[
https://issues.apache.org/jira/browse/FLINK-19006?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17185727#comment-17185727
]
Aljoscha Krettek commented on FLINK-19006:
------------------------------------------
This code is more than 6 years old by now. 😅 We can fix it, but my general
opinion is that we should remove these database-like or relational methods from
the DataStream API. See also this thread:
https://lists.apache.org/thread.html/r16bc0d1755a72e29087a97f11d26a41c35f873ef29b8fd5e0357e286%40%3Cdev.flink.apache.org%3E
> project transformation does not support conversion to Tuple25 type
> ------------------------------------------------------------------
>
> Key: FLINK-19006
> URL: https://issues.apache.org/jira/browse/FLINK-19006
> Project: Flink
> Issue Type: Improvement
> Components: API / DataStream
> Affects Versions: 1.11.1
> Reporter: ming li
> Priority: Major
>
> In the {{DataStream#project}} method, it will judge whether the length of
> {{fieldIndexes}} is between 1 and {{Tuple.MAX_ARITY-1}}, and then call
> {{projectTupleXX}} according to the length of {{fieldIndexes}}. This limits
> the maximum length of {{Tuple}} to 24.
> {code:java}
> protected StreamProjection(DataStream<IN> dataStream, int[] fieldIndexes) {
> if (!dataStream.getType().isTupleType()) {
> throw new RuntimeException("Only Tuple DataStreams can be projected");
> }
> if (fieldIndexes.length == 0) {
> throw new IllegalArgumentException("project() needs to select at least
> one (1) field.");
> } else if (fieldIndexes.length > Tuple.MAX_ARITY - 1) {
> throw new IllegalArgumentException(
> "project() may select only up to (" + (Tuple.MAX_ARITY - 1) + ")
> fields.");
> }
> int maxFieldIndex = (dataStream.getType()).getArity();
> for (int i = 0; i < fieldIndexes.length; i++) {
> Preconditions.checkElementIndex(fieldIndexes[i], maxFieldIndex);
> }
> this.dataStream = dataStream;
> this.fieldIndexes = fieldIndexes;
> }{code}
> This problem also appears in {{ProjectOperator}}.
> {code:java}
> public Projection(DataSet<T> ds, int[] fieldIndexes) {
> if (!(ds.getType() instanceof TupleTypeInfo)) {
> throw new UnsupportedOperationException("project() can only be applied
> to DataSets of Tuples.");
> }
> if (fieldIndexes.length == 0) {
> throw new IllegalArgumentException("project() needs to select at least
> one (1) field.");
> } else if (fieldIndexes.length > Tuple.MAX_ARITY - 1) {
> throw new IllegalArgumentException(
> "project() may select only up to (" + (Tuple.MAX_ARITY - 1) + ")
> fields.");
> }
> int maxFieldIndex = ds.getType().getArity();
> for (int fieldIndexe : fieldIndexes) {
> Preconditions.checkElementIndex(fieldIndexe, maxFieldIndex);
> }
> this.ds = ds;
> this.fieldIndexes = fieldIndexes;
> }{code}
> I think the length we limit should be 1 to {{Tuple.MAX_ARITY}} instead of 1
> to {{Tuple.MAX_ARITY-1}}.
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
