terraform/terraform/resource_address.go

619 lines
18 KiB
Go
Raw Normal View History

package terraform
import (
"fmt"
"reflect"
"regexp"
"strconv"
"strings"
"github.com/hashicorp/terraform/addrs"
"github.com/hashicorp/terraform/configs"
)
// ResourceAddress is a way of identifying an individual resource (or,
// eventually, a subset of resources) within the state. It is used for Targets.
type ResourceAddress struct {
// Addresses a resource falling somewhere in the module path
// When specified alone, addresses all resources within a module path
Path []string
// Addresses a specific resource that occurs in a list
Index int
InstanceType InstanceType
InstanceTypeSet bool
Name string
Type string
Mode ResourceMode // significant only if InstanceTypeSet
}
// Copy returns a copy of this ResourceAddress
func (r *ResourceAddress) Copy() *ResourceAddress {
if r == nil {
return nil
}
n := &ResourceAddress{
Path: make([]string, 0, len(r.Path)),
Index: r.Index,
InstanceType: r.InstanceType,
Name: r.Name,
Type: r.Type,
Mode: r.Mode,
}
2017-09-13 15:04:55 +02:00
n.Path = append(n.Path, r.Path...)
return n
}
// String outputs the address that parses into this address.
func (r *ResourceAddress) String() string {
var result []string
for _, p := range r.Path {
result = append(result, "module", p)
}
switch r.Mode {
case ManagedResourceMode:
// nothing to do
case DataResourceMode:
result = append(result, "data")
default:
panic(fmt.Errorf("unsupported resource mode %s", r.Mode))
}
if r.Type != "" {
result = append(result, r.Type)
}
if r.Name != "" {
name := r.Name
if r.InstanceTypeSet {
switch r.InstanceType {
case TypePrimary:
name += ".primary"
case TypeDeposed:
name += ".deposed"
case TypeTainted:
name += ".tainted"
}
}
if r.Index >= 0 {
name += fmt.Sprintf("[%d]", r.Index)
}
result = append(result, name)
}
return strings.Join(result, ".")
}
// HasResourceSpec returns true if the address has a resource spec, as
// defined in the documentation:
// https://www.terraform.io/docs/internals/resource-addressing.html
// In particular, this returns false if the address contains only
// a module path, thus addressing the entire module.
func (r *ResourceAddress) HasResourceSpec() bool {
return r.Type != "" && r.Name != ""
}
// WholeModuleAddress returns the resource address that refers to all
// resources in the same module as the receiver address.
func (r *ResourceAddress) WholeModuleAddress() *ResourceAddress {
return &ResourceAddress{
Path: r.Path,
Index: -1,
InstanceTypeSet: false,
}
}
// MatchesResourceConfig returns true if the receiver matches the given
// configuration resource within the given _static_ module path. Note that
// the module path in a resource address is a _dynamic_ module path, and
// multiple dynamic resource paths may map to a single static path if
// count and for_each are in use on module calls.
//
// Since resource configuration blocks represent all of the instances of
// a multi-instance resource, the index of the address (if any) is not
// considered.
func (r *ResourceAddress) MatchesResourceConfig(path addrs.Module, rc *configs.Resource) bool {
if r.HasResourceSpec() {
// FIXME: Some ugliness while we are between worlds. Functionality
// in "addrs" should eventually replace this ResourceAddress idea
// completely, but for now we'll need to translate to the old
// way of representing resource modes.
switch r.Mode {
case ManagedResourceMode:
if rc.Mode != addrs.ManagedResourceMode {
return false
}
case DataResourceMode:
if rc.Mode != addrs.DataResourceMode {
return false
}
}
if r.Type != rc.Type || r.Name != rc.Name {
return false
}
}
addrPath := r.Path
// normalize
if len(addrPath) == 0 {
addrPath = nil
}
if len(path) == 0 {
path = nil
}
rawPath := []string(path)
return reflect.DeepEqual(addrPath, rawPath)
}
2016-09-14 02:52:09 +02:00
// stateId returns the ID that this resource should be entered with
// in the state. This is also used for diffs. In the future, we'd like to
// move away from this string field so I don't export this.
