Hey all,
there was a project course at Technical University here in Berlin, where the users had to implement various ML algorithms in Flink. The following is a list of what the users provided as feedback:
1) Collection fields in tuples (
https://issues.apache.org/jira/browse/FLINK-610)
Two groups ran into this problem and worked around it with a manual String encoding. :(
2) Data sinks in iterations (there is an issue for this, but I couldn't find it at the moment)
3) First(n) operator (
https://issues.apache.org/jira/browse/FLINK-970)
4) For loop with incremental rollout (as in Spark) is more intuitive that our iterate operators
---
I think [~aalexandrov] also took some notes.
Best,
Ufuk
Locked
