rdsr commented on a change in pull request #111: Split files when planning scan
tasks
URL: https://github.com/apache/incubator-iceberg/pull/111#discussion_r260430004
##########
File path: core/src/main/java/com/netflix/iceberg/BaseTableScan.java
##########
@@ -231,4 +234,69 @@ public String toString() {
.add("filter", rowFilter)
.toString();
}
+
+ private CloseableIterable<FileScanTask> planSplits(long splitSize) {
+ final CloseableIterable<FileScanTask> fileScanTasks = planFiles();
+ final Iterable<FileScanTask> splitTasks = FluentIterable
+ .from(fileScanTasks)
+ .transformAndConcat(input -> getSplits(input, splitSize));
+ // Capture manifests which can be closed after scan planning
+ final CloseableIterable<FileScanTask> closeableSplitTasks =
CloseableIterable
+ .combine(splitTasks, ImmutableList.of(fileScanTasks));
+ return closeableSplitTasks;
+ }
+
+ private static Iterable<FileScanTask> getSplits(FileScanTask fileScanTask,
long splitSize) {
+ return () -> new Iterator<FileScanTask>() {
+ long offset = 0;
+ long remainingLen = fileScanTask.length();
+
+ @Override
+ public boolean hasNext() {
+ return remainingLen > 0;
+ }
+
+ @Override
+ public FileScanTask next() {
+ long len = Math.min(splitSize, remainingLen);
+ final FileScanTask splitTask = splitTask(offset, len, fileScanTask);
+ offset += len;
+ remainingLen -= len;
+ return splitTask;
Review comment:
I've commented out specific asserts under `TestFilteredScan` which rely on
the assumption that each task processes a single file. These changes are part
of the PR.
----------------------------------------------------------------
This is an automated message from the Apache Git Service.
To respond to the message, please log on 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
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
