opentofu/internal/instances/expander.go

528 lines
20 KiB
Go
Raw Normal View History

package instances
import (
"fmt"
"sort"
"sync"
"github.com/hashicorp/terraform/internal/addrs"
"github.com/zclconf/go-cty/cty"
)
// Expander instances serve as a coordination point for gathering object
// repetition values (count and for_each in configuration) and then later
// making use of them to fully enumerate all of the instances of an object.
//
// The two repeatable object types in Terraform are modules and resources.
// Because resources belong to modules and modules can nest inside other
// modules, module expansion in particular has a recursive effect that can
// cause deep objects to expand exponentially. Expander assumes that all
// instances of a module have the same static objects inside, and that they
// differ only in the repetition count for some of those objects.
//
// Expander is a synchronized object whose methods can be safely called
// from concurrent threads of execution. However, it does expect a certain
// sequence of operations which is normally obtained by the caller traversing
// a dependency graph: each object must have its repetition mode set exactly
// once, and this must be done before any calls that depend on the repetition
// mode. In other words, the count or for_each expression value for a module
// must be provided before any object nested directly or indirectly inside
// that module can be expanded. If this ordering is violated, the methods
// will panic to enforce internal consistency.
//
// The Expand* methods of Expander only work directly with modules and with
// resources. Addresses for other objects that nest within modules but
// do not themselves support repetition can be obtained by calling ExpandModule
// with the containing module path and then producing one absolute instance
// address per module instance address returned.
type Expander struct {
mu sync.RWMutex
exps *expanderModule
}
// NewExpander initializes and returns a new Expander, empty and ready to use.
func NewExpander() *Expander {
return &Expander{
exps: newExpanderModule(),
}
}
// SetModuleSingle records that the given module call inside the given parent
// module does not use any repetition arguments and is therefore a singleton.
func (e *Expander) SetModuleSingle(parentAddr addrs.ModuleInstance, callAddr addrs.ModuleCall) {
e.setModuleExpansion(parentAddr, callAddr, expansionSingleVal)
}
// SetModuleCount records that the given module call inside the given parent
// module instance uses the "count" repetition argument, with the given value.
func (e *Expander) SetModuleCount(parentAddr addrs.ModuleInstance, callAddr addrs.ModuleCall, count int) {
e.setModuleExpansion(parentAddr, callAddr, expansionCount(count))
}
// SetModuleForEach records that the given module call inside the given parent
// module instance uses the "for_each" repetition argument, with the given
// map value.
//
// In the configuration language the for_each argument can also accept a set.
// It's the caller's responsibility to convert that into an identity map before
// calling this method.
func (e *Expander) SetModuleForEach(parentAddr addrs.ModuleInstance, callAddr addrs.ModuleCall, mapping map[string]cty.Value) {
e.setModuleExpansion(parentAddr, callAddr, expansionForEach(mapping))
}
// SetResourceSingle records that the given resource inside the given module
// does not use any repetition arguments and is therefore a singleton.
func (e *Expander) SetResourceSingle(moduleAddr addrs.ModuleInstance, resourceAddr addrs.Resource) {
e.setResourceExpansion(moduleAddr, resourceAddr, expansionSingleVal)
}
// SetResourceCount records that the given resource inside the given module
// uses the "count" repetition argument, with the given value.
func (e *Expander) SetResourceCount(moduleAddr addrs.ModuleInstance, resourceAddr addrs.Resource, count int) {
e.setResourceExpansion(moduleAddr, resourceAddr, expansionCount(count))
}
// SetResourceForEach records that the given resource inside the given module
// uses the "for_each" repetition argument, with the given map value.
//
// In the configuration language the for_each argument can also accept a set.
// It's the caller's responsibility to convert that into an identity map before
// calling this method.
func (e *Expander) SetResourceForEach(moduleAddr addrs.ModuleInstance, resourceAddr addrs.Resource, mapping map[string]cty.Value) {
e.setResourceExpansion(moduleAddr, resourceAddr, expansionForEach(mapping))
}
// ExpandModule finds the exhaustive set of module instances resulting from
// the expansion of the given module and all of its ancestor modules.
