Guosmilesmile commented on code in PR #15633:
URL: https://github.com/apache/iceberg/pull/15633#discussion_r2939189724
##########
data/src/test/java/org/apache/iceberg/data/BaseFormatModelTests.java:
##########
@@ -317,6 +314,306 @@ void
testPositionDeleteWriterEngineWriteGenericRead(FileFormat fileFormat) throw
DataTestHelpers.assertEquals(positionDeleteSchema.asStruct(), records,
readRecords);
}
+ @ParameterizedTest
+ @FieldSource("FORMAT_AND_GENERATOR")
+ /** Write with Generic Record, read with projected engine type T (narrow
schema) */
+ void testReaderBuilderProjection(FileFormat fileFormat, DataGenerator
dataGenerator)
+ throws IOException {
+ Schema fullSchema = dataGenerator.schema();
+
+ List<Types.NestedField> columns = fullSchema.columns();
+ Schema projectedSchema = new Schema(columns.get(columns.size() - 1));
+
+ List<Record> genericRecords = dataGenerator.generateRecords();
+ writeGenericRecords(fileFormat, fullSchema, genericRecords);
+
+ List<Record> projectedGenericRecords = projectRecords(genericRecords,
projectedSchema);
+ List<T> expectedEngineRecords =
+ convertToEngineRecords(projectedGenericRecords, projectedSchema);
+
+ InputFile inputFile = encryptedFile.encryptingOutputFile().toInputFile();
+ List<T> readRecords;
+ try (CloseableIterable<T> reader =
+ FormatModelRegistry.readBuilder(fileFormat, engineType(), inputFile)
+ .project(projectedSchema)
+ .engineProjection(engineSchema(projectedSchema))
+ .build()) {
+ readRecords = ImmutableList.copyOf(reader);
+ }
+
+ assertEquals(projectedSchema, expectedEngineRecords, readRecords);
+ }
+
+ @ParameterizedTest
+ @FieldSource("FORMAT_AND_GENERATOR")
+ void testReaderBuilderFilter(FileFormat fileFormat, DataGenerator
dataGenerator)
+ throws IOException {
+
+ // Avro does not support filter push down
+ // Skip this test for Avro to avoid false failures.
+ assumeThat(fileFormat != FileFormat.AVRO).isTrue();
+
+ Schema schema = dataGenerator.schema();
+
+ List<Record> genericRecords = dataGenerator.generateRecords();
+ writeGenericRecords(fileFormat, schema, genericRecords);
+
+ // Construct a filter condition that is smaller than the minimum value to
achieve file-level
+ // filtering.
+ Types.NestedField firstField = schema.columns().get(0);
+ Expression filter = filterFieldExpression(firstField, schema,
genericRecords);
+
+ InputFile inputFile = encryptedFile.encryptingOutputFile().toInputFile();
+ List<T> readRecords;
+ try (CloseableIterable<T> reader =
+ FormatModelRegistry.readBuilder(fileFormat, engineType(), inputFile)
+ .project(schema)
+ .engineProjection(engineSchema(schema))
+ .filter(filter)
+ .build()) {
+ readRecords = ImmutableList.copyOf(reader);
+ }
+
+ assertThat(readRecords).isEmpty();
+ }
+
+ @ParameterizedTest
+ @FieldSource("FORMAT_AND_GENERATOR")
+ /**
+ * Write with Generic Record, then read using an upper-cased column name in
the filter to verify
+ * caseSensitive behavior.
+ */
+ void testReaderBuilderCaseSensitive(FileFormat fileFormat, DataGenerator
dataGenerator)
+ throws IOException {
+
+ // Avro does not support filter push down; caseSensitive has no effect on
it.
+ // Skip this test for Avro to avoid false failures.
+ assumeThat(fileFormat != FileFormat.AVRO).isTrue();
+
+ Schema schema = dataGenerator.schema();
+
+ List<Record> genericRecords = dataGenerator.generateRecords();
+ writeGenericRecords(fileFormat, schema, genericRecords);
+
+ // Build a filter using the upper-cased name of the first column.
+ Types.NestedField firstField = schema.columns().get(0);
+ Object filterValue = genericRecords.get(0).getField(firstField.name());
+ Expression upperCaseFilter =
Expressions.equal(firstField.name().toUpperCase(), filterValue);
+
+ InputFile inputFile = encryptedFile.encryptingOutputFile().toInputFile();
+
+ // caseSensitive=false: upper-cased column name must be resolved correctly.
+ List<T> readRecords;
+ try (CloseableIterable<T> reader =
+ FormatModelRegistry.readBuilder(fileFormat, engineType(), inputFile)
+ .project(schema)
+ .engineProjection(engineSchema(schema))
+ .filter(upperCaseFilter)
+ .caseSensitive(false)
+ .build()) {
+ readRecords = ImmutableList.copyOf(reader);
+ }
+
+ assertThat(readRecords).isNotEmpty();
+
+ // caseSensitive=true: upper-cased column name cannot be resolved → must
throw.
+ assertThatThrownBy(
+ () -> {
+ try (CloseableIterable<T> reader =
+ FormatModelRegistry.readBuilder(fileFormat, engineType(),
inputFile)
+ .project(schema)
+ .engineProjection(engineSchema(schema))
+ .filter(upperCaseFilter)
+ .caseSensitive(true)
+ .build()) {
+ ImmutableList.copyOf(reader);
+ }
+ })
+ .isInstanceOf(ValidationException.class)
+ .hasMessageContaining("Cannot find field '%s'",
firstField.name().toUpperCase());
+ }
+
+ @ParameterizedTest
+ @FieldSource("FORMAT_AND_GENERATOR")
+ /**
+ * Write with Generic Record, then read using split to verify that the split
range is respected.
+ * Reading with a zero-length split at the end of the file should return no
records, while reading
+ * with the full file range should return all records.
Review Comment:
The way to trigger multiple groups within a single DataFile varies across
different file formats — such as row groups in Parquet, stripes in ORC, etc.
Would we trigger this by writing 1000 or more records? Though explicitly
setting this size feels a bit awkward.
--
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]