func (r *ResourceAddress) stateId() string {
result := fmt.Sprintf("%s.%s", r.Type, r.Name)
switch r.Mode {
case ManagedResourceMode:
2016-09-14 02:52:09 +02:00
// Done
case DataResourceMode:
2016-09-14 02:52:09 +02:00
result = fmt.Sprintf("data.%s", result)
default:
panic(fmt.Errorf("unknown resource mode: %s", r.Mode))
}
if r.Index >= 0 {
result += fmt.Sprintf(".%d", r.Index)
}
2016-09-14 02:52:09 +02:00
return result
}
// parseResourceAddressInternal parses the somewhat bespoke resource
// identifier used in states and diffs, such as "instance.name.0".
func parseResourceAddressInternal(s string) (*ResourceAddress, error) {
// Split based on ".". Every resource address should have at least two
// elements (type and name).
parts := strings.Split(s, ".")
if len(parts) < 2 || len(parts) > 4 {
return nil, fmt.Errorf("Invalid internal resource address format: %s", s)
}
// Data resource if we have at least 3 parts and the first one is data
mode := ManagedResourceMode
if len(parts) > 2 && parts[0] == "data" {
mode = DataResourceMode
parts = parts[1:]
}
// If we're not a data resource and we have more than 3, then it is an error
if len(parts) > 3 && mode != DataResourceMode {
return nil, fmt.Errorf("Invalid internal resource address format: %s", s)
}
// Build the parts of the resource address that are guaranteed to exist
addr := &ResourceAddress{
Type: parts[0],
Name: parts[1],
Index: -1,
InstanceType: TypePrimary,
Mode: mode,
}
// If we have more parts, then we have an index. Parse that.
if len(parts) > 2 {
idx, err := strconv.ParseInt(parts[2], 0, 0)
if err != nil {
return nil, fmt.Errorf("Error parsing resource address %q: %s", s, err)
}
addr.Index = int(idx)
}
return addr, nil
}
func ParseResourceAddress(s string) (*ResourceAddress, error) {
matches, err := tokenizeResourceAddress(s)
if err != nil {
return nil, err
}
mode := ManagedResourceMode
if matches["data_prefix"] != "" {
mode = DataResourceMode
}
resourceIndex, err := ParseResourceIndex(matches["index"])
if err != nil {
return nil, err
}
instanceType, err := ParseInstanceType(matches["instance_type"])
if err != nil {
return nil, err
}
path := ParseResourcePath(matches["path"])
// not allowed to say "data." without a type following
if mode == DataResourceMode && matches["type"] == "" {
return nil, fmt.Errorf(
"invalid resource address %q: must target specific data instance",
s,
)
}
return &ResourceAddress{
Path: path,
Index: resourceIndex,
InstanceType: instanceType,
InstanceTypeSet: matches["instance_type"] != "",
Name: matches["name"],
Type: matches["type"],
Mode: mode,
}, nil
}
// ParseResourceAddressForInstanceDiff creates a ResourceAddress for a
// resource name as described in a module diff.
//
// For historical reasons a different addressing format is used in this
// context. The internal format should not be shown in the UI and instead
// this function should be used to translate to a ResourceAddress and
// then, where appropriate, use the String method to produce a canonical
// resource address string for display in the UI.
//
// The given path slice must be empty (or nil) for the root module, and
// otherwise consist of a sequence of module names traversing down into
// the module tree. If a non-nil path is provided, the caller must not
// modify its underlying array after passing it to this function.
func ParseResourceAddressForInstanceDiff(path []string, key string) (*ResourceAddress, error) {
addr, err := parseResourceAddressInternal(key)
if err != nil {
return nil, err
}
addr.Path = path
return addr, nil
}
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
// NewLegacyResourceAddress creates a ResourceAddress from a new-style
// addrs.AbsResource value.
