[SPARK-9974] [BUILD] [SQL] Makes sure com.twitter:parquet-hadoop-bundle:1.6.0 is in SBT assembly jar#8198
Closed
liancheng wants to merge 1 commit into
Closed
Conversation
|
Test build #40867 timed out for PR 8198 at commit |
Contributor
Author
|
retest this please |
|
Test build #40884 has finished for PR 8198 at commit
|
Contributor
|
I've merged this. |
asfgit
pushed a commit
that referenced
this pull request
Aug 18, 2015
…le:1.6.0 is in SBT assembly jar PR #7967 enables Spark SQL to persist Parquet tables in Hive compatible format when possible. One of the consequence is that, we have to set input/output classes to `MapredParquetInputFormat`/`MapredParquetOutputFormat`, which rely on com.twitter:parquet-hadoop:1.6.0 bundled with Hive 1.2.1. When loading such a table in Spark SQL, `o.a.h.h.ql.metadata.Table` first loads these input/output format classes, and thus classes in com.twitter:parquet-hadoop:1.6.0. However, the scope of this dependency is defined as "runtime", and is not packaged into Spark assembly jar. This results in a `ClassNotFoundException`. This issue can be worked around by asking users to add parquet-hadoop 1.6.0 via the `--driver-class-path` option. However, considering Maven build is immune to this problem, I feel it can be confusing and inconvenient for users. So this PR fixes this issue by changing scope of parquet-hadoop 1.6.0 to "compile". Author: Cheng Lian <lian@databricks.com> Closes #8198 from liancheng/spark-9974/bundle-parquet-1.6.0. (cherry picked from commit 52ae952) Signed-off-by: Reynold Xin <rxin@databricks.com>
This file contains hidden or 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
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
3 participants
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Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be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.
PR #7967 enables Spark SQL to persist Parquet tables in Hive compatible format when possible. One of the consequence is that, we have to set input/output classes to
MapredParquetInputFormat/MapredParquetOutputFormat, which rely on com.twitter:parquet-hadoop:1.6.0 bundled with Hive 1.2.1.When loading such a table in Spark SQL,
o.a.h.h.ql.metadata.Tablefirst loads these input/output format classes, and thus classes in com.twitter:parquet-hadoop:1.6.0. However, the scope of this dependency is defined as "runtime", and is not packaged into Spark assembly jar. This results in aClassNotFoundException.This issue can be worked around by asking users to add parquet-hadoop 1.6.0 via the
--driver-class-pathoption. However, considering Maven build is immune to this problem, I feel it can be confusing and inconvenient for users.So this PR fixes this issue by changing scope of parquet-hadoop 1.6.0 to "compile".