fix(extract): Class.lines always 0 for all languages

[isc-tdyar]

Class.lines always 0 for all languages

Discovered while trying to sort classes by size. Every class node had lines = 0 regardless of actual length.

Root cause

In extract_class_def(), start_line and end_line are correctly set from tree-sitter node positions, but def.lines is never computed — it stays 0 from the memset that zeroes the struct. The equivalent computation exists and is correct in push_method_def():

// push_method_def — correct
def.lines = (int)(def.end_line - def.start_line + TS_LINE_OFFSET);

// extract_class_def — missing this line entirely
// def.lines was always 0

Fix

Add the same one-liner to extract_class_def() after end_line is set. Affects all languages — any tree-sitter grammar that produces class nodes goes through this path.

Regression test

tool_qg_class_lines_nonzero in tests/test_incremental.c — indexes a project in full mode and asserts that at least one class node has lines > 0 via a query_graph tool call. Before the fix this assertion always failed.

[DeusData]

Thanks for the contribution — CI is fully green here too. Class.lines always being 0 is exactly the kind of quiet quality bug worth fixing; will review when time allows.

[DeusData]

Thanks @isc-tdyar — this one reviewed clean and is ready to merge, but it now conflicts with main (it shares tests/test_incremental.c with a just-merged PR). Could you rebase on latest main? I'll merge as soon as it's conflict-free. 🙏

[DeusData]

Thanks @isc-tdyar — the one-line Class.lines fix is correct, but the diff is now stale against main (conflicting; the extraction files moved under recent changes). Could you rebase on latest main? Worth double-checking while you're there that the class-lines path isn't already covered upstream. I'll merge once it's rebased. 🙏

[DeusData]

Thanks for catching this, @isc-tdyar — a clean root-cause fix. extract_class_def was the one definition-emit site that never set def.lines, so every Class node reported body_lines = 0 across all languages while the seven sibling sites computed it. Your one-line change lines up exactly with those siblings, and the MATCH (c:Class) WHERE c.lines > 0 regression test locks it in. Appreciate the careful, minimal fix — merging now. 🙏