adriangb opened a new pull request, #24187:
URL: https://github.com/apache/datafusion/pull/24187

   ## Which issue does this PR close?
   
   - Closes #23518.
   - Part of #23494.
   
   ## Rationale for this change
   
   Continues the per-plan proto hook migration (#23494, pattern established in 
#23495 and #23683): each plan/source owns its wire logic via `try_to_proto` / 
`try_from_proto` hooks instead of the central downcast/match chains in 
`datafusion-proto`.
   
   ## What changes are included in this PR?
   
   - `DataSource::try_to_proto` implemented for `MemorySourceConfig` and 
inherent `MemorySourceConfig::try_from_proto` added in 
`datafusion/datasource/src/memory.rs` (behind the `proto` feature). The 
record-batch IPC serde is pure Arrow and is inlined locally.
   - The `MemoryScan` decode arm in `datafusion-proto` now delegates to 
`MemorySourceConfig::try_from_proto`, and the central `MemoryScan` encode arm 
in `try_from_data_source_exec` is **deleted** — proof that the hook is the only 
path.
   - `try_into_memory_scan_physical_plan` becomes a deprecated shim delegating 
to the new hook; the now-unused `serialize_record_batches` / 
`parse_record_batches` helpers are deprecated.
   
   The wire format is byte-for-byte identical: I verified locally that a memory 
scan plan (with sort information, fetch, and `show_sizes=false`) encodes to 
identical bytes before and after this change, and that bytes encoded by each 
version decode correctly with the other.
   
   ## Are these changes tested?
   
   Yes. Existing roundtrip tests (`roundtrip_memory_source`, 
`roundtrip_memory_source_empty_projection`) now exercise the hooks since the 
central arms are gone. A new test 
`roundtrip_memory_source_sort_information_and_fetch` covers `sort_information`, 
`fetch`, and `show_sizes`, asserting on the decoded `MemorySourceConfig` 
directly rather than only the display string.
   
   ## Are there any user-facing changes?
   
   No behavior changes. 
`PhysicalPlanNodeExt::try_into_memory_scan_physical_plan`, 
`parse_record_batches`, and `serialize_record_batches` are deprecated (still 
functional).
   
   🤖 Generated with [Claude Code](https://claude.com/claude-code)


-- 
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]

Reply via email to