You are viewing a plain text version of this content. The canonical link for it is here.
Posted to github@arrow.apache.org by "alamb (via GitHub)" <gi...@apache.org> on 2023/12/11 22:15:29 UTC
[PR] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
alamb opened a new pull request, #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505
## Which issue does this PR close?
This is part of https://github.com/apache/arrow-datafusion/issues/8451
## Rationale for this change
There are no good end to end tests of repartitioned scans that I could find (because the default size threshold for repartitioning is 10MB, and most tests use files smaller than this)
## What changes are included in this PR?
1. Add end to end tests for repartitioning
2. Fix a bug in sqllogictest harness that meant `--complete` mode didn't clear scratch files (I think @tustvold hit this at some point)
3. FIx a bug in `SINGLE_FILE_OUTPUT` that caused it to fail if the parent directory didn't exist
## Are these changes tested?
Yes -- there are several new tests added
## Are there any user-facing changes?
Two bug fixes as described abive
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb merged PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1425812238
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
Aah, I still need to fix this, this is a weird hold over from append
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1426782763
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
There is no reason you should need to create these files ahead of time anymore, it was a consequence of the head request being performed to accommodate the append semantics - I can add it to my list to clean it up
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#issuecomment-1854701110
Thank you for the review @tustvold 🙏
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1426784626
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
As in the object_store `LocalFile::put` implementation should automatically create the paths?
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1426772560
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
What needs to be fixed? I would like to write up a catalog
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#issuecomment-1851863641
The test is failing because the order of files listed is not consistent. I am working on a fix for these.
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "alamb (via GitHub)" <gi...@apache.org>.
alamb commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1423185942
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
Without this I can't write into `/dir/table_name/1.parquet` if `/dir/table_name` doesn't exist already
This is not enabled by deafult, it only affects when the user has specified `SINGLE_FILE_OUTPUT` explicitly
##########
datafusion/sqllogictest/test_files/repartition_scan.slt:
##########
@@ -0,0 +1,267 @@
+# Licensed to the Apache Software Foundation (ASF) under one
+# or more contributor license agreements. See the NOTICE file
+# distributed with this work for additional information
+# regarding copyright ownership. The ASF licenses this file
+# to you under the Apache License, Version 2.0 (the
+# "License"); you may not use this file except in compliance
+# with the License. You may obtain a copy of the License at
+
+# http://www.apache.org/licenses/LICENSE-2.0
+
+# Unless required by applicable law or agreed to in writing,
+# software distributed under the License is distributed on an
+# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
+# KIND, either express or implied. See the License for the
+# specific language governing permissions and limitations
+# under the License.
+
+##########
+# Tests for automatically reading files in parallel during scan
+##########
+
+# Set 4 partitions for deterministic output plans
+statement ok
+set datafusion.execution.target_partitions = 4;
+
+# automatically partition all files over 1 byte
+statement ok
+set datafusion.optimizer.repartition_file_min_size = 1;
+
+###################
+### Parquet tests
+###################
+
+# create a single parquet file
+statement ok
+COPY (VALUES (1), (2), (3), (4), (5)) TO 'test_files/scratch/repartition_scan/parquet_table/1.parquet'
+(FORMAT PARQUET, SINGLE_FILE_OUTPUT true);
+
+statement ok
+CREATE EXTERNAL TABLE parquet_table(column1 int)
+STORED AS PARQUET
+LOCATION 'test_files/scratch/repartition_scan/parquet_table/';
+
+query I
+select * from parquet_table;
+----
+1
+2
+3
+4
+5
+
+## Expect to see the scan read the file as "4" groups with even sizes (offsets)
+query TT
+EXPLAIN SELECT column1 FROM parquet_table WHERE column1 <> 42;
+----
+logical_plan
+Filter: parquet_table.column1 != Int32(42)
+--TableScan: parquet_table projection=[column1], partial_filters=[parquet_table.column1 != Int32(42)]
+physical_plan
+CoalesceBatchesExec: target_batch_size=8192
+--FilterExec: column1@0 != 42
+----ParquetExec: file_groups={4 groups: [[WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:0..101], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:101..202], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:202..303], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:303..403]]}, projection=[column1], predicate=column1@0 != 42, pruning_predicate=column1_min@0 != 42 OR 42 != column1_max@1
+
+# create a second parquet file
+statement ok
+COPY (VALUES (100), (200)) TO 'test_files/scratch/repartition_scan/parquet_table/2.parquet'
+(FORMAT PARQUET, SINGLE_FILE_OUTPUT true);
+
+## Still expect to see the scan read the file as "4" groups with even sizes. One group should read
+## parts of both files.