//
// This is provided for shimming purposes so that we can still easily call into
// older functions that expect the ResourceAddress type.
func NewLegacyResourceAddress(addr addrs.AbsResource) *ResourceAddress {
ret := &ResourceAddress{
Type: addr.Resource.Type,
Name: addr.Resource.Name,
}
switch addr.Resource.Mode {
case addrs.ManagedResourceMode:
ret.Mode = ManagedResourceMode
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
case addrs.DataResourceMode:
ret.Mode = DataResourceMode
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
default:
panic(fmt.Errorf("cannot shim %s to legacy ResourceMode value", addr.Resource.Mode))
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
}
path := make([]string, len(addr.Module))
for i, step := range addr.Module {
if step.InstanceKey != addrs.NoKey {
// At the time of writing this can't happen because we don't
// ket generate keyed module instances. This legacy codepath must
// be removed before we can support "count" and "for_each" for
// modules.
panic(fmt.Errorf("cannot shim module instance step with key %#v to legacy ResourceAddress.Path", step.InstanceKey))
}
path[i] = step.Name
}
ret.Path = path
ret.Index = -1
return ret
}
// NewLegacyResourceInstanceAddress creates a ResourceAddress from a new-style
// addrs.AbsResource value.
//
// This is provided for shimming purposes so that we can still easily call into
// older functions that expect the ResourceAddress type.
func NewLegacyResourceInstanceAddress(addr addrs.AbsResourceInstance) *ResourceAddress {
ret := &ResourceAddress{
Type: addr.Resource.Resource.Type,
Name: addr.Resource.Resource.Name,
}
switch addr.Resource.Resource.Mode {
case addrs.ManagedResourceMode:
ret.Mode = ManagedResourceMode
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
case addrs.DataResourceMode:
ret.Mode = DataResourceMode
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
default:
panic(fmt.Errorf("cannot shim %s to legacy ResourceMode value", addr.Resource.Resource.Mode))
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
}
path := make([]string, len(addr.Module))
for i, step := range addr.Module {
if step.InstanceKey != addrs.NoKey {
// At the time of writing this can't happen because we don't
// ket generate keyed module instances. This legacy codepath must
// be removed before we can support "count" and "for_each" for
// modules.
panic(fmt.Errorf("cannot shim module instance step with key %#v to legacy ResourceAddress.Path", step.InstanceKey))
}
path[i] = step.Name
}
ret.Path = path
if addr.Resource.Key == addrs.NoKey {
ret.Index = -1
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
} else if ik, ok := addr.Resource.Key.(addrs.IntKey); ok {
ret.Index = int(ik)
2019-06-12 17:07:32 +02:00
} else if _, ok := addr.Resource.Key.(addrs.StringKey); ok {
ret.Index = -1
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
} else {
panic(fmt.Errorf("cannot shim resource instance with key %#v to legacy ResourceAddress.Index", addr.Resource.Key))
}
return ret
}
// AbsResourceInstanceAddr converts the receiver, a legacy resource address, to
// the new resource address type addrs.AbsResourceInstance.
//
// This method can be used only on an address that has a resource specification.
// It will panic if called on a module-path-only ResourceAddress. Use
// method HasResourceSpec to check before calling, in contexts where it is
// unclear.
//
// addrs.AbsResourceInstance does not represent the "tainted" and "deposed"
// states, and so if these are present on the receiver then they are discarded.
//
// This is provided for shimming purposes so that we can easily adapt functions
// that are returning the legacy ResourceAddress type, for situations where
// the new type is required.
