address with an explicit hostname.
return ret, fmt.Errorf("source address must have three more components after the hostname: the namespace, the name, and the target system")
}
return ret, fmt.Errorf("invalid namespace %q: %s", parts[0], err)
}
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
if ret.PackageAddr.Name, err = parseModuleRegistryName(parts[1]); err != nil {
return ret, fmt.Errorf("invalid module name %q: %s", parts[1], err)
}
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
if ret.PackageAddr.TargetSystem, err = parseModuleRegistryTargetSystem(parts[2]); err != nil {
if strings.Contains(parts[2], "?") {
// The user was trying to include a query string, probably?
return ret, fmt.Errorf("module registry addresses may not include a query string portion")
}
return ret, fmt.Errorf("invalid target system %q: %s", parts[2], err)
}
return ret, nil
}
// parseModuleRegistryName validates and normalizes a string in either the
// "namespace" or "name" position of a module registry source address.
func parseModuleRegistryName(given string) (string, error) {
// Similar to the names in provider source addresses, we defined these
// to be compatible with what filesystems and typical remote systems
// like GitHub allow in names. Unfortunately we didn't end up defining
// these exactly equivalently: provider names can only use dashes as
// punctuation, whereas module names can use underscores. So here we're
// using some regular expressions from the original module source
// implementation, rather than using the IDNA rules as we do in
// ParseProviderPart.
if !moduleRegistryNamePattern.MatchString(given) {
return "", fmt.Errorf("must be between one and 64 characters, including ASCII letters, digits, dashes, and underscores, where dashes and underscores may not be the prefix or suffix")
}
// We also skip normalizing the name to lowercase, because we historically
// didn't do that and so existing module registries might be doing
// case-sensitive matching.
return given, nil
}
// parseModuleRegistryTargetSystem validates and normalizes a string in the
// "target system" position of a module registry source address. This is
// what we historically called "provider" but never actually enforced as
// being a provider address, and now _cannot_ be a provider address because
// provider addresses have three slash-separated components of their own.
func parseModuleRegistryTargetSystem(given string) (string, error) {
// Similar to the names in provider source addresses, we defined these
// to be compatible with what filesystems and typical remote systems
// like GitHub allow in names. Unfortunately we didn't end up defining
// these exactly equivalently: provider names can't use dashes or
// underscores. So here we're using some regular expressions from the
// original module source implementation, rather than using the IDNA rules
// as we do in ParseProviderPart.
if !moduleRegistryTargetSystemPattern.MatchString(given) {
return "", fmt.Errorf("must be between one and 64 ASCII letters or digits")
}
// We also skip normalizing the name to lowercase, because we historically
// didn't do that and so existing module registries might be doing
// case-sensitive matching.
return given, nil
}
func (s ModuleSourceRegistry) moduleSource() {}
func (s ModuleSourceRegistry) String() string {
if s.Subdir != "" {
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
return s.PackageAddr.String() + "//" + s.Subdir
}
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
return s.PackageAddr.String()
}
func (s ModuleSourceRegistry) ForDisplay() string {
if s.Subdir != "" {
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
return s.PackageAddr.ForDisplay() + "//" + s.Subdir
}
addrs: ModuleRegistryPackage for representing module registry packages Previously we had a separation between ModuleSourceRemote and ModulePackage as a way to represent within the type system that there's an important difference between a module source address and a package address, because module packages often contain multiple modules and so a ModuleSourceRemote combines a ModulePackage with a subdirectory to represent one specific module. This commit applies that same strategy to ModuleSourceRegistry, creating a new type ModuleRegistryPackage to represent the different sort of package that we use for registry modules. Again, the main goal here is to try to reflect the conceptual modelling more directly in the type system so that we can more easily verify that uses of these different address types are correct. To make use of that, I've also lightly reworked initwd's module installer to use addrs.ModuleRegistryPackage directly, instead of a string representation thereof. This was in response to some earlier commits where I found myself accidentally mixing up package addresses and source addresses in the installRegistryModule method; with this new organization those bugs would've been caught at compile time, rather than only at unit and integration testing time. While in the area anyway, I also took this opportunity to fix some historical confusing names of fields in initwd.ModuleInstaller, to be clearer that they are only for registry packages and not for all module source address types.
2021-06-02 21:26:35 +02:00
return s.PackageAddr.ForDisplay()
}
// ModuleSourceRemote is a ModuleSource representing a remote location from
// which we can retrieve a module package.
//
// A ModuleSourceRemote can optionally include a "subdirectory" path, which
// means that it's selecting a sub-directory of the given package to use as
// the entry point into the package.
type ModuleSourceRemote struct {
// PackageAddr is the address of the remote package that the requested
// module belongs to.