//
// All of the modules on the path to the identified module must already have
// had their expansion registered using one of the SetModule* methods before
// calling, or this method will panic.
func (e *Expander) ExpandModule(addr addrs.Module) []addrs.ModuleInstance {
return e.expandModule(addr, false)
}
// expandModule allows skipping unexpanded module addresses by setting skipUnknown to true.
// This is used by instances.Set, which is only concerned with the expanded
// instances, and should not panic when looking up unknown addresses.
func (e *Expander) expandModule(addr addrs.Module, skipUnknown bool) []addrs.ModuleInstance {
if len(addr) == 0 {
// Root module is always a singleton.
return singletonRootModule
}
e.mu.RLock()
defer e.mu.RUnlock()
// We're going to be dynamically growing ModuleInstance addresses, so
// we'll preallocate some space to do it so that for typical shallow
// module trees we won't need to reallocate this.
// (moduleInstances does plenty of allocations itself, so the benefit of
// pre-allocating this is marginal but it's not hard to do.)
parentAddr := make(addrs.ModuleInstance, 0, 4)
ret := e.exps.moduleInstances(addr, parentAddr, skipUnknown)
sort.SliceStable(ret, func(i, j int) bool {
return ret[i].Less(ret[j])
})
return ret
}
instances: Non-existing module instance has no resource instances Previously we were treating it as a programming error to ask for the instances of a resource inside an instance of a module that is declared but whose declaration doesn't include the given instance key. However, that's actually a valid situation which can arise if, for example, the user has changed the repetition/expansion mode for an existing module call and so now all of the resource instances addresses it previously contained are "orphaned". To represent that, we'll instead say that an invalid instance key of a declared module behaves as if it contains no resource instances at all, regardless of the configurations of any resources nested inside. This then gives the result needed to successfully detect all of the former resource instances as "orphaned" and plan to destroy them. However, this then introduces a new case for NodePlannableResourceInstanceOrphan.deleteActionReason to deal with: the resource configuration still exists (because configuration isn't aware of individual module/resource instances) but the module instance does not. This actually allows us to resolve, at least partially, a previous missing piece of explaining to the user why the resource instances are planned for deletion in that case, finally allowing us to be explicit to the user that it's because of the module instance being removed, which internally we call plans.ResourceInstanceDeleteBecauseNoModule. Co-authored-by: Alisdair McDiarmid <alisdair@users.noreply.github.com>
2021-12-08 14:40:32 -06:00
// GetDeepestExistingModuleInstance is a funny specialized function for
// determining how many steps we can traverse through the given module instance
// address before encountering an undeclared instance of a declared module.
//
// The result is the longest prefix of the given address which steps only
// through module instances that exist.
//
// All of the modules on the given path must already have had their
// expansion registered using one of the SetModule* methods before calling,
// or this method will panic.
func (e *Expander) GetDeepestExistingModuleInstance(given addrs.ModuleInstance) addrs.ModuleInstance {
exps := e.exps // start with the root module expansions
for i := 0; i < len(given); i++ {
step := given[i]
callName := step.Name
if _, ok := exps.moduleCalls[addrs.ModuleCall{Name: callName}]; !ok {
// This is a bug in the caller, because it should always register
// expansions for an object and all of its ancestors before requesting
// expansion of it.
panic(fmt.Sprintf("no expansion has been registered for %s", given[:i].Child(callName, addrs.NoKey)))
}
var ok bool
exps, ok = exps.childInstances[step]
if !ok {
// We've found a non-existing instance, so we're done.
return given[:i]
}
}
// If we complete the loop above without returning early then the entire
// given address refers to a declared module instance.
return given
}
// ExpandModuleResource finds the exhaustive set of resource instances resulting from
// the expansion of the given resource and all of its containing modules.
