twalthr commented on code in PR #258:
URL:
https://github.com/apache/flink-connector-kafka/pull/258#discussion_r3326680408
##########
flink-connector-kafka/src/main/java/org/apache/flink/streaming/connectors/kafka/table/KafkaDynamicSink.java:
##########
@@ -457,6 +465,47 @@ public Object read(RowData row, int pos) {
}
}),
+ /**
+ * Wire-faithful alternative to {@code headers}: preserves duplicate
keys and insertion
+ * order using {@code ARRAY<ROW<key STRING, value BYTES>>} instead of
a lossy MAP.
+ */
+ MULTI_HEADERS(
+ "multi-headers",
+ DataTypes.ARRAY(
+ DataTypes.ROW(
+ DataTypes.FIELD(
+ "key",
DataTypes.STRING().nullable()),
+ DataTypes.FIELD(
+ "value",
DataTypes.BYTES().nullable()))
+ .notNull())
+ .nullable(),
+ new MetadataConverter() {
+ private static final long serialVersionUID = 1L;
+
+ @Override
+ public Object read(RowData row, int pos) {
+ if (row.isNullAt(pos)) {
+ return null;
+ }
+ final ArrayData arrayData = row.getArray(pos);
+ final List<Header> headers = new ArrayList<>();
+ for (int i = 0; i < arrayData.size(); i++) {
+ if (!arrayData.isNullAt(i)) {
+ final RowData headerRow = arrayData.getRow(i,
2);
+ if (headerRow.isNullAt(0)) {
+ throw new IllegalArgumentException(
Review Comment:
The reason for this is implicit casting. In SQL all columns are nullable by
default. If we make headers not null, the implicit casting might have issues
because casting from nullable to not nullable is not lossless. People would
need to add IFNULL function calls.
--
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]
