Define Set type

[kmoe]

Backed by an actual Set implementation.

[paddycarver]

I originally thought we wanted to use a third-party Go implementation for this, but the more I think about it, the more convinced I am that we actually want to build our own, in-house set implementation using attr.Value as the elements of the set.

[bflad]

This might have been developer error on my part, but in trying to implement an "ordered set" type (list with required ordering, but element-uniqueness constraints), I was having trouble with creating a map[tftypes.Value]struct{} due to tftypes.primitive not being hashable (resultant in a Go panic). If that does turn out to be a problem, we might need to figure out the best way to hash values.

Panic:

panic: runtime error: hash of unhashable type tftypes.primitive [recovered]
	panic: runtime error: hash of unhashable type tftypes.primitive

goroutine 14 [running]:
testing.tRunner.func1.2(0x1631d00, 0xc000412090)
	/usr/local/Cellar/go/1.16.6/libexec/src/testing/testing.go:1143 +0x332
testing.tRunner.func1(0xc0003bd800)
	/usr/local/Cellar/go/1.16.6/libexec/src/testing/testing.go:1146 +0x4b6
panic(0x1631d00, 0xc000412090)
	/usr/local/Cellar/go/1.16.6/libexec/src/runtime/panic.go:965 +0x1b9
type..hash.git.luolix.top/hashicorp/terraform-plugin-go/tftypes.Value(0xc0000df168, 0x0, 0x18)
	<autogenerated>:1 +0x35
github.com/hashicorp/terraform-plugin-framework/types.OrderedSetType.Validate(0x0, 0x0, 0x17aee90, 0xc00011a008, 0x17b2428, 0xc0003ab230, 0x15f87a0, 0xc00000e978, 0x169d940, 0x1, ...)
	/Users/bflad/src/github.com/hashicorp/terraform-plugin-framework/types/orderedset.go:47 +0x2a5

Code:

package types

import (
	"context"
	"fmt"
	"strings"

	"github.com/hashicorp/terraform-plugin-framework/attr"
	"github.com/hashicorp/terraform-plugin-go/tfprotov6"
	"github.com/hashicorp/terraform-plugin-go/tftypes"
)

var (
	_ attr.TypeWithValidate = OrderedSetType{}
)

type OrderedSetType struct {
	ListType
}

func (t OrderedSetType) Validate(ctx context.Context, in tftypes.Value) []*tfprotov6.Diagnostic {
	var diags []*tfprotov6.Diagnostic

	if !in.Type().Is(tftypes.List{}) {
		err := fmt.Errorf("expected List value, received %T with value: %v", in, in)
		return append(diags, &tfprotov6.Diagnostic{
			Severity: tfprotov6.DiagnosticSeverityError,
			Summary:  "OrderedSet Type Validation Error",
			Detail:   "An unexpected error was encountered trying to validate an attribute value. This is always an error in the provider. Please report the following to the provider developer:\n\n" + err.Error(),
		})
	}

	var vals []tftypes.Value

	if err := in.As(&vals); err != nil {
		return append(diags, &tfprotov6.Diagnostic{
			Severity: tfprotov6.DiagnosticSeverityError,
			Summary:  "OrderedSet Type Validation Error",
			Detail:   "An unexpected error was encountered trying to validate an attribute value. This is always an error in the provider. Please report the following to the provider developer:\n\n" + err.Error(),
		})
	}

	duplicatesMap := make(map[tftypes.Value]struct{})
	valsMap := make(map[tftypes.Value]struct{})

	for _, val := range vals {
		if _, ok := valsMap[val]; ok {
			if _, ok := duplicatesMap[val]; ok {
				continue
			}

			duplicatesMap[val] = struct{}{}
			continue
		}
		valsMap[val] = struct{}{}
	}

	if len(duplicatesMap) > 0 {
		var duplicates []string

		for duplicate := range duplicatesMap {
			duplicates = append(duplicates, duplicate.String())
		}

		return append(diags, &tfprotov6.Diagnostic{
			Severity: tfprotov6.DiagnosticSeverityError,
			Summary:  "Duplicate Set Elements",
			Detail:   fmt.Sprintf("This attribute contains duplicate elements of:\n\n%s", strings.Join(duplicates, "\n")),
		})
	}

	return nil
}

[paddycarver]

I think my hot take is that we should start with unordered sets, as that's what the Terraform type expects, and right now there's no way to surface a tftypes.Set that is expected to be unordered.

The string representation of the value should be consistent, but I think what we really want is to use the Equal method instead of relying on Go's type system's concept of equality, to avoid any gotchas in the future around that.

[bflad]

Certainly, opened #99 to track the potential for an "ordered set" type, unordered sets (surfacing tftypes.Set) is a much more important goal. I only documented the above since I think it means we'll need to create our own hashing (or introduce more complex algorithms to work with the data like a set) unless we adjust upstream.

[github-actions]

I'm going to lock this issue because it has been closed for 30 days ⏳. This helps our maintainers find and focus on the active issues.
If you have found a problem that seems similar to this, please open a new issue and complete the issue template so we can capture all the details necessary to investigate further.