func (addr *ResourceAddress) AbsResourceInstanceAddr() addrs.AbsResourceInstance {
if !addr.HasResourceSpec() {
panic("AbsResourceInstanceAddr called on ResourceAddress with no resource spec")
}
ret := addrs.AbsResourceInstance{
Module: addr.ModuleInstanceAddr(),
Resource: addrs.ResourceInstance{
Resource: addrs.Resource{
Type: addr.Type,
Name: addr.Name,
},
},
}
switch addr.Mode {
case ManagedResourceMode:
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
ret.Resource.Resource.Mode = addrs.ManagedResourceMode
case DataResourceMode:
terraform: ugly huge change to weave in new HCL2-oriented types Due to how deeply the configuration types go into Terraform Core, there isn't a great way to switch out to HCL2 gradually. As a consequence, this huge commit gets us from the old state to a _compilable_ new state, but does not yet attempt to fix any tests and has a number of known missing parts and bugs. We will continue to iterate on this in forthcoming commits, heading back towards passing tests and making Terraform fully-functional again. The three main goals here are: - Use the configuration models from the "configs" package instead of the older models in the "config" package, which is now deprecated and preserved only to help us write our migration tool. - Do expression inspection and evaluation using the functionality of the new "lang" package, instead of the Interpolator type and related functionality in the main "terraform" package. - Represent addresses of various objects using types in the addrs package, rather than hand-constructed strings. This is not critical to support the above, but was a big help during the implementation of these other points since it made it much more explicit what kind of address is expected in each context. Since our new packages are built to accommodate some future planned features that are not yet implemented (e.g. the "for_each" argument on resources, "count"/"for_each" on modules), and since there's still a fair amount of functionality still using old-style APIs, there is a moderate amount of shimming here to connect new assumptions with old, hopefully in a way that makes it easier to find and eliminate these shims later. I apologize in advance to the person who inevitably just found this huge commit while spelunking through the commit history.
2018-04-30 19:33:53 +02:00
ret.Resource.Resource.Mode = addrs.DataResourceMode
default:
panic(fmt.Errorf("cannot shim %s to addrs.ResourceMode value", addr.Mode))
}
if addr.Index != -1 {
ret.Resource.Key = addrs.IntKey(addr.Index)
}
return ret
}
// ModuleInstanceAddr returns the module path portion of the receiver as a
// addrs.ModuleInstance value.
func (addr *ResourceAddress) ModuleInstanceAddr() addrs.ModuleInstance {
path := make(addrs.ModuleInstance, len(addr.Path))
for i, name := range addr.Path {
path[i] = addrs.ModuleInstanceStep{Name: name}
}
return path
}
// Contains returns true if and only if the given node is contained within
// the receiver.
//
// Containment is defined in terms of the module and resource heirarchy:
// a resource is contained within its module and any ancestor modules,
// an indexed resource instance is contained with the unindexed resource, etc.
func (addr *ResourceAddress) Contains(other *ResourceAddress) bool {
ourPath := addr.Path
givenPath := other.Path
if len(givenPath) < len(ourPath) {
return false
}
for i := range ourPath {
if ourPath[i] != givenPath[i] {
return false
}
}
// If the receiver is a whole-module address then the path prefix
// matching is all we need.
if !addr.HasResourceSpec() {
return true
}
if addr.Type != other.Type || addr.Name != other.Name || addr.Mode != other.Mode {
return false
}
if addr.Index != -1 && addr.Index != other.Index {
return false
}
if addr.InstanceTypeSet && (addr.InstanceTypeSet != other.InstanceTypeSet || addr.InstanceType != other.InstanceType) {
return false
}
return true
}
// Equals returns true if the receiver matches the given address.
//
// The name of this method is a misnomer, since it doesn't test for exact
// equality. Instead, it tests that the _specified_ parts of each
// address match, treating any unspecified parts as wildcards.
//
// See also Contains, which takes a more heirarchical approach to comparing
// addresses.
func (addr *ResourceAddress) Equals(raw interface{}) bool {
other, ok := raw.(*ResourceAddress)
if !ok {
return false
}
2015-10-08 14:48:04 +02:00
pathMatch := len(addr.Path) == 0 && len(other.Path) == 0 ||
reflect.DeepEqual(addr.Path, other.Path)
2015-10-08 14:48:04 +02:00
indexMatch := addr.Index == -1 ||
other.Index == -1 ||
2015-10-08 14:48:04 +02:00
addr.Index == other.Index
2015-10-08 14:48:04 +02:00
nameMatch := addr.Name == "" ||
other.Name == "" ||
2015-10-08 14:48:04 +02:00
addr.Name == other.Name
2015-10-08 14:48:04 +02:00
typeMatch := addr.Type == "" ||
other.Type == "" ||
2015-10-08 14:48:04 +02:00
addr.Type == other.Type
// mode is significant only when type is set
modeMatch := addr.Type == "" ||
other.Type == "" ||
addr.Mode == other.Mode
2015-10-08 14:48:04 +02:00
return pathMatch &&
indexMatch &&
addr.InstanceType == other.InstanceType &&
nameMatch &&
typeMatch &&
modeMatch
}
// Less returns true if and only if the receiver should be sorted before
// the given address when presenting a list of resource addresses to
// an end-user.
