Change array agg result from empty list to null if no row qualifed #11299
Conversation
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
jayzhan211
changed the title
jayzhan211
changed the title
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
|
|
||
| # TODO: Expect NULL, got empty list | ||
| query ? | ||
| select array_agg(distinct a) from t where a > 3; |
There was a problem hiding this comment.
This is the only query that has unexpected result.
I thought this should be equivalent to select array_agg(a) from t where a > 3 group by a;
There was a problem hiding this comment.
Thanks @jayzhan211 for calling this out!
Given there are no rows with a > 3, the aggregation input relation from t where a > 3 is empty, so this should produce same result as array_agg from empty table, ie NULL. currently it shows [] as a result.
There was a problem hiding this comment.
Maybe it is related to a multi-stage grouping (where the first phase outputs a NULL and then the second phase gets that NULL and interprets it as [] 🤔
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
| query III | ||
| WITH indices AS ( | ||
| SELECT 1 AS idx UNION ALL | ||
| SELECT 2 AS idx UNION ALL | ||
| SELECT 3 AS idx UNION ALL | ||
| SELECT 4 AS idx UNION ALL | ||
| SELECT 5 AS idx | ||
| ) | ||
| SELECT data.arr[indices.idx] as element, array_length(data.arr) as array_len, dummy | ||
| FROM ( | ||
| SELECT array_agg(distinct c2) as arr, count(1) as dummy FROM aggregate_test_100 | ||
| ) data | ||
| CROSS JOIN indices | ||
| ORDER BY 1 | ||
| ---- | ||
| 1 5 100 | ||
| 2 5 100 | ||
| 3 5 100 | ||
| 4 5 100 | ||
| 5 5 100 |
There was a problem hiding this comment.
I rewrite it to the simpler one!
|
|
||
| # TODO: Expect NULL, got empty list | ||
| query ? | ||
| select array_agg(distinct a) from t where a > 3; |
There was a problem hiding this comment.
Thanks @jayzhan211 for calling this out!
Given there are no rows with a > 3, the aggregation input relation from t where a > 3 is empty, so this should produce same result as array_agg from empty table, ie NULL. currently it shows [] as a result.
| statement ok | ||
| drop table t; | ||
|
|
||
| # test with no values |
There was a problem hiding this comment.
add array_agg(distinct case on empty table
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
Signed-off-by: jayzhan211 <jayzhan211@gmail.com>
Which issue does this PR close?
As @findepi pointed out in #11274 (comment) that most of the aggregate function does not return non-null result if no row qualified. I double check the result in Postgres and Duckdb and find out they does not return empty list for
array_agg. I think we can follow the behaviour as they did.I also hope this can make dealing with nullability simpler
Closes #.
Rationale for this change
What changes are included in this PR?
Are these changes tested?
Are there any user-facing changes?
The result of array agg is changed