datacatalog: don't drop bool_value when ordered field is present

[jbbqqf]

Summary

Fixes a regression in the custom encoder for google_data_catalog_tag
where bool_value was silently dropped whenever the matching field had
any companion key — including pure metadata like order — instead of
only when another value-typed key (stringValue, doubleValue, …) was
present.

Fixes hashicorp/terraform-provider-google#16856 — see hashicorp/terraform-provider-google#16856

Why

The DataCatalog API rejects a field payload that contains more than one
value-typed entry (one of boolValue, stringValue, doubleValue,
timestampValue, enumValue, richtextValue). To prevent this on the
provider side, the custom encoder added in
GoogleCloudPlatform/magic-modules#4004
strips boolValue whenever a field has multiple entries, on the
assumption that a multi-entry field implies a value conflict.

That assumption is wrong. A user who writes:

fields {
  field_id     = "bool_field"
  display_name = "..."
  type {
    primitive_type = "BOOL"
  }
  order = 1   # any non-zero value
}

…and then a tag with:

fields {
  field_name = "bool_field"
  bool_value = false
}

…produces a field map with {boolValue: false, order: 1} after the
expand pass. The old encoder saw len(values) > 1 and stripped
boolValue — even though order is a metadata key, not a value-typed
key. The resulting payload had no value at all, breaking updates with
either an API error or a silent state mismatch.

Maintainer @roaks3 confirmed in the issue thread:

It looks like some of this behavior was explicitly changed with PR
GoogleCloudPlatform#4004, and we most likely missed an edge case.

GCP API reference:

• TagField schema: https://cloud.google.com/data-catalog/docs/reference/rest/v1/projects.locations.entryGroups.entries.tags#tagfield
• Original PR that introduced the bug: fixed error when a boolean field set to false GoogleCloudPlatform/magic-modules#4004

What changed

mmv1 custom encoder template only — replaces the unconditional
"more-than-one-key" strip with a check that at least one other
value-typed key is present. Pure metadata companions (order,
displayName, fieldName, column) no longer trigger the strip.

 mmv1/templates/terraform/encoders/data_catalog_tag.go.tmpl | 41 ++++++++++++++++---
 1 file changed, 33 insertions(+), 8 deletions(-)

The original "value-conflict" defense is preserved (e.g. if a user really
did set both bool_value and string_value on the same field, we still
strip boolValue — same as today).

Edge cases tested

#	Scenario	HCL excerpt	Expected	Verified by
1	bool_value alone, ordered field	bool_value = false + tag template order = 1	Payload contains boolValue: false, update succeeds	static — encoder retains the key because no other value-typed key is present
2	bool_value + string_value (legitimate conflict)	both set on same field	boolValue stripped (preserves PR GoogleCloudPlatform#4004 behavior); only stringValue reaches API	static — encoder still strips when another value-typed key present
3	string_value only, ordered field	string_value = "x" + order = 1	stringValue reaches API; boolValue was never present so encoder is no-op	static — encoder skips early when no boolValue

Test protocol

Test	Result	Notes
Encoder logic review	OK	Iterates fields once; only strips boolValue when another value-typed key shares the same field
Backwards compatibility	OK	Behavior unchanged for the conflict case PR GoogleCloudPlatform#4004 was designed to handle
Live BEFORE/AFTER smoke	not run	DataCatalog tag template provisioning + tag update needs a working entry group; live smoke deferred for this batch. The fix is logic-tightening on a clearly-identified bug; reviewer can validate with the issue's exact reproducer.

Resources

• Original issue: DataCatalog tag update fails when ordered boolean field is present. hashicorp/terraform-provider-google#16856
• TagField API: https://cloud.google.com/data-catalog/docs/reference/rest/v1/projects.locations.entryGroups.entries.tags#tagfield
• PR that introduced the regression: fixed error when a boolean field set to false GoogleCloudPlatform/magic-modules#4004

Disclosure

This PR was drafted with assistance from Claude Code as part of a parallel
contribution batch. The encoder change was reviewed against the
DataCatalog API documentation and against the original maintainer
acknowledgement in the issue thread. Live smoke was not run; the author
(a human) will review the diff and the modular-magician downstream PRs
before requesting maintainer review.