RussellSpitzer commented on code in PR #12840:
URL: https://github.com/apache/iceberg/pull/12840#discussion_r2054896054
##########
spark/v3.5/spark/src/test/java/org/apache/iceberg/spark/actions/TestRewriteManifestsAction.java:
##########
@@ -531,6 +536,158 @@ public void testRewriteLargeManifestsPartitionedTable()
throws IOException {
assertThat(newManifests).hasSizeGreaterThanOrEqualTo(2);
}
+ @TestTemplate
+ public void
testRewriteManifestsPartitionedTableWithInvalidClusteringColumns()
+ throws IOException {
+ PartitionSpec spec =
PartitionSpec.builderFor(SCHEMA).identity("c1").bucket("c3", 10).build();
+ Map<String, String> options = Maps.newHashMap();
+ options.put(TableProperties.FORMAT_VERSION, String.valueOf(formatVersion));
+ options.put(TableProperties.SNAPSHOT_ID_INHERITANCE_ENABLED,
snapshotIdInheritanceEnabled);
+ Table table = TABLES.create(SCHEMA, spec, options, tableLocation);
+
+ SparkActions actions = org.apache.iceberg.spark.actions.SparkActions.get();
+
+ List<String> hasNonexistentFields = ImmutableList.of("c1", "c2");
+ assertThatThrownBy(
+ () ->
+ actions
+ .rewriteManifests(table)
+ .rewriteIf(manifest -> true)
+ .clusterBy(hasNonexistentFields)
+ .execute())
+ .isInstanceOf(IllegalArgumentException.class)
+ .hasMessage(
+ "Cannot set manifest clustering because specified field(s) [c2]
were not found in "
+ + "current partition spec 0.");
+
+ // c3_bucket is the correct internal partition name to use, c3 is the
untransformed column name,
+ // clusterBy() expects the hidden partition column names
+ List<String> hasIncorrectPartitionFieldNames = ImmutableList.of("c1",
"c3");
+ assertThatThrownBy(
+ () ->
+ actions
+ .rewriteManifests(table)
+ .rewriteIf(manifest -> true)
+ .clusterBy(hasIncorrectPartitionFieldNames)
+ .execute())
+ .isInstanceOf(IllegalArgumentException.class)
+ .hasMessage(
+ "Cannot set manifest clustering because specified field(s) [c3]
were not found in "
+ + "current partition spec 0.");
+ }
+
+ @TestTemplate
+ public void testRewriteManifestsPartitionedTableWithCustomClustering()
throws IOException {
+ Random random = new Random();
+
+ PartitionSpec spec =
+ PartitionSpec.builderFor(SCHEMA).identity("c1").truncate("c2",
3).bucket("c3", 10).build();
+ Table table = TABLES.create(SCHEMA, spec, tableLocation);
+
+ List<DataFile> dataFiles = Lists.newArrayList();
+ for (int fileOrdinal = 0; fileOrdinal < 1000; fileOrdinal++) {
+ dataFiles.add(
+ newDataFile(
+ table,
+ TestHelpers.Row.of(
+ new Object[] {
+ fileOrdinal, String.valueOf(random.nextInt() * 100),
random.nextInt(10)
+ })));
+ }
+ ManifestFile appendManifest = writeManifest(table, dataFiles);
+ table.newFastAppend().appendManifest(appendManifest).commit();
+
+ List<ManifestFile> manifests =
table.currentSnapshot().allManifests(table.io());
+ assertThat(manifests).as("Should have 1 manifests before
rewrite").hasSize(1);
+
+ // Capture the c3 partition's lower and upper bounds - used for later test
assertions
Review Comment:
The checks here are pretty complicated, I think it probably would be a lot
simpler to just check that the rows within the manifest are ordered rather than
going through all the manifest metadata. Our readers should already read
through entries in order from the file I think ...
Then we could just check "contents before = contents after" - "row before
unsorted on c3 != rows after sorted on c3 transform"
--
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]