PackageAddr ModulePackage
// If Subdir is non-empty then it represents a sub-directory within the
// remote package which will serve as the entry-point for the package.
//
// Subdir uses a normalized forward-slash-based path syntax within the
// virtual filesystem represented by the final package. It will never
// include `../` or `./` sequences.
Subdir string
}
func parseModuleSourceRemote(raw string) (ModuleSourceRemote, error) {
var subDir string
raw, subDir = getmodules.SplitPackageSubdir(raw)
if strings.HasPrefix(subDir, "../") {
return ModuleSourceRemote{}, fmt.Errorf("subdirectory path %q leads outside of the module package", subDir)
}
// A remote source address is really just a go-getter address resulting
// from go-getter's "detect" phase, which adds on the prefix specifying
// which protocol it should use and possibly also adjusts the
// protocol-specific part into different syntax.
//
// Note that for historical reasons this can potentially do network
// requests in order to disambiguate certain address types, although
// that's a legacy thing that is only for some specific, less-commonly-used
// address types. Most just do local string manipulation. We should
// aim to remove the network requests over time, if possible.
norm, moreSubDir, err := getmodules.NormalizePackageAddress(raw)
if err != nil {
Refactoring of module source addresses and module installation It's been a long while since we gave close attention to the codepaths for module source address parsing and external module package installation. Due to their age, these codepaths often diverged from our modern practices such as representing address types in the addrs package, and encapsulating package installation details only in a particular location. In particular, this refactor makes source address parsing a separate step from module installation, which therefore makes the result of that parsing available to other Terraform subsystems which work with the configuration representation objects. This also presented the opportunity to better encapsulate our use of go-getter into a new package "getmodules" (echoing "getproviders"), which is intended to be the only part of Terraform that directly interacts with go-getter. This is largely just a refactor of the existing functionality into a new code organization, but there is one notable change in behavior here: the source address parsing now happens during configuration loading rather than module installation, which may cause errors about invalid addresses to be returned in different situations than before. That counts as backward compatible because we only promise to remain compatible with configurations that are _valid_, which means that they can be initialized, planned, and applied without any errors. This doesn't introduce any new error cases, and instead just makes a pre-existing error case be detected earlier. Our module registry client is still using its own special module address type from registry/regsrc for now, with a small shim from the new addrs.ModuleSourceRegistry type. Hopefully in a later commit we'll also rework the registry client to work with the new address type, but this commit is already big enough as it is.
2021-05-28 04:24:59 +02:00
// We must pass through the returned error directly here because
// the getmodules package has some special error types it uses
// for certain cases where the UI layer might want to include a
// more helpful error message.
return ModuleSourceRemote{}, err
}
if moreSubDir != "" {
switch {
case subDir != "":
// The detector's own subdir goes first, because the
// subdir we were given is conceptually relative to
// the subdirectory that we just detected.
subDir = path.Join(moreSubDir, subDir)
default:
subDir = path.Clean(moreSubDir)
}
if strings.HasPrefix(subDir, "../") {
// This would suggest a bug in a go-getter detector, but
// we'll catch it anyway to avoid doing something confusing
// downstream.
return ModuleSourceRemote{}, fmt.Errorf("detected subdirectory path %q of %q leads outside of the module package", subDir, norm)
}
}
return ModuleSourceRemote{
PackageAddr: ModulePackage(norm),
Subdir: subDir,
}, nil
}
func (s ModuleSourceRemote) moduleSource() {}
func (s ModuleSourceRemote) String() string {
if s.Subdir != "" {
return s.PackageAddr.String() + "//" + s.Subdir
}
return s.PackageAddr.String()
}
func (s ModuleSourceRemote) ForDisplay() string {
// The two string representations are identical for this address type.
// This isn't really entirely true to the idea of "ForDisplay" since
// it'll often include some additional components added in by the
// go-getter detectors, but we don't have any function to turn a
// "detected" string back into an idiomatic shorthand the user might've
// entered.
return s.String()
}
// FromRegistry can be called on a remote source address that was returned
// from a module registry, passing in the original registry source address
// that the registry was asked about, in order to get the effective final
// remote source address.
//
// Specifically, this method handles the situations where one or both of
// the two addresses contain subdirectory paths, combining both when necessary
// in order to ensure that both the registry's given path and the user's
// given path are both respected.
//
// This will return nonsense if given a registry address other than the one
// that generated the reciever via a registry lookup.
func (s ModuleSourceRemote) FromRegistry(given ModuleSourceRegistry) ModuleSourceRemote {
ret := s // not a pointer, so this is a shallow copy
switch {
case s.Subdir != "" && given.Subdir != "":
ret.Subdir = path.Join(s.Subdir, given.Subdir)
case given.Subdir != "":
ret.Subdir = given.Subdir
}
return ret
}