andygrove opened a new pull request, #1633:
URL: https://github.com/apache/datafusion-ballista/pull/1633
# Which issue does this PR close?
Closes #.
# Rationale for this change
#1623 made sort-based shuffle the default, but follow-up benchmarking shows
the hash-based writer is consistently faster on TPC-H even after the regression
fix in #1626. Until sort-based shuffle catches up on performance, it should
remain opt-in so users get the faster path by default.
# What changes are included in this PR?
Reverts #1623. The default value of `ballista.shuffle.sort_based.enabled`
returns to `false`, and the test golden plans, dot-graph labels, and user-guide
docs are restored to describe hash-based shuffle as the default.
# Are there any user-facing changes?
Yes. Users who were relying on sort-based shuffle being on by default will
need to opt in explicitly:
```rust
let session_config = SessionConfig::new_with_ballista()
.set_bool("ballista.shuffle.sort_based.enabled", true);
```
No public API signatures change, so no `api change` label is needed.
--
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]