//
// All of the modules on the path to the identified resource and the resource
// itself must already have had their expansion registered using one of the
// SetModule*/SetResource* methods before calling, or this method will panic.
func (e *Expander) ExpandModuleResource(moduleAddr addrs.Module, resourceAddr addrs.Resource) []addrs.AbsResourceInstance {
e.mu.RLock()
defer e.mu.RUnlock()
// We're going to be dynamically growing ModuleInstance addresses, so
// we'll preallocate some space to do it so that for typical shallow
// module trees we won't need to reallocate this.
// (moduleInstances does plenty of allocations itself, so the benefit of
// pre-allocating this is marginal but it's not hard to do.)
moduleInstanceAddr := make(addrs.ModuleInstance, 0, 4)
ret := e.exps.moduleResourceInstances(moduleAddr, resourceAddr, moduleInstanceAddr)
sort.SliceStable(ret, func(i, j int) bool {
return ret[i].Less(ret[j])
})
return ret
}
// ExpandResource finds the set of resource instances resulting from
// the expansion of the given resource within its module instance.
//
// All of the modules on the path to the identified resource and the resource
// itself must already have had their expansion registered using one of the
// SetModule*/SetResource* methods before calling, or this method will panic.
instances: Non-existing module instance has no resource instances Previously we were treating it as a programming error to ask for the instances of a resource inside an instance of a module that is declared but whose declaration doesn't include the given instance key. However, that's actually a valid situation which can arise if, for example, the user has changed the repetition/expansion mode for an existing module call and so now all of the resource instances addresses it previously contained are "orphaned". To represent that, we'll instead say that an invalid instance key of a declared module behaves as if it contains no resource instances at all, regardless of the configurations of any resources nested inside. This then gives the result needed to successfully detect all of the former resource instances as "orphaned" and plan to destroy them. However, this then introduces a new case for NodePlannableResourceInstanceOrphan.deleteActionReason to deal with: the resource configuration still exists (because configuration isn't aware of individual module/resource instances) but the module instance does not. This actually allows us to resolve, at least partially, a previous missing piece of explaining to the user why the resource instances are planned for deletion in that case, finally allowing us to be explicit to the user that it's because of the module instance being removed, which internally we call plans.ResourceInstanceDeleteBecauseNoModule. Co-authored-by: Alisdair McDiarmid <alisdair@users.noreply.github.com>
2021-12-08 14:40:32 -06:00
//
// ExpandModuleResource returns all instances of a resource across all
// instances of its containing module, whereas this ExpandResource function
// is more specific and only expands within a single module instance. If
// any of the module instances selected in the module path of the given address
// aren't valid for that module's expansion then ExpandResource returns an
// empty result, reflecting that a non-existing module instance can never
// contain any existing resource instances.
func (e *Expander) ExpandResource(resourceAddr addrs.AbsResource) []addrs.AbsResourceInstance {
e.mu.RLock()
defer e.mu.RUnlock()
moduleInstanceAddr := make(addrs.ModuleInstance, 0, 4)
ret := e.exps.resourceInstances(resourceAddr.Module, resourceAddr.Resource, moduleInstanceAddr)
sort.SliceStable(ret, func(i, j int) bool {
return ret[i].Less(ret[j])
})
return ret
}
// GetModuleInstanceRepetitionData returns an object describing the values
// that should be available for each.key, each.value, and count.index within
// the call block for the given module instance.
func (e *Expander) GetModuleInstanceRepetitionData(addr addrs.ModuleInstance) RepetitionData {
if len(addr) == 0 {
// The root module is always a singleton, so it has no repetition data.
