[
https://issues.apache.org/jira/browse/HIVE-22977?focusedWorklogId=837801&page=com.atlassian.jira.plugin.system.issuetabpanels:worklog-tabpanel#worklog-837801
]
ASF GitHub Bot logged work on HIVE-22977:
-----------------------------------------
Author: ASF GitHub Bot
Created on: 09/Jan/23 05:12
Start Date: 09/Jan/23 05:12
Worklog Time Spent: 10m
Work Description: SourabhBadhya commented on code in PR #3801:
URL: https://github.com/apache/hive/pull/3801#discussion_r1064288785
##########
ql/src/java/org/apache/hadoop/hive/ql/txn/compactor/QueryCompactor.java:
##########
@@ -281,5 +283,146 @@ static void overrideConfProps(HiveConf conf,
CompactionInfo ci, Map<String, Stri
conf.set(property, entry.getValue());
});
}
+
+ /**
+ * Returns whether merge compaction must be enabled or not.
+ * @param conf Hive configuration
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @param storageDescriptor storage descriptor of the underlying table
+ * @return true, if merge compaction must be enabled
+ */
+ static boolean isMergeCompaction(HiveConf conf, AcidDirectory directory,
+ ValidWriteIdList validWriteIdList,
+ StorageDescriptor storageDescriptor) {
+ return conf.getBoolVar(HiveConf.ConfVars.HIVE_MERGE_COMPACTION_ENABLED)
+ && !hasDeleteOrAbortedDirectories(directory, validWriteIdList)
+ &&
storageDescriptor.getOutputFormat().equalsIgnoreCase("org.apache.hadoop.hive.ql.io.orc.OrcOutputFormat");
+ }
+
+ /**
+ * Scan a directory for delete deltas or aborted directories.
+ * @param directory the directory to be scanned
+ * @param validWriteIdList list of valid write IDs
+ * @return true, if delete or aborted directory found
+ */
+ static boolean hasDeleteOrAbortedDirectories(AcidDirectory directory,
ValidWriteIdList validWriteIdList) {
+ if (!directory.getCurrentDirectories().isEmpty()) {
+ final long minWriteId = validWriteIdList.getMinOpenWriteId() == null ?
1 : validWriteIdList.getMinOpenWriteId();
+ final long maxWriteId = validWriteIdList.getHighWatermark();
+ return directory.getCurrentDirectories().stream()
+ .filter(AcidUtils.ParsedDeltaLight::isDeleteDelta)
+ .filter(delta -> delta.getMinWriteId() >= minWriteId)
+ .anyMatch(delta -> delta.getMaxWriteId() <= maxWriteId) ||
!directory.getAbortedDirectories().isEmpty();
+ }
+ return true;
+ }
+
+ /**
+ * Collect the list of all bucket file paths, which belong to the same
bucket Id. This method scans all the base
+ * and delta dirs.
+ * @param conf hive configuration, must be not null
+ * @param dir the root directory of delta dirs
+ * @param includeBaseDir true, if the base directory should be scanned
+ * @param isMm
+ * @return map of bucket ID -> bucket files
+ * @throws IOException an error happened during the reading of the
directory/bucket file
+ */
+ private static Map<Integer, List<Reader>>
matchBucketIdToBucketFiles(HiveConf conf, AcidDirectory dir,
+ boolean
includeBaseDir, boolean isMm) throws IOException {
+ Map<Integer, List<Reader>> result = new HashMap<>();
+ if (includeBaseDir && dir.getBaseDirectory() != null) {
+ getBucketFiles(conf, dir.getBaseDirectory(), isMm, result);
+ }
+ for (AcidUtils.ParsedDelta deltaDir : dir.getCurrentDirectories()) {
+ Path deltaDirPath = deltaDir.getPath();
+ getBucketFiles(conf, deltaDirPath, isMm, result);
+ }
+ return result;
+ }
+
+ /**
+ * Collect the list of all bucket file paths, which belong to the same
bucket Id. This method checks only one
+ * directory.
+ * @param conf hive configuration, must be not null
+ * @param dirPath the directory to be scanned.
+ * @param isMm collect bucket files fron insert only directories
+ * @param bucketIdToBucketFilePath the result of the scan
+ * @throws IOException an error happened during the reading of the
directory/bucket file
+ */
+ private static void getBucketFiles(HiveConf conf, Path dirPath, boolean
isMm, Map<Integer, List<Reader>> bucketIdToBucketFilePath) throws IOException {
+ FileSystem fs = dirPath.getFileSystem(conf);
+ FileStatus[] fileStatuses =
+ fs.listStatus(dirPath, isMm ? AcidUtils.originalBucketFilter :
AcidUtils.bucketFileFilter);
+ for (FileStatus f : fileStatuses) {
+ final Path fPath = f.getPath();
+ Matcher matcher = isMm ? AcidUtils.LEGACY_BUCKET_DIGIT_PATTERN
+ .matcher(fPath.getName()) :
AcidUtils.BUCKET_PATTERN.matcher(fPath.getName());
+ if (!matcher.find()) {
+ String errorMessage = String
+ .format("Found a bucket file matching the bucket pattern! %s
Matcher=%s", fPath.toString(),
Review Comment:
Added the following - "Found a bucket file which did not match the bucket
pattern!" . Done.
Issue Time Tracking
-------------------
Worklog Id: (was: 837801)
Time Spent: 2h 40m (was: 2.5h)
> Merge delta files instead of running a query in major/minor compaction
> ----------------------------------------------------------------------
>
> Key: HIVE-22977
> URL: https://issues.apache.org/jira/browse/HIVE-22977
> Project: Hive
> Issue Type: Improvement
> Reporter: László Pintér
> Assignee: Sourabh Badhya
> Priority: Major
> Labels: pull-request-available
> Attachments: HIVE-22977.01.patch, HIVE-22977.02.patch
>
> Time Spent: 2h 40m
> Remaining Estimate: 0h
>
> [Compaction Optimiziation]
> We should analyse the possibility to move a delta file instead of running a
> major/minor compaction query.
> Please consider the following use cases:
> - full acid table but only insert queries were run. This means that no
> delete delta directories were created. Is it possible to merge the delta
> directory contents without running a compaction query?
> - full acid table, initiating queries through the streaming API. If there
> are no abort transactions during the streaming, is it possible to merge the
> delta directory contents without running a compaction query?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
