Fix flaky test test_fuzz_async_reader_selection
#2656
Conversation
|
@tustvold FYI. This should fix the flaky test. |
| @@ -1015,7 +1014,7 @@ mod tests { | |||
| let mut selectors = vec![]; | |||
|
|
|||
| while total_rows < 7300 { | |||
| let row_count: usize = rand.gen_range(0..100); | |||
| let row_count: usize = rand.gen_range(1..100); | |||
There was a problem hiding this comment.
So RowSelector.row_count cannot be 0?
There was a problem hiding this comment.
It can if you manually construct one as we do in this test but in practice it would be non-sensical (and you wouldn't ever get one if you built a RowSelection from ranges or BooleanArrays). So depending on your perspective we could consider this a bug in the ParquetRecordBatchReader (probably, not sure where the actual issue comes up) but I think if we did the solution would be to just disallow row counts of 0 explicitly.
There was a problem hiding this comment.
Perhaps you could file a ticket describing the issue, I agree it isn't urgent but it should probably be looked into eventually (as it might unveil other issues).
|
Benchmark runs are scheduled for baseline = 31aaef2 and contender = a37c038. a37c038 is a master commit associated with this PR. Results will be available as each benchmark for each run completes. |
Which issue does this PR close?
Closes #.
Rationale for this change
Fixes flaky fuzz test for row selection. Was generating random row selections with
row_countequal to 0 which mucks things up.What changes are included in this PR?
Are there any user-facing changes?