return RepetitionData{}
}
e.mu.RLock()
defer e.mu.RUnlock()
parentMod := e.findModule(addr[:len(addr)-1])
lastStep := addr[len(addr)-1]
exp, ok := parentMod.moduleCalls[addrs.ModuleCall{Name: lastStep.Name}]
if !ok {
panic(fmt.Sprintf("no expansion has been registered for %s", addr))
}
return exp.repetitionData(lastStep.InstanceKey)
}
// GetResourceInstanceRepetitionData returns an object describing the values
// that should be available for each.key, each.value, and count.index within
// the definition block for the given resource instance.
func (e *Expander) GetResourceInstanceRepetitionData(addr addrs.AbsResourceInstance) RepetitionData {
e.mu.RLock()
defer e.mu.RUnlock()
parentMod := e.findModule(addr.Module)
exp, ok := parentMod.resources[addr.Resource.Resource]
if !ok {
panic(fmt.Sprintf("no expansion has been registered for %s", addr.ContainingResource()))
}
return exp.repetitionData(addr.Resource.Key)
}
// AllInstances returns a set of all of the module and resource instances known
// to the expander.
//
// It generally doesn't make sense to call this until everything has already
// been fully expanded by calling the SetModule* and SetResource* functions.
// After that, the returned set is a convenient small API only for querying
// whether particular instance addresses appeared as a result of those
// expansions.
func (e *Expander) AllInstances() Set {
return Set{e}
}
func (e *Expander) findModule(moduleInstAddr addrs.ModuleInstance) *expanderModule {
// We expect that all of the modules on the path to our module instance
// should already have expansions registered.
mod := e.exps
for i, step := range moduleInstAddr {
next, ok := mod.childInstances[step]
if !ok {
// Top-down ordering of registration is part of the contract of
// Expander, so this is always indicative of a bug in the caller.
panic(fmt.Sprintf("no expansion has been registered for ancestor module %s", moduleInstAddr[:i+1]))
}
mod = next
}
return mod
}
func (e *Expander) setModuleExpansion(parentAddr addrs.ModuleInstance, callAddr addrs.ModuleCall, exp expansion) {
e.mu.Lock()
defer e.mu.Unlock()
mod := e.findModule(parentAddr)
if _, exists := mod.moduleCalls[callAddr]; exists {
panic(fmt.Sprintf("expansion already registered for %s", parentAddr.Child(callAddr.Name, addrs.NoKey)))
}
// We'll also pre-register the child instances so that later calls can
// populate them as the caller traverses the configuration tree.
for _, key := range exp.instanceKeys() {
step := addrs.ModuleInstanceStep{Name: callAddr.Name, InstanceKey: key}
mod.childInstances[step] = newExpanderModule()
}
mod.moduleCalls[callAddr] = exp
}
func (e *Expander) setResourceExpansion(parentAddr addrs.ModuleInstance, resourceAddr addrs.Resource, exp expansion) {
e.mu.Lock()
defer e.mu.Unlock()
mod := e.findModule(parentAddr)
if _, exists := mod.resources[resourceAddr]; exists {
panic(fmt.Sprintf("expansion already registered for %s", resourceAddr.Absolute(parentAddr)))
}
mod.resources[resourceAddr] = exp
}
func (e *Expander) knowsModuleInstance(want addrs.ModuleInstance) bool {
if want.IsRoot() {
return true // root module instance is always present
}
e.mu.Lock()
defer e.mu.Unlock()
return e.exps.knowsModuleInstance(want)
}
func (e *Expander) knowsModuleCall(want addrs.AbsModuleCall) bool {
e.mu.Lock()
defer e.mu.Unlock()
return e.exps.knowsModuleCall(want)
}
func (e *Expander) knowsResourceInstance(want addrs.AbsResourceInstance) bool {
e.mu.Lock()
defer e.mu.Unlock()
return e.exps.knowsResourceInstance(want)
}
func (e *Expander) knowsResource(want addrs.AbsResource) bool {
e.mu.Lock()
defer e.mu.Unlock()
return e.exps.knowsResource(want)
}
type expanderModule struct {
moduleCalls map[addrs.ModuleCall]expansion
resources map[addrs.Resource]expansion
childInstances map[addrs.ModuleInstanceStep]*expanderModule
}
func newExpanderModule() *expanderModule {
return &expanderModule{
moduleCalls: make(map[addrs.ModuleCall]expansion),
resources: make(map[addrs.Resource]expansion),
childInstances: make(map[addrs.ModuleInstanceStep]*expanderModule),
}
}
var singletonRootModule = []addrs.ModuleInstance{addrs.RootModuleInstance}
// if moduleInstances is being used to lookup known instances after all
// expansions have been done, set skipUnknown to true which allows addrs which
// may not have been seen to return with no instances rather than panicking.
