You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "mustafasrepo (via GitHub)" <gi...@apache.org> on 2023/11/20 14:41:05 UTC
[PR] Complex exprs requirement support [arrow-datafusion]
mustafasrepo opened a new pull request, #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281
## Which issue does this PR close?
<!--
We generally require a GitHub issue to be filed for all bug fixes and enhancements and this helps us generate change logs for our releases. You can link an issue to this PR using the GitHub syntax. For example `Closes #123` indicates that this PR will close issue #123.
-->
Improves the situation on [#8064](https://github.com/apache/arrow-datafusion/issues/8064)
## Rationale for this change
Currently we cannot determine whether ordering requirement of the complex expressions are satisfied.
Consider the case where requirement is `[a+b ASC]`, and we know that existing orderings are `[a ASC], [b ASC]`. It should be possible to understand that given `[a ASC] and [b ASC]`, requirement: `[a+b ASC]` is satisfied (Please note that we already have mechanism to make deduction. However, it is not considered in requirement satisfaction analysis. Currently we search requirement among existing orderings, this limits our ability during analysis).
<!--
Why are you proposing this change? If this is already explained clearly in the issue then this section is not needed.
Explaining clearly why changes are proposed helps reviewers understand your changes and offer better suggestions for fixes.
-->
## What changes are included in this PR?
This PR re-implements the `find_longest_permutation` and `ordering_satisfy_requirement` so that ordering of the complex(composed) expressions are calculated correctly, and considered during analysis. Most of the changes comes from the tests, and test related changes.
<!--
There is no need to duplicate the description in the issue here but it is sometimes worth providing a summary of the individual changes in this PR.
-->
## Are these changes tested?
Yes.
<!--
We typically require tests for all PRs in order to:
1. Prevent the code from being accidentally broken by subsequent changes
2. Serve as another way to document the expected behavior of the code
If tests are not included in your PR, please explain why (for example, are they covered by existing tests)?
-->
## Are there any user-facing changes?
<!--
If there are user-facing changes then we may require documentation to be updated before approving the PR.
-->
<!--
If there are any breaking changes to public APIs, please add the `api change` label.
-->
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak merged PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on code in PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#discussion_r1402684334
##########
datafusion/core/src/physical_optimizer/replace_with_order_preserving_variants.rs:
##########
@@ -357,15 +357,19 @@ mod tests {
let physical_plan =
sort_preserving_merge_exec(vec![sort_expr("a", &schema)], sort);
- let expected_input = ["SortPreservingMergeExec: [a@0 ASC NULLS LAST]",
+ let expected_input = [
Review Comment:
Just whitespace changes. Agree with your suggestion in general, we will try to improve our self discipline on that
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on code in PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#discussion_r1402675203
##########
datafusion/sqllogictest/test_files/groupby.slt:
##########
@@ -3842,6 +3842,51 @@ ProjectionExec: expr=[SUM(alias1)@1 as SUM(DISTINCT t1.x), MAX(alias1)@2 as MAX(
--------------------ProjectionExec: expr=[CAST(x@0 AS Float64) as CAST(t1.x AS Float64)t1.x, y@1 as y]
----------------------MemoryExec: partitions=1, partition_sizes=[1]
+# create an unbounded table that contains ordered timestamp.
+statement ok
+CREATE UNBOUNDED EXTERNAL TABLE csv_with_timestamps (
+ name VARCHAR,
+ ts TIMESTAMP
+)
+STORED AS CSV
+WITH ORDER (ts DESC)
+LOCATION '../core/tests/data/timestamps.csv'
+
+# below query should work in streaming mode.
