fix: make write_csv/json/parquet cancel-safe
#5196
Conversation
| // Create cancellation-token to interrupt background execution on drop. | ||
| let cancellation_token = CancellationToken::new(); |
There was a problem hiding this comment.
Another way is to use abort-on-drop wrapper, I see this pattern is used in datafusion and there are built-in wrapper already. Let me know if you think it would be better
There was a problem hiding this comment.
I've decided not to introduce a new pattern here and changed it to AbortOnDropSingle per task.
You still can check how it looked like before: 3918d49
DDtKey
changed the title
write_csv/json/parquetwrite_csv/json/parquet cancel-safe
|
Another place with similar issue: #5197 |
|
Benchmark runs are scheduled for baseline = f63b972 and contender = d5077b5. d5077b5 is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
3 participants
Which issue does this PR close?
Closes #5178
Rationale for this change
Currently interruption from outside of datafusion don't stop spawned tasks and it continue to execute query.
It's very helpful to be able to restrict these methods by
timeoutorselectfor example.What changes are included in this PR?
Are these changes tested?
It's quite hard to cover with union-tests and such test could be too flaky (any ideas are welcome).
However all existed tests are passing successfully.
Are there any user-facing changes?