feat: Enhance symbol search with co-location boosting and receiver type support

Improves search accuracy by boosting results when multiple query symbols appear in the same file, addressing cases where common names like "run" return too many results. Adds Go method receiver type extraction to qualified names for better searchability (e.g., "scrapeLoop.run"). Optimizes database queries with two-pass approach to handle distinctive vs common symbol names efficiently.
This commit is contained in:
Colby McHenry
2026-04-06 11:16:56 -05:00
parent d256af3a23
commit 7a3afc9124
16 changed files with 466 additions and 877 deletions
+23 -1
View File
@@ -25,17 +25,39 @@ export const STOP_WORDS = new Set([
'over', 'only', 'new', 'out', 'its', 'so', 'up', 'as', 'if',
// Code-specific noise
'code', 'file', 'files', 'function', 'method', 'class', 'type',
'build', 'run', 'test', 'fix', 'bug', 'call', 'called', 'set', 'add',
'build', 'fix', 'bug', 'called', 'set', 'add',
]);
/**
* Extract meaningful search terms from a natural language query.
* Splits camelCase, PascalCase, snake_case, SCREAMING_SNAKE, and dot.notation
* into individual tokens before filtering.
*
* Preserves original compound identifiers (e.g., "scrapeLoop") alongside
* their split parts so that FTS can match both the full symbol name and
* individual words within it.
*/
export function extractSearchTerms(query: string): string[] {
const tokens = new Set<string>();
// First, extract and preserve compound identifiers before splitting
// CamelCase: scrapeLoop, UserService, getCallGraph
const compoundPattern = /\b([a-zA-Z][a-zA-Z0-9]*(?:[A-Z][a-z]+)+|[A-Z][a-z]+(?:[A-Z][a-z]*)+)\b/g;
let match;
while ((match = compoundPattern.exec(query)) !== null) {
if (match[1] && match[1].length >= 3) {
tokens.add(match[1].toLowerCase()); // preserve full compound: "scrapeloop"
}
}
// snake_case: scrape_loop, user_service
const snakePattern = /\b([a-zA-Z][a-zA-Z0-9]*(?:_[a-zA-Z0-9]+)+)\b/g;
while ((match = snakePattern.exec(query)) !== null) {
if (match[1] && match[1].length >= 3) {
tokens.add(match[1].toLowerCase());
}
}
// Split camelCase / PascalCase: "getUserName" → "get User Name"
const camelSplit = query
.replace(/([a-z])([A-Z])/g, '$1 $2')