PARQUET-138: Allow merging more restrictive field in less restrictive field #550
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
rdblue
reviewed
Nov 14, 2018
parquet-column/src/test/java/org/apache/parquet/schema/TestMessageType.java
Outdated
Show resolved
Hide resolved
|
With the change I made, merging something less restrictive into something
more restrictive would fail, while the opposite would now succeed. So
merging is dependent on the “direction” of the merge: you merge X into Y.
If we want merging to always succeed, it is also possible, and in this case
there would not be any “direction”: you merge X and Y.
What do you think the behavior should be?
…On Wednesday, November 14, 2018, Ryan Blue ***@***.***> wrote:
***@***.**** commented on this pull request.
------------------------------
In parquet-column/src/test/java/org/apache/parquet/schema/
TestMessageType.java
<#550 (comment)>:
> @@ -90,15 +90,15 @@ public void testMergeSchema() {
new PrimitiveType(REQUIRED, BINARY, "a"));
try {
- t3.union(t4);
+ t4.union(t3);
Why is this dependent on which schema is the left-hand operand?
—
You are receiving this because you authored the thread.
Reply to this email directly, view it on GitHub
<#550 (review)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGvZWRPeRf6tNazWox_OFtuK0PhBb8IGks5uvF_wgaJpZM4YdwMn>
.
|
|
I see no reason why union would not by symmetric. Otherwise, this is just changing the direction, which isn't very useful. |
|
I see no reason either, actually. Updated the PR to change that! |
parquet-column/src/test/java/org/apache/parquet/schema/TestType.java
Outdated
Show resolved
Hide resolved
|
@julienledem, can you think of any problems that may result from making this change? I think it just allows unioning more types together, but I'm not sure where exactly we rely on this and whether parts assume the old behavior. |
|
Hey team, I have came across this issue in my project and I was wondering if you are likely to make this change anytime soon? Thanks! |
|
@tpmusielak, we need to make sure we don't think this behavior change will cause problems first. I'd like to see it in the next release. |
|
What's required to get this in for the next release? Do we need more tests to ensure we don't break existing behavior? |
|
Can we put this behind a flag to stay backward compatible and users can opt-in to use it? |
|
Hey @julienledem @rdblue - friendly ping here and agreed it would be nice to get this into the next release. If this looks good, can we merge? |
|
Thanks for the reminder, @vinooganesh. I see no reason why it would not be safe for union to return the same result when operand position is swapped. |
|
hey guys, |
|
@boriskar I guess this is more a question for apache/spark. This change is available in parquet 1.11.0 or higher, so spark needs to bump its version to that (it's currently on 1.10.1). |
shangxinli
added a commit
to shangxinli/parquet-mr
that referenced
this pull request
Mar 1, 2023
Summary: This is to merge from upstream PARQUET-1305: Backward incompatible change introduced in 1.8 (apache#483) PARQUET-1452: Deprecate old logical types API (apache#535) PARQUET-1414: Simplify next row count check calculation (apache#537) PARQUET-1435: Benchmark filtering column-indexes (apache#536) PARQUET-1365: Don't write page level statistics (apache#549) Page level statistics were never used in production and became pointless after adding column indexes. PARQUET-1456: Use page index, ParquetFileReader throw ArrayIndexOutOfBoundsException (apache#548) The usage of static caching in the page index implementation did not allow using multiple readers at the same time. PARQUET-1407: Avro: Fix binary values returned from dictionary encoding (apache#552) * PARQUET-1407: Add test case for PARQUET-1407 to demonstrate the issue * PARQUET-1407: Fix binary values from dictionary encoding. Closes apache#551. PARQUET-1460: Fix javadoc errors and include javadoc checking in Travis checks (apache#554) Experiment. Revert "Experiment." This reverts commit 97a880c. PARQUET-1434: Update CHANGES.md for 1.11.0 release. [maven-release-plugin] prepare release apache-parquet-1.11.0 [maven-release-plugin] prepare for next development iteration PARQUET-1461: Third party code does not compile after parquet-mr minor version update (apache#556) PARQUET-1434: Update CHANGES.md for 1.11.0 release candidate 2. PARQUET-1258: Update scm developer connection to github [maven-release-plugin] prepare release apache-parquet-1.11.0 [maven-release-plugin] prepare for next development iteration PARQUET-1462: Allow specifying new development version in prepare-release.sh (apache#557) Before this change, prepare-release.sh only took the release version as a parameter, the new development version was asked interactively for each individual pom.xml file, which made answering them tedious. PARQUET-1472: Dictionary filter fails on FIXED_LEN_BYTE_ARRAY (apache#562) PARQUET-1474: Less verbose and lower level logging for missing column/offset indexes (apache#563) PARQUET-1476: