Conversation
![chatgpt-codex-connector[bot]](https://avatars.githubusercontent.com/in/1144995?s=60&v=4){kind=small}
There was a problem hiding this comment.
💡 Codex Review
Here are some automated review suggestions for this pull request.
Reviewed commit: bdb9429b25
ℹ️ About Codex in GitHub
Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you
- Open a pull request for review
- Mark a draft as ready
- Comment "@codex review".
If Codex has suggestions, it will comment; otherwise it will react with 👍.
When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".
| true, | ||
| None, | ||
| )?; | ||
| all_queries_empty &= query.as_any().is::<EmptyQuery>(); |
There was a problem hiding this comment.
When an index contains splits created before a tokenizer mapping update, this query is compiled only against the latest doc_mapper.schema(). For example, if an old split indexed "-" with the raw tokenizer and the field later switched to the default tokenizer, a full-text query for "-" becomes an EmptyQuery under the latest schema even though compiling it against the old split schema would match that document. Rewriting the request to MatchNone then skips every split and silently drops valid hits; emptiness must be established independently of schema evolution or checked against the actual split schemas.
Useful? React with 👍 / 👎.
| true, | ||
| None, | ||
| )?; | ||
| all_queries_empty &= query.as_any().is::<EmptyQuery>(); |
There was a problem hiding this comment.
When a no-match clause is wrapped in QueryAst::Boost, such as an explicitly boosted MatchNone or a boosted full-text query that produces zero terms, query construction returns a BoostQuery containing an EmptyQuery; consequently this concrete-type check is false. Boolean simplification also treats that wrapper as an opaque leaf, so these requests still list and search every split despite being guaranteed to return no hits. Propagate semantic emptiness through query wrappers rather than relying only on a top-level EmptyQuery downcast.
Useful? React with 👍 / 👎.
2 participants
Summary
MatchNonequeries before querying the metastore for splits.Testing
make fmtcd quickwit && cargo nextest run -p quickwit-search(210 tests passed)