fix(parsers): parse C# fields and locals - #1584

Fixes #1525. Full suite: 1136 passed, 7 skipped, no failures.

The bug

variables was hardcoded to [] and CSHARP_QUERIES had no variables entry at all — C# produced zero Variable nodes.

The larger cost was downstream. resolution/calls.py uses local-variable declarations to infer a receiver's type, so:

var svc = new UserService();
svc.Save();

could not resolve Save to UserService.Save, and C# method calls fell back to name-only heuristics.

The fix

Fields and locals share a shape — variable_declaration carries the type and holds one variable_declarator per name — so one query covers both:

  L2 _count   type=int      value=1                   ctx=None  cls=UserService
  L3 Name     type=string   value=None                ctx=None  cls=UserService
  L4 A        type=int      value=1                   ctx=None  cls=UserService
  L4 B        type=int      value=2                   ctx=None  cls=UserService
  L6 total    type=int      value=0                   ctx=Save  cls=UserService
  L7 svc      type=var      value=new UserService()   ctx=Save  cls=UserService

int A = 1, B = 2; correctly yields two rows sharing a type, and locals carry their enclosing method as context.

One grammar detail worth recording: tree-sitter-c-sharp has no equals_value_clause — the initialiser is a direct sibling after the = token. My first attempt looked for that node and returned value: None for everything. Capturing it is what makes var svc = new UserService() usable for receiver-type inference.

Golden — please read this bit

The raw diff looks larger than the change. The committed C# golden was already stale, which I only found by controlling for it: regenerating on an unmodified main alone produces

Directory   6 ->  9      CALLS     12 ->  7
File       15 -> 20      IMPORTS   21 -> 32
Module      9 -> 12      CONTAINS 132 -> 140

That drift is unrelated to this PR and is carried along because the file had to be regenerated. Measured against a clean-main regeneration, this change contributes exactly:

nodes  Variable    0 -> 15  (+15)
edges  CONTAINS  140 -> 155 (+15)
(nothing else changed)

Purely additive, and in particular the CALLS 12 -> 7 drop is not from this change — I checked, because a CALLS regression is what made me abandon the JS/TS import fix in #1526.

I also confirmed the regeneration is deterministic (two identical runs, identical counts), so none of the above is golden noise.

The stale baseline is worth its own issue — a golden that no longer matches what the code produces means the C# parity signal has been silently degraded for a while.

Tests

tests/unit/parsers/test_csharp_variables.py — 8 tests covering fields, locals, multi-declarator, method/class context, initialisers, and the var x = new T() inference case specifically.

🤖 Generated with Claude Code