rdblue commented on a change in pull request #3400:
URL: https://github.com/apache/iceberg/pull/3400#discussion_r740529694
##########
File path:
spark/v3.2/spark/src/main/java/org/apache/iceberg/spark/source/SparkBatchQueryScan.java
##########
@@ -104,30 +144,100 @@
}
}
- if (splitSize != null) {
- scan = scan.option(TableProperties.SPLIT_SIZE, splitSize.toString());
+ for (Expression filter : filterExpressions()) {
+ scan = scan.filter(filter);
}
- if (splitLookback != null) {
- scan = scan.option(TableProperties.SPLIT_LOOKBACK,
splitLookback.toString());
+ try (CloseableIterable<FileScanTask> filesIterable = scan.planFiles()) {
+ this.files = Lists.newArrayList(filesIterable);
+ } catch (IOException e) {
+ throw new UncheckedIOException("Failed to close table scan: " + scan,
e);
}
+ }
- if (splitOpenFileCost != null) {
- scan = scan.option(TableProperties.SPLIT_OPEN_FILE_COST,
splitOpenFileCost.toString());
- }
+ return files;
+ }
- for (Expression filter : filterExpressions()) {
- scan = scan.filter(filter);
+ @Override
+ protected List<CombinedScanTask> tasks() {
+ if (tasks == null) {
+ CloseableIterable<FileScanTask> splitFiles = TableScanUtil.splitFiles(
+ CloseableIterable.withNoopClose(files()),
+ splitSize);
+ CloseableIterable<CombinedScanTask> scanTasks = TableScanUtil.planTasks(
+ splitFiles, splitSize,
+ splitLookback, splitOpenFileCost);
+ tasks = Lists.newArrayList(scanTasks);
+ }
+
+ return tasks;
+ }
+
+ @Override
+ public NamedReference[] filterAttributes() {
+ Set<Integer> partitionFieldSourceIds = Sets.newHashSet();
+
+ for (Integer specId : specIds()) {
+ PartitionSpec spec = table().specs().get(specId);
+ for (PartitionField field : spec.fields()) {
+ partitionFieldSourceIds.add(field.sourceId());
}
+ }
- try (CloseableIterable<CombinedScanTask> tasksIterable =
scan.planTasks()) {
- this.tasks = Lists.newArrayList(tasksIterable);
- } catch (IOException e) {
- throw new RuntimeIOException(e, "Failed to close table scan: %s",
scan);
+ Map<Integer, String> nameById =
TypeUtil.indexQuotedNameById(table().schema().asStruct());
Review comment:
I don't think there's a problem here. Iceberg is going to use a
consistent set of names for a table that is determined at table load time. And
Iceberg metadata is tracked using field IDs, so even if a name changes it's
okay.
If you have a table partitioned by 1:a and change that to b in the schema,
then you have a table partitioned by 1:b. Because only one schema is reported
to Spark, you'll either use a or b and when Spark passes the filter back you'll
consistently get a or b in the predicate.
--
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]