//
// This sort uses lexicographic sorting for most components, but uses
// numeric sort for indices, thus causing index 10 to sort after
// index 9, rather than after index 1.
func (addr *ResourceAddress) Less(other *ResourceAddress) bool {
switch {
case len(addr.Path) != len(other.Path):
return len(addr.Path) < len(other.Path)
case !reflect.DeepEqual(addr.Path, other.Path):
// If the two paths are the same length but don't match, we'll just
// cheat and compare the string forms since it's easier than
// comparing all of the path segments in turn, and lexicographic
// comparison is correct for the module path portion.
addrStr := addr.String()
otherStr := other.String()
return addrStr < otherStr
case addr.Mode != other.Mode:
return addr.Mode == DataResourceMode
case addr.Type != other.Type:
return addr.Type < other.Type
case addr.Name != other.Name:
return addr.Name < other.Name
case addr.Index != other.Index:
// Since "Index" is -1 for an un-indexed address, this also conveniently
// sorts unindexed addresses before indexed ones, should they both
// appear for some reason.
return addr.Index < other.Index
case addr.InstanceTypeSet != other.InstanceTypeSet:
return !addr.InstanceTypeSet
case addr.InstanceType != other.InstanceType:
// InstanceType is actually an enum, so this is just an arbitrary
// sort based on the enum numeric values, and thus not particularly
// meaningful.
return addr.InstanceType < other.InstanceType
default:
return false
}
}
func ParseResourceIndex(s string) (int, error) {
if s == "" {
return -1, nil
}
return strconv.Atoi(s)
}
func ParseResourcePath(s string) []string {
if s == "" {
return nil
}
parts := strings.Split(s, ".")
path := make([]string, 0, len(parts))
for _, s := range parts {
// Due to the limitations of the regexp match below, the path match has
// some noise in it we have to filter out :|
if s == "" || s == "module" {
continue
}
path = append(path, s)
}
return path
}
func ParseInstanceType(s string) (InstanceType, error) {
switch s {
case "", "primary":
return TypePrimary, nil
case "deposed":
return TypeDeposed, nil
case "tainted":
return TypeTainted, nil
default:
return TypeInvalid, fmt.Errorf("Unexpected value for InstanceType field: %q", s)
}
}
func tokenizeResourceAddress(s string) (map[string]string, error) {
// Example of portions of the regexp below using the
// string "aws_instance.web.tainted[1]"
re := regexp.MustCompile(`\A` +
// "module.foo.module.bar" (optional)
`(?P<path>(?:module\.(?P<module_name>[^.]+)\.?)*)` +
// possibly "data.", if targeting is a data resource
`(?P<data_prefix>(?:data\.)?)` +
// "aws_instance.web" (optional when module path specified)
`(?:(?P<type>[^.]+)\.(?P<name>[^.[]+))?` +
// "tainted" (optional, omission implies: "primary")
`(?:\.(?P<instance_type>\w+))?` +
// "1" (optional, omission implies: "0")
`(?:\[(?P<index>\d+)\])?` +
`\z`)
groupNames := re.SubexpNames()
rawMatches := re.FindAllStringSubmatch(s, -1)
if len(rawMatches) != 1 {
return nil, fmt.Errorf("invalid resource address %q", s)
}
matches := make(map[string]string)
for i, m := range rawMatches[0] {
matches[groupNames[i]] = m
}
return matches, nil
}