Don't emit a warning message for files without new logical type (apache#577) Update CHANGES.md for 1.11.0 release candidate 2. [maven-release-plugin] prepare release apache-parquet-1.11.0 [maven-release-plugin] prepare for next development iteration PARQUET-1478: Can't read spec compliant, 3-level lists via parquet-proto (apache#578) PARQUET-1489: Insufficient documentation for UserDefinedPredicate.keep(T) (apache#588) PARQUET-1487: Do not write original type for timezone-agnostic timestamps (apache#585) Update CHANGES.md for 1.11.0 release candidate 3. [maven-release-plugin] prepare release apache-parquet-1.11.0 [maven-release-plugin] prepare for next development iteration PARQUET-1490: Add branch-specific Travis steps (apache#590) The possiblity of branch-specific scripts allows feature branches to build SNAPSHOT versions of parquet-format (and depend on them in the POM files). Even if such branch-specific scripts get merged into master accidentally, they will not have any effect there. The script for the main branch checks the POM files to make sure that SNAPSHOT dependencies are not added to or merged into master accidentally. PARQUET-1280: [parquet-protobuf] Use maven protoc plugin (apache#506) PARQUET-1466: Upgrade to the latest guava 27.0-jre (apache#559) PARQUET-1475: Fix lack of cause propagation in DirectCodecFactory.ParquetCompressionCodecException. (apache#564) PARQUET-1492: Remove protobuf build (apache#592) We do not need to build protobuf (protoc) ourselves since we rely on maven protoc plugin to compile protobuf. This should save about 10 minutes travis build time (time for building protobuf itself). PARQUET-1498: Add instructions to install thrift via homebrew (apache#595) PARQUET-1502: Convert FIXED_LEN_BYTE_ARRAY to arrow type in logicalTypeAnnotation if it is not null (apache#593) [PARQUET-1506] Migrate maven-thrift-plugin to thrift-maven-plugin (apache#600) maven-thrift-plugin (Aug 13, 2013) https://mvnrepository.com/artifact/org.apache.thrift.tools/maven-thrift-plugin/0.1.11 thrift-maven-plugin (Jan 18, 2017) https://mvnrepository.com/artifact/org.apache.thrift/thrift-maven-plugin/0.10.0 The maven-thrift-plugin is the old one which has been migrated to the ASF and continued as thrift-maven-plugin: https://issues.apache.org/jira/browse/THRIFT-4083 [PARQUET-1500] Replace Closeables with try-with-resources (apache#597) PARQUET-1503: Remove Ints Utility Class (apache#598) PARQUET-1513: Update HiddenFileFilter to avoid extra startsWith (apache#606) PARQUET-1504: Add an option to convert Int96 to Arrow Timestamp (apache#594) PARQUET-1504: Add an option to convert Parquet Int96 to Arrow Timestamp PARQUET-1509: Note Hive deprecation in README. (apache#602) PARQUET-1510: Fix notEq for optional columns with null values. (apache#603) Dictionaries cannot contain null values, so notEq filters cannot conclude that a block cannot match using only the dictionary. Instead, it must also check whether the block may have at least one null value. If there are no null values, then the existing check is correct. [PARQUET-1507] Bump Apache Thrift to 0.12.0 (apache#601) PARQUET-1518: Use Jackson2 version 2.9.8 in parquet-cli (apache#609) There are some vulnerabilities: https://ossindex.sonatype.org/vuln/1205a1ec-0837-406f-b081-623b9fb02992 https://ossindex.sonatype.org/vuln/b85a00e3-7d9b-49cf-9b19-b73f8ee60275 https://ossindex.sonatype.org/vuln/4f7e98ad-2212-45d3-ac21-089b3b082e6c https://ossindex.sonatype.org/vuln/ab9013f0-09a2-4f01-bce5-751dc7437494 https://ossindex.sonatype.org/vuln/3f596fc0-9615-4b93-b30a-d4e0532e667f https://ossindex.sonatype.org/vuln/4f7e98ad-2212-45d3-ac21-089b3b082e6c PARQUET-138: Allow merging more restrictive field in less restrictive field (apache#550) * Allow merging more restrictive field in less restrictive field * Make class and function names more explicit Add javax.annotation-api dependency for JDK >= 9 (apache#604) PARQUET-1470: Inputstream leakage in ParquetFileWriter.appendFile (apache#611) PARQUET-1514: ParquetFileWriter Records Compressed Bytes instead of Uncompressed Bytes (apache#607) PARQUET-1505: Use Java 7 NIO StandardCharsets (apache#599) PARQUET-1480 INT96 to avro not yet implemented error should mention deprecation (apache#579) PARQUET-1485: Fix Snappy direct memory leak (apache#581) PARQUET-1527: [parquet-tools] cat command throw java.lang.ClassCastException (apache#612) PARQUET-1529: Shade fastutil in all modules where used (apache#617) Update CHANGES.md for 1.11.0rc4 [maven-release-plugin] prepare release apache-parquet-1.11.0 [maven-release-plugin] prepare for next development iteration Merge from upstream Test Plan: Testing in Spark/Hive/Presto wil be performed before roll out to production! Reviewers: pavi, leisun Reviewed By: leisun Differential Revision: https://code.uberinternal.com/D2512639 Revert "PARQUET-1485: Fix Snappy direct memory leak (apache#581)" This reverts commit 7dcdcdc.
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
https://issues.apache.org/jira/browse/PARQUET-138