func (m *expanderModule) moduleInstances(addr addrs.Module, parentAddr addrs.ModuleInstance, skipUnknown bool) []addrs.ModuleInstance {
callName := addr[0]
exp, ok := m.moduleCalls[addrs.ModuleCall{Name: callName}]
if !ok {
if skipUnknown {
return nil
}
// This is a bug in the caller, because it should always register
// expansions for an object and all of its ancestors before requesting
// expansion of it.
panic(fmt.Sprintf("no expansion has been registered for %s", parentAddr.Child(callName, addrs.NoKey)))
}
var ret []addrs.ModuleInstance
// If there's more than one step remaining then we need to traverse deeper.
if len(addr) > 1 {
for step, inst := range m.childInstances {
if step.Name != callName {
continue
}
instAddr := append(parentAddr, step)
ret = append(ret, inst.moduleInstances(addr[1:], instAddr, skipUnknown)...)
}
return ret
}
// Otherwise, we'll use the expansion from the final step to produce
// a sequence of addresses under this prefix.
for _, k := range exp.instanceKeys() {
// We're reusing the buffer under parentAddr as we recurse through
// the structure, so we need to copy it here to produce a final
// immutable slice to return.
full := make(addrs.ModuleInstance, 0, len(parentAddr)+1)
full = append(full, parentAddr...)
full = full.Child(callName, k)
ret = append(ret, full)
}
return ret
}
func (m *expanderModule) moduleResourceInstances(moduleAddr addrs.Module, resourceAddr addrs.Resource, parentAddr addrs.ModuleInstance) []addrs.AbsResourceInstance {
if len(moduleAddr) > 0 {
var ret []addrs.AbsResourceInstance
// We need to traverse through the module levels first, so we can
// then iterate resource expansions in the context of each module
// path leading to them.
callName := moduleAddr[0]
if _, ok := m.moduleCalls[addrs.ModuleCall{Name: callName}]; !ok {
// This is a bug in the caller, because it should always register
// expansions for an object and all of its ancestors before requesting
// expansion of it.
panic(fmt.Sprintf("no expansion has been registered for %s", parentAddr.Child(callName, addrs.NoKey)))
}
for step, inst := range m.childInstances {
if step.Name != callName {
continue
}
moduleInstAddr := append(parentAddr, step)
ret = append(ret, inst.moduleResourceInstances(moduleAddr[1:], resourceAddr, moduleInstAddr)...)
}
return ret
}
return m.onlyResourceInstances(resourceAddr, parentAddr)
}
func (m *expanderModule) resourceInstances(moduleAddr addrs.ModuleInstance, resourceAddr addrs.Resource, parentAddr addrs.ModuleInstance) []addrs.AbsResourceInstance {
if len(moduleAddr) > 0 {
// We need to traverse through the module levels first, using only the
// module instances for our specific resource, as the resource may not
// yet be expanded in all module instances.
step := moduleAddr[0]
callName := step.Name
if _, ok := m.moduleCalls[addrs.ModuleCall{Name: callName}]; !ok {
// This is a bug in the caller, because it should always register
// expansions for an object and all of its ancestors before requesting
// expansion of it.
