Query validation materializes results multiple times

[Widthdom]

Summary

DbSymbolReader.SearchSymbols() / CountSearchSymbols() (src/CodeIndex/Database/DbSymbolReader.cs:122-141, 378) normalize the queries input array by chaining .Select().Where().Distinct().ToList(). At lines 124 and 139 the .ToList() forces a materialization, but when validQueries.Count > 1 line 141 immediately recurses into SearchSymbols(validQueries, ...), which re-runs the same normalization on already-normalized input. Repeated normalization adds GC pressure for queries with many distinct terms.

Where

• src/CodeIndex/Database/DbSymbolReader.cs:122-141 (AnySearchSymbols / CountSearchSymbols)
• src/CodeIndex/Database/DbSymbolReader.cs:378 (ExcludeUsedSymbols — same pattern)
• src/CodeIndex/Database/DbSymbolReader.cs:235, 330 (AnalyzeSymbol — .Where().Distinct().ToList())

Suggested approach

1. Extract query normalization into one method returning IReadOnlyList<string>; cache the result on first call per input.
2. Change recursion call sites to pass the already-normalized list and have the callee detect it.
3. Use HashSet<string> where order is unimportant; reserve Distinct().ToList() for cases that need ordered output.
4. Make the public method accept IReadOnlyList<string> and document that callers should pre-normalize when possible.
5. Add a unit test that asserts normalization runs once per distinct input set.
6. Run a benchmark with a 50-term query and verify allocations drop materially.

[Widthdom]

Work started on branch codex/fix-issue1827-1954.