fix(parser): scale P/T by opponents below half starting life (Anya, Merciless Angel #5920)

[xfodev]

Fixes #5920.

Summary

Anya, Merciless Angel ({3}{R}{W}, 4/4): "Anya gets +3/+3 for each opponent whose life total is less than half their starting life total." She was being buffed unconditionally — the reporter saw +3/+3 while every opponent was well above half their starting life.

Root cause: the static "gets +N/+N for each opponent whose …" dynamic-count dispatch (parse_for_each_opponent_player_attribute_clause) had no life-total arm. The for-each clause failed to parse, so control fell through to parse_continuous_modifications, which extracted a fixed +3/+3 and silently dropped the count scaling.

Fix (parser-only)

Adds parse_life_total_who_attr_clause — the direct sibling of the existing parse_hand_size_who_attr_clause — composing two shared building blocks already used by the existential "an opponent's life total is <cmp> <qty>" static condition (i.e. Anya's own second/indestructible ability):

• parse_life_total_comparator (the shared life-total ordering-comparator grammar)
• nom_quantity::parse_quantity (the shared threshold-quantity grammar; "half their starting life total" → DivideRounded(StartingLifeTotal, 2, Down))

The clause now lowers to:

PlayerCount { PlayerFilter::PlayerAttribute {
  relation: Opponent, attr: LifeTotal(ScopedPlayer),
  comparator: LT, value: DivideRounded(StartingLifeTotal, 2, Down) } }

No new engine variant, no runtime/walker changes. PlayerFilter::PlayerAttribute already exists and its runtime is exercised by shipping cards (Bandit's Talent, Wolfcaller's Howl, Glissa's Retriever). The add-engine-variant gate was run and returned "use the existing parameterized slot" — a scope-baked OpponentLifeBelowHalfStarting sibling would violate "Parameterize, don't proliferate."

Class, not card

Covers "for each opponent whose life total is <comparator> <quantity>" for the full comparator × threshold matrix (via parse_life_total_comparator + parse_quantity), not just Anya.

Anchored on

• crates/engine/src/parser/oracle_quantity.rs:1448 — parse_hand_size_who_attr_clause, the sibling tag ∘ comparator ∘ quantity arm this mirrors (same combinator family, same return tuple (QuantityRef, Comparator, QuantityExpr)).
• crates/engine/src/parser/oracle_quantity.rs:1409 — parse_for_each_opponent_player_attribute_clause, the existing alt(...) this adds one branch to.

Verification

Local cargo test on the full engine test binaries OOMs on this host (7.7 GB; no C toolchain / no root — I link Rust via a bootstrapped zig cc). So I verified by building the engine rlib (non-test builds fine) and running standalone binaries against it:

Parser AST — parse_static_line("~ gets +3/+3 for each opponent whose life total is less than half their starting life total.") produces AddDynamicPower/AddDynamicToughness { Multiply{ 3, PlayerCount{ PlayerAttribute{ Opponent, LifeTotal(ScopedPlayer), LT, DivideRounded(StartingLifeTotal,2,Down) } } } } (both P and T, affected SelfRef). The sibling "who has one or fewer cards in hand" clause still parses (no arm shadowing).

Runtime P/T — Anya (base 4/4) on the battlefield, evaluate_layers, starting_life 20 (half 10):

opponents (life)	qualifying	Anya P/T
both at 20 (≥ half)	0	4/4 ← the reporter's bug case
one at 9 (< half)	1	7/7
two below half	2	10/10
one at exactly 10 (= half)	0 (strict LT)	4/4

Tests added

• crates/engine/tests/integration/anya_merciless_angel_5920.rs — the runtime P/T matrix above through the real Oracle parse → static synthesis → layer pipeline (registered in tests/integration/main.rs).
• oracle_static/tests.rs::static_for_each_opponent_below_half_starting_life_is_dynamic_anya — asserts the dynamic AST (revert-failing: pre-fix emits a fixed pump).

Gate A

Gate A PASS head=2778441f7a61091d4553f04a06b955b9fbafca7c base=45c6b30b85b046ccf5d28741807c95e91356de01

Model: claude-opus-4-8
Thinking: High
Tier: Frontier

[gemini-code-assist]

Code Review

This pull request implements parsing for player attribute clauses checking life totals (e.g., "whose life total is ") in oracle_quantity.rs, resolving issue #5920. This allows cards like Anya, Merciless Angel to correctly scale their power and toughness based on opponents' life totals. The implementation includes comprehensive integration tests verifying the behavior under various life total thresholds. The changes are highly idiomatic, use nom combinators correctly, and include proper CR annotations. I have no feedback or issues to raise.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[github-actions]

Parse changes introduced by this PR · 1 card(s), 1 signature(s) (baseline: main 45c6b30b85b0)

🟡 Modified fields (1 signature)

• 1 card · 🔄 static/Continuous · changed field mods: power +3, toughness +3 → add dynamic power, add dynamic toughness
  • Affected (first 3): Anya, Merciless Angel

[matthewevans]

Approved after current-head review: the parse diff is the claimed Anya dynamic P/T correction, and the runtime test exercises zero, one, and two qualifying opponents plus the strict-half boundary.