+query TT
+EXPLAIN SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+logical_plan
+Limit: skip=0, fetch=5
+--Sort: time_chunks DESC NULLS FIRST, fetch=5
+----Projection: date_bin(Utf8("15 minutes"),csv_with_timestamps.ts) AS time_chunks
+------Aggregate: groupBy=[[date_bin(IntervalMonthDayNano("900000000000"), csv_with_timestamps.ts) AS date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)]], aggr=[[]]
+--------TableScan: csv_with_timestamps projection=[ts]
+physical_plan
+GlobalLimitExec: skip=0, fetch=5
+--SortPreservingMergeExec: [time_chunks@0 DESC], fetch=5
+----ProjectionExec: expr=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as time_chunks]
+------AggregateExec: mode=FinalPartitioned, gby=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------CoalesceBatchesExec: target_batch_size=2
+----------SortPreservingRepartitionExec: partitioning=Hash([date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0], 8), input_partitions=8, sort_exprs=date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 DESC
+------------AggregateExec: mode=Partial, gby=[date_bin(900000000000, ts@0) as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------------RepartitionExec: partitioning=RoundRobinBatch(8), input_partitions=1
+----------------StreamingTableExec: partition_sizes=1, projection=[ts], infinite_source=true, output_ordering=[ts@0 DESC]
+
+query P
+SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+2018-12-13T12:00:00
+2018-11-13T17:00:00
+
Review Comment:
Good idea. I added a test that shows the query works when the function is monotonic but not otherwise for unbounded sources. I also added a test that shows the query works regardless of the function's monotonicity for bounded sources.
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Complex exprs requirement support [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#issuecomment-1821172588
I will do another deep dive on this tomorrow
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Complex exprs requirement support [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#issuecomment-1821649236
Thank you @mustafasrepo -- I plan to review this carefully tomorrow
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on code in PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#discussion_r1402682340
##########
datafusion/core/src/physical_optimizer/enforce_distribution.rs:
##########
@@ -3787,7 +3787,7 @@ pub(crate) mod tests {
fn repartition_transitively_past_sort_with_projection_and_filter() -> Result<()> {
let schema = schema();
let sort_key = vec![PhysicalSortExpr {
- expr: col("c", &schema).unwrap(),
+ expr: col("a", &schema).unwrap(),
Review Comment:
In the plan, there is a filter enforcing `c = 0` -- so having a sort on `c` is a no-op logically. After recent improvements we are able to recognize things like that and remove sorts completely, so the test became moot in its old form.
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#issuecomment-1823469894
Thanks for the review @alamb. I answered your questions and added a couple tests to address your suggestions (verifying behavior difference for monotonic/nonmonotonic functions). Regarding your question, complex expression orderings are calculated via the `get_expr_ordering`/`update_ordering` functions which handle monotonic/nonmonotonic expressions.
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#issuecomment-1823480609
For some reason sqllogictests are failing in CI (they pass locally for me). Looking at the logs, it seems like there expected/actual are the same, so I can't figure out why this is the case. I'll let @mustafasrepo figure it out and fix tomorrow and we can merge then.
In the meantime if there is more suggestions for improvements from the community we will incorporate them as well
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#discussion_r1402147301
##########
datafusion/core/src/physical_optimizer/replace_with_order_preserving_variants.rs:
##########
@@ -357,15 +357,19 @@ mod tests {
let physical_plan =
sort_preserving_merge_exec(vec![sort_expr("a", &schema)], sort);
- let expected_input = ["SortPreservingMergeExec: [a@0 ASC NULLS LAST]",
+ let expected_input = [
Review Comment:
Did these plans actually change? Or is this just whitespace changes?
If it is just whitespace changes I would really appreciate breaking such changes out into their own PRs as they are much faster / easier to review and merge.
##########
datafusion/core/src/physical_optimizer/enforce_distribution.rs:
##########
@@ -3787,7 +3787,7 @@ pub(crate) mod tests {
fn repartition_transitively_past_sort_with_projection_and_filter() -> Result<()> {
let schema = schema();
let sort_key = vec![PhysicalSortExpr {
- expr: col("c", &schema).unwrap(),
+ expr: col("a", &schema).unwrap(),
Review Comment:
why was this test changed?
