openinx commented on a change in pull request #3784:
URL: https://github.com/apache/iceberg/pull/3784#discussion_r819242211
##########
File path: orc/src/main/java/org/apache/iceberg/orc/OrcFileAppender.java
##########
@@ -99,9 +104,27 @@ public Metrics metrics() {
@Override
public long length() {
- Preconditions.checkState(isClosed,
- "Cannot return length while appending to an open file.");
- return file.toInputFile().getLength();
+ if (isClosed) {
+ return file.toInputFile().getLength();
+ }
+ if (this.treeWriter == null) {
+ throw new RuntimeException("Can't get the length!");
+ }
+ long estimateMemory = this.treeWriter.estimateMemory();
+
+ long dataLength = 0;
+ try {
+ List<StripeInformation> stripes = writer.getStripes();
+ if (!stripes.isEmpty()) {
+ StripeInformation stripeInformation = stripes.get(stripes.size() - 1);
+ dataLength = stripeInformation != null ? stripeInformation.getOffset()
+ stripeInformation.getLength() : 0;
+ }
+ } catch (IOException e) {
+ throw new UncheckedIOException(String.format("Can't get stripes from
file %s", file.location()), e);
+ }
+
+ // This value is estimated, not actual.
+ return dataLength + estimateMemory + batch.size;
Review comment:
I read this
https://github.com/apache/iceberg/pull/3784#issuecomment-1022891787 here. For
the file-persisted bytes , I think using the last strip's offset plus the its
length should be correct. For he the memory encoded batch vector , I think the
`TreeWriter#estimateMemory` should be okay.
But for the batch vector whose rows did not flush to encoded memory, using
the batch.size shouldn't be correct. Because the rows can be any data type,
such as Integer, Long, Timestamp, String etc. As their width are not the same,
I think we may need to use an average width minus the batch.size (which is row
count actually).
--
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]
