Urmas Tamassy created HIVE-22660:
------------------------------------
Summary: HDFS path is encoded twice when inserting to a
partitioned table using dynamic partitions.
Key: HIVE-22660
URL: https://issues.apache.org/jira/browse/HIVE-22660
Project: Hive
Issue Type: Bug
Components: HiveServer2, Query Planning
Affects Versions: 2.1.1
Reporter: Urmas Tamassy
Attachments: details.txt
Tested on CDH 5.16.1 (Hive 1.1) and CDH 6.3 (Hive 2.2.1).
When file merging is enabled, ie. hive.merge.mapfiles or hive.merge.mapredfiles
is true (or in case of spark hive.merge.sparkfiles), under specific conditions
(see:
[ConditionalResolverMergeFiles|https://github.com/apache/hive/blob/e06bf8d273f3e391b269da4d3217f12bbc165cfc/ql/src/java/org/apache/hadoop/hive/ql/plan/ConditionalResolverMergeFiles.java#L207],
option 3.) on insert commands to partitioned tables, if Hive merges some
partition files in a second job and moves the ones that do not need to be
merged from the first jobs result, then the ones that can be moved are moved
twice. This seems to cause file paths to be encoded twice as well
([FileUtils#escapePathName()|https://github.com/apache/hive/blob/f37c5de6c32b9395d1b34fa3c02ed06d1bfbf6eb/common/src/java/org/apache/hadoop/hive/common/FileUtils.java#L271])
. For example, this can cause issues with partition values such as 'A:B' and
'C:D' to be encoded as 'A%3AB' for one partition path correctly and as
'C%253AD' for the other partition incorrectly.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