##########
datafusion/sqllogictest/test_files/groupby.slt:
##########
@@ -3842,6 +3842,51 @@ ProjectionExec: expr=[SUM(alias1)@1 as SUM(DISTINCT t1.x), MAX(alias1)@2 as MAX(
--------------------ProjectionExec: expr=[CAST(x@0 AS Float64) as CAST(t1.x AS Float64)t1.x, y@1 as y]
----------------------MemoryExec: partitions=1, partition_sizes=[1]
+# create an unbounded table that contains ordered timestamp.
+statement ok
+CREATE UNBOUNDED EXTERNAL TABLE csv_with_timestamps (
+ name VARCHAR,
+ ts TIMESTAMP
+)
+STORED AS CSV
+WITH ORDER (ts DESC)
+LOCATION '../core/tests/data/timestamps.csv'
+
+# below query should work in streaming mode.
+query TT
+EXPLAIN SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+logical_plan
+Limit: skip=0, fetch=5
+--Sort: time_chunks DESC NULLS FIRST, fetch=5
+----Projection: date_bin(Utf8("15 minutes"),csv_with_timestamps.ts) AS time_chunks
+------Aggregate: groupBy=[[date_bin(IntervalMonthDayNano("900000000000"), csv_with_timestamps.ts) AS date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)]], aggr=[[]]
+--------TableScan: csv_with_timestamps projection=[ts]
+physical_plan
+GlobalLimitExec: skip=0, fetch=5
+--SortPreservingMergeExec: [time_chunks@0 DESC], fetch=5
+----ProjectionExec: expr=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as time_chunks]
+------AggregateExec: mode=FinalPartitioned, gby=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------CoalesceBatchesExec: target_batch_size=2
+----------SortPreservingRepartitionExec: partitioning=Hash([date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0], 8), input_partitions=8, sort_exprs=date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 DESC
+------------AggregateExec: mode=Partial, gby=[date_bin(900000000000, ts@0) as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------------RepartitionExec: partitioning=RoundRobinBatch(8), input_partitions=1
+----------------StreamingTableExec: partition_sizes=1, projection=[ts], infinite_source=true, output_ordering=[ts@0 DESC]
+
+query P
+SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+2018-12-13T12:00:00
+2018-11-13T17:00:00
+
Review Comment:
Could you also please add a negative test that can not work in streaming mode with an expression that doesn't preserve the order?
perhaps something like
```
SELECT extract(month from ts) as months,
FROM csv_with_timestamps
GROUP BY extract(month from ts)
ORDER BY time_chunks DESC
LIMIT 5
```
##########
datafusion/sqllogictest/test_files/groupby.slt:
##########
@@ -3842,6 +3842,51 @@ ProjectionExec: expr=[SUM(alias1)@1 as SUM(DISTINCT t1.x), MAX(alias1)@2 as MAX(
--------------------ProjectionExec: expr=[CAST(x@0 AS Float64) as CAST(t1.x AS Float64)t1.x, y@1 as y]
----------------------MemoryExec: partitions=1, partition_sizes=[1]
+# create an unbounded table that contains ordered timestamp.
+statement ok
+CREATE UNBOUNDED EXTERNAL TABLE csv_with_timestamps (
+ name VARCHAR,
+ ts TIMESTAMP
+)
+STORED AS CSV
+WITH ORDER (ts DESC)
+LOCATION '../core/tests/data/timestamps.csv'
+
+# below query should work in streaming mode.
