Can you use one attribute that is an alias to an existing pre-defined field for completeness? I feel like you're not understanding what I'm asking.

apologies, there was a reference to use a "predefined" field (not TemporalInt01 etc) in test_component_test.go. I also added a reference to the predefined field SearchAttributeTemporalScheduledByID.

As for whether to include it in Payload.go, there was a previous TODO to track Count and Size as search attributes, and I wasn't sure that adding a reference to TemporalScheduledById made sense within the scope of the Payload component. This component only tracks the current count and size of payloads, and a map from the PayloadKey to its Payload blob. What would be the values provided to the search attribute?

I agree that functional tests should cover this case, I just wasn't sure if it was acceptable style to add a search attribute that doesn't seem related to the component. Added a reference in payload.go to this.

I suppose you can't just skip the CHASM search attributes based on this comment.

This is temporary, we won't skip CHASM search attributes once we consolidate the type maps on the query path.

Ditto, I think it's better that you check if it's a CHASM search attribute before checking the type. Also, in this code, if it's a CHASM search attribute, saType will be UNSPECIFIED, so you'd be setting the wrong type for CHASM search attributes.

If UNSPECIFIED, SetMetadataType will return early. To be clear, this is just temp code that will need to change when the query path is implemented.

func SetMetadataType(p *commonpb.Payload, t enumspb.IndexedValueType) {
	if t == enumspb.INDEXED_VALUE_TYPE_UNSPECIFIED {
		return
	}

	_, isValidT := enumspb.IndexedValueType_name[int32(t)]
	if !isValidT {
		panic(fmt.Sprintf("unknown index value type %v", t))
	}
	p.Metadata[MetadataType] = []byte(enumspb.IndexedValueType(t).String())
}

This is in the write path, not query path. What I mean is this code will always set the type for CHASM search attribute as UNSPECIFIED, and as you pointed out, the function will just ignore, ie., the CHASM search attributes will never have its type set in the metadata.
Or did I miss something?

Encode is on the read path. I suppose either way, the MetadataType is somewhat ignored on the write path, and we basically reconstruct the MetadataType solely from the saTypeMap right now? Either way, without merging the saTypeMap with the CHASM component's saTypeMap, current functionality won't work. On the next PR to support querying, we can re-evaluate how we want to support setting the MetadataType. This portion of code is mainly to get functional tests to succeed.

ditto

this isn't necessary, DecodeValue handles this case

func DecodeValue(
	value *commonpb.Payload,
	t enumspb.IndexedValueType,
	allowList bool,
) (any, error) {
	if t == enumspb.INDEXED_VALUE_TYPE_UNSPECIFIED {
		var err error
		t, err = enumspb.IndexedValueTypeFromString(string(value.Metadata[MetadataType]))
		if err != nil {
			return nil, fmt.Errorf("%w: %v", ErrInvalidType, t)
		}
	}

	switch t {
	case enumspb.INDEXED_VALUE_TYPE_BOOL:
		return decodeValueTyped[bool](value, allowList)
	case enumspb.INDEXED_VALUE_TYPE_DATETIME:
		return decodeValueTyped[time.Time](value, allowList)
	case enumspb.INDEXED_VALUE_TYPE_DOUBLE:
		return decodeValueTyped[float64](value, allowList)
	case enumspb.INDEXED_VALUE_TYPE_INT:
		return decodeValueTyped[int64](value, allowList)
	case enumspb.INDEXED_VALUE_TYPE_KEYWORD:
		return validateStrings(decodeValueTyped[string](value, allowList))
	case enumspb.INDEXED_VALUE_TYPE_TEXT:
		return validateStrings(decodeValueTyped[string](value, allowList))
	case enumspb.INDEXED_VALUE_TYPE_KEYWORD_LIST:
		return validateStrings(decodeValueTyped[[]string](value, false))
	default:
		return nil, fmt.Errorf("%w: %v", ErrInvalidType, t)
	}
}

We need to revert this change. This is swallowing errors when the custom search attribute is not defined in the typeMap.

The correct behavior is: if the typeMap input is not nil, Decode checks everything against typeMap; otherwise, it uses the payload metadata if available.

just fmi, in which cases would the typeMap be nil, and we would then be permitted to decode the value based off the payload Metadata? I would have assumed we would've thrown an error when translating from alias to field upstream, but it seems that isn't the case.

Please add unit tests for changes you did.

there are unit test validations for the changes. Previously, key2 would have failed decoding due it missing in the saTypeMap