Github user fhueske commented on a diff in the pull request:
https://github.com/apache/flink/pull/3511#discussion_r139917275
--- Diff:
flink-core/src/main/java/org/apache/flink/api/common/ExecutionConfig.java ---
@@ -127,6 +127,12 @@
private long autoWatermarkInterval = 0;
/**
+ * The flag determines whether a custom NormalizedKeySorter will be
dynamically created
+ * for underlying data
+ */
+ private boolean codeGenerationForSorterEnabled = false;
--- End diff --
Add documentation for this flag to
`https://ci.apache.org/projects/flink/flink-docs-release-1.3/dev/execution_configuration.html`.
Since the flag will be hidden in a larger list of options, it might make
sense to mention it in the DataSet API documentation more prominently. Maybe we
can restructure the docs a bit and add a section about performance tweaking
that would include semantic annotations, optimizer hints (incl. hash combine),
sorter code gen, and object reusage. This could be done as a follow up issue,
IMO.
---
