Github user Tagar commented on the issue:
1) My main point was that this exception should be thrown to the user, so
he or she has a chance to increase this limit. Currently if it breaks, only way
to find out about this limitation is to enable debugging and not a lot of users
can do that.
2) You're right .. it's 200M not sure how that user got that much data.
That wasn't from my code, but from a colleague of mine. I guess it was a larger
table of data. Would you mine making default somewhere in the range 16-32M? I
think a lot of folks would run into the 4M limit.
3) Also, it would be great if IPythonInterpreter would catch exceptions
better. Found another problem -
https://issues.apache.org/jira/browse/ZEPPELIN-3239 - unrelated to this one,
but it also shows the same symptoms to the user - Spark interpreter just