bhasudha commented on a change in pull request #689: [HUDI-25] Optimize
HoodieInputFormat.listStatus for faster Hive Incremental queries
URL: https://github.com/apache/incubator-hudi/pull/689#discussion_r287534561
##########
File path:
hoodie-hadoop-mr/src/main/java/com/uber/hoodie/hadoop/HoodieInputFormat.java
##########
@@ -62,65 +66,53 @@
@Override
public FileStatus[] listStatus(JobConf job) throws IOException {
- // Get all the file status from FileInputFormat and then do the filter
- FileStatus[] fileStatuses = super.listStatus(job);
- Map<HoodieTableMetaClient, List<FileStatus>> groupedFileStatus =
groupFileStatus(fileStatuses);
- LOG.info("Found a total of " + groupedFileStatus.size() + " groups");
+ // Segregate inputPaths[] to incremental, non incremental and non hoodie
paths
+ List<String> incrementalTables =
HoodieHiveUtil.getIncrementalTableNames(Job.getInstance(job));
+ InputPathHandler inputPathHandler = new
InputPathHandler(getInputPaths(job), incrementalTables);
List<FileStatus> returns = new ArrayList<>();
- for (Map.Entry<HoodieTableMetaClient, List<FileStatus>> entry :
groupedFileStatus.entrySet()) {
- HoodieTableMetaClient metadata = entry.getKey();
- if (metadata == null) {
- // Add all the paths which are not hoodie specific
- returns.addAll(entry.getValue());
+
+ Map<String, HoodieTableMetaClient> tableMetaClientMap =
inputPathHandler.tableMetaClientMap;
+ // process incremental pulls first
+ for (String table : incrementalTables) {
+ HoodieTableMetaClient metaClient = tableMetaClientMap.get(table);
+ if (metaClient == null) {
continue;
}
+ List<Path> inputPaths =
inputPathHandler.groupedIncrementalPaths.get(metaClient);
+ List<FileStatus> result = listStatusForIncrementalMode(job, metaClient,
inputPaths);
+ if (result != null) {
+ returns.addAll(result);
+ }
+ }
- FileStatus[] statuses = entry.getValue().toArray(new
FileStatus[entry.getValue().size()]);
- if (LOG.isDebugEnabled()) {
- LOG.debug("Hoodie Metadata initialized with completed commit Ts as :"
+ metadata);
+ // process non hoodie Paths next.
+ List<Path> nonHoodiePaths = inputPathHandler.nonHoodieInputPaths;
+ if (nonHoodiePaths.size() > 0) {
+ setInputPaths(job, nonHoodiePaths.toArray(new
Path[nonHoodiePaths.size()]));
Review comment:
The input Paths are pre classified as either - incremental, non-incremental
or non-hoodie. The incremental processing takes care of handling incremental
queries. After that the JobConf is mutated to set to non-hoodie paths and
processed. So the case of mixed types should be handled. I have added a unit
test in InputPathHandlerTest to verify this works.
----------------------------------------------------------------
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.
For queries about this service, please contact Infrastructure at:
[email protected]
With regards,
Apache Git Services
