Fix manasight/manasight-parser#221: emit ChoiceResult / LinkInfo annotation details

[timc-enthrall]

Summary

• Parser now extracts AnnotationType_ChoiceResult and AnnotationType_LinkInfo detail fields instead of dropping them at the no-op catch-all, so "name a card" choices and their link metadata reach the desktop.

Changes Made

• Added ANNOTATION_TYPE_CHOICE_RESULT / ANNOTATION_TYPE_LINK_INFO consts and match arms in add_type_specific_fields.
• Added add_choice_result_fields / add_link_info_fields extractors (permissive, mirroring add_designation_fields — generic per-domain emit, no special-casing CardName).
• Emitted fields — ChoiceResult: choice_value (always), choice_domain, choice_sentiment; LinkInfo: choose_link_type, source_ability_grp_id, link_type.
• Updated the extract_annotations doc-comment to document both new types.

Testing

• New mod choice_result_extraction (3 tests: Domain=5, value-only, Domain=13 CardName live capture shape) and mod link_info_extraction (2 tests: ChooseLinkType="Type" with sourceAbilityGRPID + LinkType; ChooseLinkType="CardName" minimal case).
• All tests passing; lint clean; formatted.
• Coverage estimate: Δ ≈ +5 tests, ~30 new tested LOC vs ~30 new production LOC — net neutral to slight positive on coverage ratio.

Stacked PR

Base: main — single PR, no stack.

Closes #221

🤖 Generated with Claude Code

[timc-enthrall]

Code review

Found 1 issue:

1. Missing edge-case test coverage for the two new permissive extractors. add_choice_result_fields falls back to choice_value: 0 via unwrap_or(0) when Choice_Value is absent, but no test exercises that absent-value fallback path. Likewise add_link_info_fields can emit zero type-specific fields when all three detail keys are absent, and that all-keys-absent case is untested. The repo CLAUDE.md testing policy requires "New/updated tests for every code change" and "Test behavior, not implementation" — the permissive fallback behavior is new and should be locked in. (CLAUDE.md says "New/updated tests for every code change")

manasight-parser/src/parsers/gre/annotations.rs

Lines 492 to 531 in 753e08b

	/// variants pass through cleanly with whatever fields they happen to carry.
	fn add_choice_result_fields(
	obj: &mut serde_json::Map<String, serde_json::Value>,
	details: &[serde_json::Value],
	) {
	obj.insert(
	"choice_value".into(),
	serde_json::json!(detail_int(details, "Choice_Value").unwrap_or(0)),
	);
	if let Some(d) = detail_int(details, "Choice_Domain") {
	obj.insert("choice_domain".into(), serde_json::json!(d));
	}
	if let Some(s) = detail_int(details, "Choice_Sentiment") {
	obj.insert("choice_sentiment".into(), serde_json::json!(s));
	}
	}
	/// Adds `LinkInfo` fields to an annotation result.
	///
	/// Conditionally emits `choose_link_type`, `source_ability_grp_id`, and
	/// `link_type` when found in `details`.
	///
	/// The implementation is intentionally permissive: every key is looked up
	/// unconditionally via the existing `detail_str` / `detail_int` helpers and
	/// inserted only when found. Never branches on the `ChooseLinkType` value —
	/// undocumented variants pass through cleanly with whatever fields they carry.
	fn add_link_info_fields(
	obj: &mut serde_json::Map<String, serde_json::Value>,
	details: &[serde_json::Value],
	) {
	if let Some(t) = detail_str(details, "ChooseLinkType") {
	obj.insert("choose_link_type".into(), serde_json::json!(t));
	}
	if let Some(g) = detail_int(details, "sourceAbilityGRPID") {
	obj.insert("source_ability_grp_id".into(), serde_json::json!(g));
	}
	if let Some(lt) = detail_int(details, "LinkType") {
	obj.insert("link_type".into(), serde_json::json!(lt));
	}
	}

The new test modules cover present-key cases but not the absent-key fallbacks:

manasight-parser/src/parsers/gre/annotations.rs

Lines 2018 to 2020 in 753e08b

	mod choice_result_extraction {
	use super::*;

[timc-enthrall]

Code review

Found 1 issue:

1. Missing positive-path test for the new choice_sentiment conditional emit. add_choice_result_fields conditionally emits choice_sentiment when Choice_Sentiment is present in details, but every ChoiceResult test only asserts choice_sentiment is absent — no test exercises the present-and-emitted path. The sibling field choice_domain (same if let Some(...) pattern) has a positive test (test_choice_result_domain5_value_and_domain_emitted), so this is a coverage gap relative to the file's own convention. The repo CLAUDE.md testing policy requires "New/updated tests for every code change" and "Test behavior, not implementation" — the present-path emit of choice_sentiment is new behavior that should be locked in. (CLAUDE.md says "New/updated tests for every code change")

Emit site (no positive-path test covers this branch):

manasight-parser/src/parsers/gre/annotations.rs

Lines 503 to 506 in 7990adc

	}
	if let Some(s) = detail_int(details, "Choice_Sentiment") {
	obj.insert("choice_sentiment".into(), serde_json::json!(s));
	}

Tests only ever assert absence:

manasight-parser/src/parsers/gre/annotations.rs

Lines 2062 to 2065 in 7990adc

	assert_eq!(ann["choice_domain"], 5);
	// choice_sentiment absent when not in details.
	assert!(ann.get("choice_sentiment").is_none());
	}

[timc-enthrall]

Code review

No blocking issues found. Checked for bugs and CLAUDE.md/CONVENTIONS compliance across five independent passes (CLAUDE.md adherence, shallow bug scan, git-history context, prior-PR comments, code-comment compliance).

Both findings from the earlier review rounds are resolved:

• absent-Choice_Value and LinkInfo all-keys-absent edge cases (iteration 2)
• positive-path choice_sentiment emit test (iteration 3): test_choice_result_sentiment_emitted_when_present

One sub-threshold observation was scored below the confidence bar and filtered: the add_choice_result_fields doc-comment rationale doesn't spell out the unwrap_or(0) fallback. The behavior itself is correct, intentional house style (mirrors add_zone_transfer_fields / add_designation_fields), and is unit-tested (test_choice_result_absent_choice_value_defaults_to_zero).