Search the content index. Returns matching lines with file paths, line numbers, and content. Supports path filtering, file type filtering, and case-insensitive search. Results are ranked: source files (`src/`, `lib/`, `pkg/`, `internal/`) are prioritized over config, test, build, and generated files. Within each tier, results follow path order and preserve the verifier's within-file order. The
(
repo_root: &Path,
pattern: &str,
opts: ContentSearchOptions,
)
| 185 | /// (capped to `max_results`) and the `total_matches` count so callers |
| 186 | /// know how many were found before truncation. |
| 187 | pub fn search_content( |
| 188 | repo_root: &Path, |
| 189 | pattern: &str, |
| 190 | opts: ContentSearchOptions, |
| 191 | ) -> Result<ContentSearchResult, ContentSearchError> { |
| 192 | let config = content_config(repo_root); |
| 193 | let index = Index::open(config)?; |
| 194 | |
| 195 | let display_limit = opts.max_results.unwrap_or(50); |
| 196 | |
| 197 | // Fetch a large candidate set so we can rank before truncating. |
| 198 | // We ask for 10x the display limit (capped at 2000) to get good |
| 199 | // coverage across the repo before applying our own ranking. |
| 200 | let fetch_limit = (display_limit * 10).min(2000); |
| 201 | |
| 202 | let search_opts = SearchOptions { |
| 203 | path_filter: opts.path_filter, |
| 204 | file_type: opts.file_type, |
| 205 | exclude_type: opts.exclude_type, |
| 206 | max_results: Some(fetch_limit), |
| 207 | case_insensitive: opts.case_insensitive, |
| 208 | }; |
| 209 | |
| 210 | let raw_matches = index.search(pattern, &search_opts)?; |
| 211 | |
| 212 | // Drop matches in ignored or Atomic-internal paths. syntext's index walk |
| 213 | // respects `.gitignore`/`.ignore` but NOT `.atomicignore`, and it descends |
| 214 | // into hidden dirs like `.atomic`/`.vault`, so build artifacts and |
| 215 | // dependencies excluded only by `.atomicignore` (plus Atomic internals) |
| 216 | // would otherwise surface here and, via the KG search's content-only file |
| 217 | // promotion, as `file:` results. Filter at this shared consumption point |
| 218 | // so both `query search` and `query code` stay clean regardless of what |
| 219 | // the index contains. |
| 220 | let ignore_rules = crate::ignore::IgnoreRules::load_for_enrichment(repo_root); |
| 221 | let raw_matches: Vec<syntext::SearchMatch> = raw_matches |
| 222 | .into_iter() |
| 223 | .filter(|m| { |
| 224 | !crate::ignore::is_enrichment_internal(&m.path) |
| 225 | && !ignore_rules.is_ignored(&m.path, false) |
| 226 | }) |
| 227 | .collect(); |
| 228 | Ok(rank_content_matches(raw_matches, display_limit)) |
| 229 | } |
| 230 | |
| 231 | /// Recorded-query byte and regex-automaton cutoff (10 MiB). |
| 232 | pub(crate) const RECORDED_CONTENT_LIMIT: usize = 10 * 1024 * 1024; |