Java and C# overload identities¶

Stable function IDs¶

Function IDs use <module>#<qualified-name>~<signature-digest> for all Java/C# functions, so adding or removing an overload does not rename the others. identity_signature records the normalized declaration syntax: parameter types, generic arity and relevant constructor/interface distinctions. The digest excludes line numbers, formatting, comments, local parameter names, defaults and bodies. The normal signature property retains the full declaration header, including multiline parameter lists. Type aliases, generic substitution and compiler type equivalence are not evaluated. Other languages retain their existing IDs.

Review records from an older index¶

On the first refreshed scan, old managed Java/C# Function records are retained with _source_state="obsolete", including records without graph links: they may contain authored properties or history. Generated containment is rebuilt for the new identities. Authored/unknown relationships remain attached to the old record; grag cannot infer which overload the author intended after an earlier ID collision. The ingest response explains this transition. Current search excludes the obsolete records; Cypher and evidence="all" can inspect them.

To review replacements for an old ID, query current candidates, compare their signatures and citations, then explicitly reconcile any authored references:

MATCH (f:Function)
WHERE f.id STARTS WITH '<old-function-id>~' AND f._source_state = 'current'
RETURN f.id, f.signature, f.path, f.line_start, f.identity_signature

Legacy records are intentionally retained until reviewed; ingestion does not silently delete or retarget them. New overloads use ordinary source-deletion handling, preserving obsolete nodes when authored references still need them. This improves structural identities, not Java/C# call resolution.