viirya opened a new issue, #23912:
URL: https://github.com/apache/datafusion/issues/23912
### Describe the bug
`DistinctPercentileContAccumulator` has two bugs, both stemming from reusing
the shared set-based `GenericDistinctBuffer` for an aggregate it doesn't fit:
**1. Panic on every `percentile_cont(DISTINCT ...)` query.** `update_batch`
forwards all argument columns to `GenericDistinctBuffer::update_batch`, which
asserts a single input array. `percentile_cont` always passes two columns (the
value and the percentile literal), so any distinct query panics. This path has
no existing test coverage, which is why it went unnoticed.
**2. Wrong results in sliding windows.** The buffer is a plain `HashSet`
with no per-value multiplicity, so `retract_batch` drops a value entirely when
one occurrence leaves the window frame, even when other rows in the frame still
carry that value.
### To Reproduce
```sql
CREATE TABLE t(id INT, x DOUBLE) AS VALUES (1,5),(2,5),(3,9);
-- Bug 1: panics with
-- "DistinctValuesBuffer::update_batch expects only a single input array
(left: 2, right: 1)"
SELECT percentile_cont(DISTINCT x, 0.5) FROM t;
-- Bug 2 (once Bug 1 is fixed): sliding window drops the still-present
duplicate 5
SELECT id, percentile_cont(DISTINCT x, 0.5)
OVER (ORDER BY id ROWS BETWEEN 1 PRECEDING AND CURRENT ROW)
FROM t ORDER BY id;
```
### Expected behavior
- Query 1: `7` (median of distinct {5, 9}).
- Query 2: `5, 5, 7` — at row 3 the frame is {5, 9} (the row-2 `5` is still
in-window), so the distinct set is {5, 9}, median 7.
### Actual behavior
- Query 1: **panic**.
- Query 2 (with Bug 1 worked around): row 3 returns **9** — `retract`
dropped `5` entirely when the row-1 `5` left, even though row-2's `5` remained.
### Additional context
Confirmed by running: unfixed code panics on query 1; with the fix, both
queries return the expected results and the full `aggregate.slt` suite passes.
--
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]
