Gabriel39 opened a new pull request #33520:
URL: https://github.com/apache/spark/pull/33520
<!--
Thanks for sending a pull request! Here are some tips for you:
1. If this is your first time, please read our contributor guidelines:
https://spark.apache.org/contributing.html
2. Ensure you have added or run the appropriate tests for your PR:
https://spark.apache.org/developer-tools.html
3. If the PR is unfinished, add '[WIP]' in your PR title, e.g.,
'[WIP][SPARK-XXXX] Your PR title ...'.
4. Be sure to keep the PR description updated to reflect all changes.
5. Please write your PR title to summarize what this PR proposes.
6. If possible, provide a concise example to reproduce the issue for a
faster review.
7. If you want to add a new configuration, please read the guideline first
for naming configurations in
'core/src/main/scala/org/apache/spark/internal/config/ConfigEntry.scala'.
8. If you want to add or modify an error type or message, please read the
guideline first in
'core/src/main/resources/error/README.md'.
-->
### What changes were proposed in this pull request?
Currently, RewriteDistinctAggregates rule will rewrite distinct aggregates
with extra Expand node. This causes unnecessary memory waste and performance
fallback for some specific aggregates(e.g. count distinct case when).
for query:
`
{{{
SELECT
cat1,
COUNT(DISTINCT CASE WHEN cond1 THEN cat2 ELSE null end) as cat2_cnt1,
COUNT(DISTINCT CASE WHEN cond2 THEN cat2 ELSE null end) as cat2_cnt2,
FROM
data
GROUP BY
key
}}}
`
Currently, we rewrite it to :
`
{{{
Aggregate(
key = ['key]
functions = [count(if (('gid = 1) and 'casewhen1) 'cat2_1 else null),
count(if (('gid = 2) and 'casewhen2) 'cat2_2 else null)]
output = ['key, 'cat2_cnt1, 'cat2_cnt2])
Aggregate(
key = ['key, 'casewhen1, 'casewhen2, 'gid]
functions = []
output = ['key, 'casewhen1, 'casewhen2, 'gid])
Expand(
projections = [('key, 'cat1, CASE WHEN cond1 THEN cat2 ELSE null END,
null, 1),
('key, 'cat1, 'null, CASE WHEN cond1 THEN cat2 ELSE
null END, 2)]
output = ['key, 'cat1, 'casewhen1, 'casewhen2, 'gid])
LocalTableScan [...]
}}}
`
This PR will improve performance and reduce memory waste. And query will be
rewrite to:
`
if # of case when expr < 64
{{{
Aggregate(
key = ['key]
functions = [count(if (01 & 'bit_vector != 0) 0 else null),
count(if (10 & 'bit_vector != 0) 0 else null)]
output = ['key, 'cat2_cnt1, 'cat2_cnt2])
Aggregate(
key = ['key, 'cat1]
functions = [bit_or(if (cond1) 01 else 00, if (cond2) 10 else 00)]
output = ['key, 'cat1, 'bit_vector])
LocalTableScan [...]
}}}
if # of case when expr >= 64
{{{
Aggregate(
key = ['key]
functions = [count(if ('sum1 >= 1) 1 else null),
count(if ('sum2 >= 1) 1 else null)]
output = ['key, 'cat1_cnt1, 'cat1_cnt2])
Aggregate(
key = ['key, 'cat1]
functions = [sum(if (cond1) 1 else 0), sum(if (cond2) 1 else 0)]
output = ['key, 'cat1, 'sum1, 'sum2])
LocalTableScan [...]
}}}
`
### Why are the changes needed?
we can rewrite count distinct case when more efficiently
### Does this PR introduce _any_ user-facing change?
no
### How was this patch tested?
UT in RewriteDistinctAggregatesSuite
--
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]
