Abhishek-Mohan opened a new pull request, #7162: URL: https://github.com/apache/datafusion-comet/pull/7162
## Which issue does this PR close? Part of #4419. ## Rationale for this change This applies the #4419 vocabulary to the JVM shuffle guide. The rule IDs refer to the style guide in #7151. ## What changes are included in this PR? - **"JVM code with native encoding" now says which part runs where.** Partitioning runs on the JVM, and Arrow IPC encoding runs in Rust (`writeSortedFileNative` and related functions in `jni_api.rs`). - **"Comet native operator" → "Comet operator"** for a shuffle's child, since that child can be JVM-side. "Rust-implemented operator" is used only for the direct-read consumer, which really is part of a Rust plan. - **"on the native path" → "with native shuffle".** The diagram labels are updated to match. - **Fixed the description of `spark.comet.shuffle.directRead.enabled`.** "Decode blocks natively" was misleading, because both read paths decode in Rust. Direct read decodes inside the Rust plan and bypasses Arrow FFI, which is what the config's own doc string says. - **Unchanged:** "native shuffle" (N2), "native code" (N3), and identifiers (N6). ## How are these changes tested? Docs only. - `prettier --check` passes. - I checked each changed claim against `main`. - I built the docs locally with Sphinx. AI assistance: OpenCode helped check the page against the code and draft the edits. -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected] --------------------------------------------------------------------- To unsubscribe, e-mail: [email protected] For additional commands, e-mail: [email protected]
