AI search engine and MCP server for bots and agents. One Rust binary indexes code, docs, logs, JSONL/CSV and PDFs for BM25 full-text, vector and hybrid search, code search without grep, agent memory and RAG. Elasticsearch-compatible API; Corpus Hub serves hundreds of reference memory corpora.
Rust agent-memoryai-agentsbm25claude-codeclicode-searchcursordatabaseelasticsearchfull-text-searchhybrid-searchjsonllocal-firstmcpmcp-servermodel-context-protocolragrustsearch-enginevector-search
docs(api): refresh scroll _source-suppression comments after #659
xerj-org08c69bba4ff3
diff --git a/engine/crates/xerj-api/src/es_compat.rs b/engine/crates/xerj-api/src/es_compat.rsindex abd953712..3fef19a2c 100644--- a/engine/crates/xerj-api/src/es_compat.rs+++ b/engine/crates/xerj-api/src/es_compat.rs@@ -14649,12 +14649,13 @@ async fn search_impl(// re-derives the mapping check per hit from `h.index` instead.//// The per-hit null-source case is likewise handled in- // `scroll_page_response`. `scroll_page_response` is shared with the- // `/:index/_search_scroll` route, whose first page does NOT apply the- // mapping check — so the per-hit mapping gate there is guarded by the- // `mapping_source_check` flag set below (true here, false for that- // route), keeping each route's continuation identical to its own first- // page. That route's broader `_source` gap is tracked in #659.+ // `scroll_page_response`, which is shared with the `/:index/_search_scroll`+ // route. Both real scroll-opening routes now apply this same per-hit+ // mapping suppression on their first page (`_search_scroll` since #659),+ // so both set `mapping_source_check: true`. That flag still guards the+ // shared continuation's mapping gate — it is `false` only for the+ // synthetic/test `ScrollContext`s that have no first page — keeping each+ // route's continuation identical to its own first page.let scroll_source_disabled = matches!(body.source, Some(Value::Bool(false)))|| (suppress_source_for_stored && body.source.is_none());let ctx = xerj_engine::engine::ScrollContext {@@ -20709,11 +20710,12 @@ async fn scroll_page_response(// at open). inner_hits still render from the collapse group snapshot// regardless.let source_disabled = ctx.source_disabled;- // #637: whether to ALSO apply the mapping `_source.enabled:false`- // suppression per hit below. Only true for scrolls opened by- // `search_impl` (whose first page applies it); false for the- // `_search_scroll` route, whose first page does not — so its- // continuation stays byte-identical to that first page (#659).+ // #637/#659: whether to ALSO apply the mapping `_source.enabled:false`+ // suppression per hit below. True for scrolls opened by either real+ // route — `search_impl` and, since #659, `_search_scroll` — both of+ // which apply it on their first page; false only for synthetic/test+ // contexts. Keeps the continuation byte-identical to the opening+ // route's own first page.let mapping_source_check = ctx.mapping_source_check;// Detect whether the initial search sorted by a non-score key;@@ -20849,12 +20851,12 @@ async fn scroll_page_response(// implied). The mapping `_source.enabled:false` case is// re-derived PER HIT from this hit's own index (a scroll// can span indices with divergent `_source` settings, and- // search_impl's first page suppresses it per-hit — #658- // skeptic), but ONLY when the opening route applies it on- // its first page (`mapping_source_check`); the- // `_search_scroll` route does not, so its continuation- // stays identical to its first page (#659). inner_hits are- // emitted regardless.+ // the first page suppresses it per-hit — #658 skeptic),+ // but ONLY when the opening route applies it on its first+ // page (`mapping_source_check`). Both real scroll routes+ // now do (search_impl, and `_search_scroll` since #659);+ // the flag is false only for synthetic contexts.+ // inner_hits are emitted regardless.if !source_disabled&& !(mapping_source_check && mapping_source_disabled(state, &h.index)){