panic(fmt.Sprintf("no expansion has been registered for %s", parentAddr.Child(callName, addrs.NoKey)))
}
instances: Non-existing module instance has no resource instances Previously we were treating it as a programming error to ask for the instances of a resource inside an instance of a module that is declared but whose declaration doesn't include the given instance key. However, that's actually a valid situation which can arise if, for example, the user has changed the repetition/expansion mode for an existing module call and so now all of the resource instances addresses it previously contained are "orphaned". To represent that, we'll instead say that an invalid instance key of a declared module behaves as if it contains no resource instances at all, regardless of the configurations of any resources nested inside. This then gives the result needed to successfully detect all of the former resource instances as "orphaned" and plan to destroy them. However, this then introduces a new case for NodePlannableResourceInstanceOrphan.deleteActionReason to deal with: the resource configuration still exists (because configuration isn't aware of individual module/resource instances) but the module instance does not. This actually allows us to resolve, at least partially, a previous missing piece of explaining to the user why the resource instances are planned for deletion in that case, finally allowing us to be explicit to the user that it's because of the module instance being removed, which internally we call plans.ResourceInstanceDeleteBecauseNoModule. Co-authored-by: Alisdair McDiarmid <alisdair@users.noreply.github.com>
2021-12-08 14:40:32 -06:00
if inst, ok := m.childInstances[step]; ok {
moduleInstAddr := append(parentAddr, step)
return inst.resourceInstances(moduleAddr[1:], resourceAddr, moduleInstAddr)
} else {
// If we have the module _call_ registered (as we checked above)
// but we don't have the given module _instance_ registered, that
// suggests that the module instance key in "step" is not declared
// by the current definition of this module call. That means the
// module instance doesn't exist at all, and therefore it can't
// possibly declare any resource instances either.
//
// For example, if we were asked about module.foo[0].aws_instance.bar
// but module.foo doesn't currently have count set, then there is no
// module.foo[0] at all, and therefore no aws_instance.bar
// instances inside it.
return nil
}
}
return m.onlyResourceInstances(resourceAddr, parentAddr)
}
func (m *expanderModule) onlyResourceInstances(resourceAddr addrs.Resource, parentAddr addrs.ModuleInstance) []addrs.AbsResourceInstance {
var ret []addrs.AbsResourceInstance
exp, ok := m.resources[resourceAddr]
if !ok {
panic(fmt.Sprintf("no expansion has been registered for %s", resourceAddr.Absolute(parentAddr)))
}
for _, k := range exp.instanceKeys() {
// We're reusing the buffer under parentAddr as we recurse through
// the structure, so we need to copy it here to produce a final
// immutable slice to return.
moduleAddr := make(addrs.ModuleInstance, len(parentAddr))
copy(moduleAddr, parentAddr)
ret = append(ret, resourceAddr.Instance(k).Absolute(moduleAddr))
}
return ret
}
func (m *expanderModule) getModuleInstance(want addrs.ModuleInstance) *expanderModule {
current := m
for _, step := range want {
next := current.childInstances[step]
if next == nil {
return nil
}
current = next
}
return current
}
func (m *expanderModule) knowsModuleInstance(want addrs.ModuleInstance) bool {
return m.getModuleInstance(want) != nil
}
func (m *expanderModule) knowsModuleCall(want addrs.AbsModuleCall) bool {
modInst := m.getModuleInstance(want.Module)
if modInst == nil {
return false
}
_, ret := modInst.moduleCalls[want.Call]
return ret
}
func (m *expanderModule) knowsResourceInstance(want addrs.AbsResourceInstance) bool {
modInst := m.getModuleInstance(want.Module)
if modInst == nil {
return false
}
resourceExp := modInst.resources[want.Resource.Resource]
if resourceExp == nil {
return false
}
for _, key := range resourceExp.instanceKeys() {
if key == want.Resource.Key {
return true
}
}
return false
}
func (m *expanderModule) knowsResource(want addrs.AbsResource) bool {
modInst := m.getModuleInstance(want.Module)
if modInst == nil {
return false
}
_, ret := modInst.resources[want.Resource]
return ret
}