morningman commented on code in PR #45148:
URL: https://github.com/apache/doris/pull/45148#discussion_r1874352612
##########
fe/fe-core/src/main/java/org/apache/doris/datasource/SplitAssignment.java:
##########
@@ -156,7 +156,7 @@ public BlockingQueue<Collection<TScanRangeLocations>>
getAssignedSplits(Backend
return splits;
}
- public void setException(UserException e) {
+ public synchronized void setException(UserException e) {
Review Comment:
Why using `synchronized `? add comment in code
##########
fe/fe-core/src/main/java/org/apache/doris/datasource/maxcompute/source/MaxComputeScanNode.java:
##########
@@ -181,32 +174,118 @@ boolean createTableBatchReadSession() throws
UserException {
orderedRequiredDataColumns.add(columnName);
}
}
+ }
+ /**
+ * For no partition table: request requiredPartitionSpecs is empty
+ * For partition table: if requiredPartitionSpecs is empty, get all
partition data.
+ */
+ TableBatchReadSession createTableBatchReadSession(List<PartitionSpec>
requiredPartitionSpecs) throws IOException {
+ MaxComputeExternalCatalog mcCatalog = (MaxComputeExternalCatalog)
table.getCatalog();
+ readTimeout = mcCatalog.getReadTimeout();
+ connectTimeout = mcCatalog.getConnectTimeout();
+ retryTimes = mcCatalog.getRetryTimes();
+
+ TableReadSessionBuilder scanBuilder = new TableReadSessionBuilder();
+ return scanBuilder.identifier(TableIdentifier.of(table.getDbName(),
table.getName()))
+ .withSettings(mcCatalog.getSettings())
+ .withSplitOptions(mcCatalog.getSplitOption())
+ .requiredPartitionColumns(requiredPartitionColumns)
+ .requiredDataColumns(orderedRequiredDataColumns)
+ .withFilterPredicate(filterPredicate)
+ .requiredPartitions(requiredPartitionSpecs)
+ .withArrowOptions(
+ ArrowOptions.newBuilder()
+ .withDatetimeUnit(TimestampUnit.MILLI)
+ .withTimestampUnit(TimestampUnit.NANO)
+ .build()
+ ).buildBatchReadSession();
+ }
- MaxComputeExternalCatalog mcCatalog = (MaxComputeExternalCatalog)
table.getCatalog();
+ @Override
+ public boolean isBatchMode() {
+ if (table.getPartitionColumns().isEmpty()) {
+ return false;
+ }
- try {
- TableReadSessionBuilder scanBuilder = new
TableReadSessionBuilder();
- tableBatchReadSession =
-
scanBuilder.identifier(TableIdentifier.of(table.getDbName(), table.getName()))
- .withSettings(mcCatalog.getSettings())
- .withSplitOptions(mcCatalog.getSplitOption())
- .requiredPartitionColumns(requiredPartitionColumns)
- .requiredPartitions(requiredPartitionSpecs)
- .requiredDataColumns(orderedRequiredDataColumns)
- .withArrowOptions(
- ArrowOptions.newBuilder()
-
.withDatetimeUnit(TimestampUnit.MILLI)
-
.withTimestampUnit(TimestampUnit.NANO)
- .build()
- )
- .withFilterPredicate(filterPredicate)
- .buildBatchReadSession();
- } catch (java.io.IOException e) {
- throw new RuntimeException(e);
+ com.aliyun.odps.Table odpsTable = table.getOdpsTable();
+ if (desc.getSlots().isEmpty() || odpsTable.getFileNum() <= 0) {
+ return false;
}
- return true;
+
+ int numPartitions =
ConnectContext.get().getSessionVariable().getNumPartitionsInBatchMode();
+ return numPartitions > 0
+ && selectedPartitions != SelectedPartitions.NOT_PRUNED
+ && selectedPartitions.selectedPartitions.size() >=
numPartitions;
+ }
+
+ @Override
+ public int numApproximateSplits() {
+ return selectedPartitions.selectedPartitions.size();
+ }
+
+ @Override
+ public void startSplit() {
+ this.totalPartitionNum = selectedPartitions.totalPartitionNum;
+ this.selectedPartitionNum =
selectedPartitions.selectedPartitions.size();
+
+ if (selectedPartitions.selectedPartitions.isEmpty()) {
+ //no need read any partition data.
+ return;
+ }
+
+ createRequiredColumns();
+ List<PartitionSpec> requiredPartitionSpecs = new ArrayList<>();
+ selectedPartitions.selectedPartitions.forEach(
+ (key, value) -> requiredPartitionSpecs.add(new
PartitionSpec(key))
+ );
+
+
+ int batchNumPartitions =
ConnectContext.get().getSessionVariable().getNumPartitionsInBatchMode();
+
+ Executor scheduleExecutor =
Env.getCurrentEnv().getExtMetaCacheMgr().getScheduleExecutor();
+ AtomicReference<UserException> batchException = new
AtomicReference<>(null);
+ AtomicInteger numFinishedPartitions = new AtomicInteger(0);
+
+ CompletableFuture.runAsync(() -> {
+ for (int beginIndex = 0; beginIndex <
requiredPartitionSpecs.size(); beginIndex += batchNumPartitions) {
+ int endIndex = Math.min(beginIndex + batchNumPartitions,
requiredPartitionSpecs.size());
+ if (batchException.get() != null || splitAssignment.isStop()) {
+ break;
+ }
+ List<PartitionSpec> requiredBatchPartitionSpecs =
requiredPartitionSpecs.subList(beginIndex, endIndex);
+ int curBatchSize = endIndex - beginIndex;
+
+ try {
+ CompletableFuture.runAsync(() -> {
+ try {
+ TableBatchReadSession tableBatchReadSession =
+
createTableBatchReadSession(requiredBatchPartitionSpecs);
Review Comment:
Looks like in this new implementation, you create read session for each
partition?
Is `create read session` a heavy operation?
--
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.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]