laskoviymishka commented on code in PR #16828:
URL: https://github.com/apache/iceberg/pull/16828#discussion_r3461140592
##########
kafka-connect/kafka-connect/src/test/java/org/apache/iceberg/connect/data/TestSchemaUtils.java:
##########
@@ -331,4 +332,20 @@ public void testInferIcebergTypeEmpty() {
assertThat(SchemaUtils.inferIcebergType(ImmutableMap.of("nested",
ImmutableMap.of()), config))
.isNull();
}
+
+ @Test
+ public void testToIcebergTypeUUIDLogicalTypeOnString() {
+ IcebergSinkConfig config = mock(IcebergSinkConfig.class);
+
+ Schema uuidSchema = SchemaBuilder.string().name("uuid").build();
+ assertThat(SchemaUtils.toIcebergType(uuidSchema,
config)).isInstanceOf(UUIDType.class);
+ }
+
+ @Test
+ public void testToIcebergTypeUUIDLogicalTypeOnBytes() {
+ IcebergSinkConfig config = mock(IcebergSinkConfig.class);
Review Comment:
This test passes, but it's why the BYTES bug above stays invisible — it only
checks the type-inference result, never the conversion. If we keep the
BYTES→UUIDType mapping, I'd want a companion test that builds a Connect Struct
with a BYTES "uuid" field, runs `RecordConverter.convert`, and asserts the
resulting record holds the right UUID. That's the test that fails today.
Same gap exists for the STRING path — `TestRecordConverter` only uses
`Schema.STRING_SCHEMA` (no logical name), so the new STRING+uuid path is never
round-tripped either. One end-to-end test per kept branch would close this.
##########
kafka-connect/kafka-connect/src/main/java/org/apache/iceberg/connect/data/SchemaUtils.java:
##########
@@ -235,6 +236,8 @@ Type toIcebergType(Schema valueSchema) {
if (Decimal.LOGICAL_NAME.equals(valueSchema.name())) {
int scale =
Integer.parseInt(valueSchema.parameters().get(Decimal.SCALE_FIELD));
return DecimalType.of(38, scale);
+ } else if ("uuid".equals(valueSchema.name())) {
Review Comment:
I think this BYTES branch is broken at write time. We create a UUIDType
column here, but `RecordConverter.convertUUID` (RecordConverter.java:413-431)
only handles `String` and `java.util.UUID` — a Connect BYTES field comes
through as a `ByteBuffer`, so every record throws `Cannot convert to UUID`.
That's worse than the pre-fix behavior, where the data at least landed as
binary.
Either we extend `convertUUID` to accept a 16-byte big-endian
`ByteBuffer`/`byte[]` (`org.apache.iceberg.util.UUIDUtil.convert` handles
that), or we drop this branch and leave BYTES+uuid as BinaryType with a
comment. The STRING branch is the one that maps to the real-world source, so
I'd lean toward dropping BYTES unless we add the converter support and a
round-trip test. 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: [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]
