pvary commented on code in PR #12824:
URL: https://github.com/apache/iceberg/pull/12824#discussion_r2080967454
##########
core/src/main/java/org/apache/iceberg/actions/BinPackRewriteFilePlanner.java:
##########
@@ -199,30 +214,48 @@ protected long defaultTargetFileSize() {
public FileRewritePlan<FileGroupInfo, FileScanTask, DataFile,
RewriteFileGroup> plan() {
StructLikeMap<List<List<FileScanTask>>> plan = planFileGroups();
RewriteExecutionContext ctx = new RewriteExecutionContext();
- Stream<RewriteFileGroup> groups =
- plan.entrySet().stream()
- .filter(e -> !e.getValue().isEmpty())
- .flatMap(
- e -> {
- StructLike partition = e.getKey();
- List<List<FileScanTask>> scanGroups = e.getValue();
- return scanGroups.stream()
- .map(
- tasks -> {
- long inputSize = inputSize(tasks);
- return newRewriteGroup(
- ctx,
- partition,
- tasks,
- inputSplitSize(inputSize),
- expectedOutputFiles(inputSize));
- });
- })
- .sorted(RewriteFileGroup.comparator(rewriteJobOrder));
+ List<RewriteFileGroup> selectedFileGroups = new ArrayList<>();
+ AtomicInteger fileCountRunner = new AtomicInteger();
+ plan.entrySet().stream()
Review Comment:
What is the biggest memory hug at this point? I expect that the
`FileScanTask`s with the many `DataFile`s. If that's true, then filtering at
this point doesn't save too much memory. We already have a `plan` which
contains a reference for them.
Suggestions:
- Accept that we don't solve the memory issue with this solution, and
concentrate on decreasing the NN calls further down the execution path. In this
case filtering after sorting aligns more with the user expectations, or
- Push down the filter to `planFileGroups`, or
- Make `planFileGroups` return a real iterator (seems very hard).
WDYT?
--
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: issues-unsubscr...@iceberg.apache.org
For queries about this service, please contact Infrastructure at:
us...@infra.apache.org
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@iceberg.apache.org
For additional commands, e-mail: issues-h...@iceberg.apache.org