+query TT
+EXPLAIN SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+logical_plan
+Limit: skip=0, fetch=5
+--Sort: time_chunks DESC NULLS FIRST, fetch=5
+----Projection: date_bin(Utf8("15 minutes"),csv_with_timestamps.ts) AS time_chunks
+------Aggregate: groupBy=[[date_bin(IntervalMonthDayNano("900000000000"), csv_with_timestamps.ts) AS date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)]], aggr=[[]]
+--------TableScan: csv_with_timestamps projection=[ts]
+physical_plan
+GlobalLimitExec: skip=0, fetch=5
+--SortPreservingMergeExec: [time_chunks@0 DESC], fetch=5
+----ProjectionExec: expr=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as time_chunks]
+------AggregateExec: mode=FinalPartitioned, gby=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------CoalesceBatchesExec: target_batch_size=2
+----------SortPreservingRepartitionExec: partitioning=Hash([date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0], 8), input_partitions=8, sort_exprs=date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 DESC
+------------AggregateExec: mode=Partial, gby=[date_bin(900000000000, ts@0) as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------------RepartitionExec: partitioning=RoundRobinBatch(8), input_partitions=1
+----------------StreamingTableExec: partition_sizes=1, projection=[ts], infinite_source=true, output_ordering=[ts@0 DESC]
+
+query P
+SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
Review Comment:
I am confused about the seeming extra `(` and `)` here and in the above query:
Is there any reason it is not:
```suggestion
GROUP BY date_bin('15 minutes', ts)
```
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org
Re: [PR] Calculate ordering equivalence for expressions (rather than just columns) [arrow-datafusion]
Posted by "ozankabak (via GitHub)" <gi...@apache.org>.
ozankabak commented on code in PR #8281:
URL: https://github.com/apache/arrow-datafusion/pull/8281#discussion_r1402676110
##########
datafusion/sqllogictest/test_files/groupby.slt:
##########
@@ -3842,6 +3842,51 @@ ProjectionExec: expr=[SUM(alias1)@1 as SUM(DISTINCT t1.x), MAX(alias1)@2 as MAX(
--------------------ProjectionExec: expr=[CAST(x@0 AS Float64) as CAST(t1.x AS Float64)t1.x, y@1 as y]
----------------------MemoryExec: partitions=1, partition_sizes=[1]
+# create an unbounded table that contains ordered timestamp.
+statement ok
+CREATE UNBOUNDED EXTERNAL TABLE csv_with_timestamps (
+ name VARCHAR,
+ ts TIMESTAMP
+)
+STORED AS CSV
+WITH ORDER (ts DESC)
+LOCATION '../core/tests/data/timestamps.csv'
+
+# below query should work in streaming mode.
+query TT
+EXPLAIN SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
+ ORDER BY time_chunks DESC
+ LIMIT 5
+----
+logical_plan
+Limit: skip=0, fetch=5
+--Sort: time_chunks DESC NULLS FIRST, fetch=5
+----Projection: date_bin(Utf8("15 minutes"),csv_with_timestamps.ts) AS time_chunks
+------Aggregate: groupBy=[[date_bin(IntervalMonthDayNano("900000000000"), csv_with_timestamps.ts) AS date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)]], aggr=[[]]
+--------TableScan: csv_with_timestamps projection=[ts]
+physical_plan
+GlobalLimitExec: skip=0, fetch=5
+--SortPreservingMergeExec: [time_chunks@0 DESC], fetch=5
+----ProjectionExec: expr=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as time_chunks]
+------AggregateExec: mode=FinalPartitioned, gby=[date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------CoalesceBatchesExec: target_batch_size=2
+----------SortPreservingRepartitionExec: partitioning=Hash([date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0], 8), input_partitions=8, sort_exprs=date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)@0 DESC
+------------AggregateExec: mode=Partial, gby=[date_bin(900000000000, ts@0) as date_bin(Utf8("15 minutes"),csv_with_timestamps.ts)], aggr=[], ordering_mode=Sorted
+--------------RepartitionExec: partitioning=RoundRobinBatch(8), input_partitions=1
+----------------StreamingTableExec: partition_sizes=1, projection=[ts], infinite_source=true, output_ordering=[ts@0 DESC]
+
+query P
+SELECT date_bin('15 minutes', ts) as time_chunks
+ FROM csv_with_timestamps
+ GROUP BY (date_bin('15 minutes', ts))
Review Comment:
No reason, it was unnecessary -- now removed
--
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: github-unsubscribe@arrow.apache.org
For queries about this service, please contact Infrastructure at:
users@infra.apache.org