[ZEPPELIN-1150] Table contents disappear 2nd run onwards

[Leemoonsoo]

What is this PR for?

Fix for ZEPPELIN-1150.

What type of PR is it?

Bug Fix

Todos

• - Recreate table on data refresh
• - Better solution for ZEPPELIN-1078 without performance degrade

What is the Jira issue?

ZEPPELIN-1150

How should this be tested?

Reproduce procedure described in the issue

Questions:

• Does the licenses files need update? no
• Is there breaking changes for older versions? no
• Does this needs documentation? no

[astroshim]

I got a following build error.

[INFO] --- frontend-maven-plugin:0.0.25:grunt (grunt build) @ zeppelin-web ---
[INFO] Running 'grunt build --no-color' in /home/nflabs/tmp/zeppelin/zeppelin-web
[INFO] Running "jscs:all" (jscs) task
[INFO] disallowMultipleLineBreaks: ; and console.log("updateParagraph oldData %o, newData %o. type %o -> %o, mode %o -> %o", $scope.paragraph, data, oldType, newType, oldGraphMode, newGraphMode); should have at most 2 line(s) between them at src/app/notebook/paragraph/paragraph.controller.js :
[INFO]    412 |        isEmpty(data.paragraph.result) !== isEmpty($scope.paragraph.result) ||
[INFO]    413 |        data.paragraph.status === 'ERROR' || (data.paragraph.status === 'FINISHED' && statusChanged) ||
[INFO]    414 |        (!newActiveApp && oldActiveApp !== newActiveApp);
[INFO] -----------------------------------------------------------------^
[INFO]    415 |
[INFO]    416 |
[INFO] >> 1 code style errors found!

[Leemoonsoo]

@astroshim Thanks for testing. Build error is addressed in last commit.

[corneadoug]

@Leemoonsoo Tested,
Fix for the table after re-run confirmed.

For the second item, I would keep the solution used in https://issues.apache.org/jira/browse/ZEPPELIN-1078
Since your changes do not bring any improvement regarding performance degrade.
(We still do a renderText, when the text result is the same as before)

[astroshim]

Build error has gone and working good!

[Leemoonsoo]

@corneadoug Try revert line 413 and 473. And then run paragraph multiple times (with table result). In my machine after 2~3 times without 413, 473 change, browser almost freeze.

[Leemoonsoo]

@corneadoug I tried more and it looks like slowing down without 413 and 473 change is not consistently reproducible.

Anyway, logically, 413 and 473 change reduce text rendering from
4 times (rendering 4 status change FINISHED -> PENDING -> RUNNING -> FINISHED)
to
1 time (RUNNING -> FINISHED)

So, i'd like to keep it while it doesn't bring any side effect or increase complexity of code.

[corneadoug]

@Leemoonsoo I tested again, originally I was testing with previous || resultRefreshed
473 does help the text rendering indeed. However I have the same improvement without 413 for both TABLE and TEXT types

[Leemoonsoo]

@corneadoug 413 and 473 "together" are alternative fix for ZEPPELIN-1078 for better performance.

Without 413, applying 473 will bring ZEPPELIN-1078 problem again.

[corneadoug]

@Leemoonsoo I was testing with %spark only, while ZEPPELIN-1078's issue shows with %python.
LGTM

[Leemoonsoo]

Thanks @astroshim @corneadoug for review this!

Merge it into master if there're no more discussions.

[prabhjyotsingh]

I missed this (which solve the same problem as #1182), let me check this.

[prabhjyotsingh]

LGTM 👍