+query TT
+EXPLAIN SELECT column1 FROM parquet_table WHERE column1 <> 42 ORDER BY column1;
+----
+logical_plan
+Sort: parquet_table.column1 ASC NULLS LAST
+--Filter: parquet_table.column1 != Int32(42)
+----TableScan: parquet_table projection=[column1], partial_filters=[parquet_table.column1 != Int32(42)]
+physical_plan
+SortPreservingMergeExec: [column1@0 ASC NULLS LAST]
+--SortExec: expr=[column1@0 ASC NULLS LAST]
+----CoalesceBatchesExec: target_batch_size=8192
+------FilterExec: column1@0 != 42
+--------ParquetExec: file_groups={4 groups: [[WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:0..200], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:200..400], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:400..403, WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/2.parquet:0..197], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/2.parquet:197..394]]}, projection=[column1], predicate=column1@0 != 42, pruning_predicate=column1_min@0 != 42 OR 42 != column1_max@1
+
+
+## Read the files as though they are ordered
+
+statement ok
+CREATE EXTERNAL TABLE parquet_table_with_order(column1 int)
+STORED AS PARQUET
+LOCATION 'test_files/scratch/repartition_scan/parquet_table'
+WITH ORDER (column1 ASC);
+
+# output should be ordered
+query I
+SELECT column1 FROM parquet_table_with_order WHERE column1 <> 42 ORDER BY column1;
+----
+1
+2
+3
+4
+5
+100
+200
+
+# explain should not have any groups with more than one file
+# https://github.com/apache/arrow-datafusion/issues/8451
+query TT
+EXPLAIN SELECT column1 FROM parquet_table_with_order WHERE column1 <> 42 ORDER BY column1;
+----
+logical_plan
+Sort: parquet_table_with_order.column1 ASC NULLS LAST
+--Filter: parquet_table_with_order.column1 != Int32(42)
+----TableScan: parquet_table_with_order projection=[column1], partial_filters=[parquet_table_with_order.column1 != Int32(42)]
+physical_plan
+SortPreservingMergeExec: [column1@0 ASC NULLS LAST]
+--CoalesceBatchesExec: target_batch_size=8192
+----FilterExec: column1@0 != 42
+------ParquetExec: file_groups={4 groups: [[WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:0..200], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:200..400], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/1.parquet:400..403, WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/2.parquet:0..197], [WORKSPACE_ROOT/datafusion/sqllogictest/test_files/scratch/repartition_scan/parquet_table/2.parquet:197..394]]}, projection=[column1], predicate=column1@0 != 42, pruning_predicate=column1_min@0 != 42 OR 42 != column1_max@1
Review Comment:
This is actually incorrect because data from file1 appended to file2 is not sorted by column1 anymore -- https://github.com/apache/arrow-datafusion/issues/8451
(note the query gets the right answer because only the first group actually makes any values for 1.parquet as it is tiny)
##########
datafusion/sqllogictest/bin/sqllogictests.rs:
##########
@@ -159,6 +159,7 @@ async fn run_test_file_with_postgres(test_file: TestFile) -> Result<()> {
relative_path,
} = test_file;
info!("Running with Postgres runner: {}", path.display());
+ setup_scratch_dir(&relative_path)?;
Review Comment:
without this complete mode didn't clear the scratch dir resulting in confusing errors
--
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] Minor: Add repartition_file.slt end to end test for repartitioning files, and supporting tweaks [arrow-datafusion]
Posted by "tustvold (via GitHub)" <gi...@apache.org>.
tustvold commented on code in PR #8505:
URL: https://github.com/apache/arrow-datafusion/pull/8505#discussion_r1426826135
##########
datafusion/core/src/datasource/listing/url.rs:
##########
@@ -131,6 +131,10 @@ impl ListingTableUrl {
if is_directory {
fs::create_dir_all(path)?;
} else {
+ // ensure parent directory exists
Review Comment:
PR in https://github.com/apache/arrow-datafusion/pull/8539
--
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