[ 
https://issues.apache.org/jira/browse/BEAM-14515?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17542282#comment-17542282
 ] 

Ning commented on BEAM-14515:
-----------------------------

Our existing logic is coupled with IPython, not Jupyter; most of the usage 
supports any IPython notebook runtime such as Colab.

The missing part: anything "displayed" by IPython can be saved to an HTML file; 
we need a new "save" API for it right before any "display" call. An example we 
have is in the screen diff integration test: 
[https://github.com/apache/beam/blob/master/sdks/python/apache_beam/runners/interactive/testing/integration/notebook_executor.py.]

 

One proposal for 2 changes:

1. Decouple the output and output channel: Output such as 
print/log/plot/HTML/JS should be decoupled from the channel that surface the 
output such as Python output/IPython console/save-to-file/notebook messaging.

2. Decide what features should be supported even if the current environment is 
not in IPython: abstract those.

 

> Decouple interactive Beam from Jupyter
> --------------------------------------
>
>                 Key: BEAM-14515
>                 URL: https://issues.apache.org/jira/browse/BEAM-14515
>             Project: Beam
>          Issue Type: Improvement
>          Components: runner-py-interactive
>            Reporter: Brian Hulette
>            Priority: P2
>
> Some aspects of interactive Beam (e.g. triggering execution with ib.collect) 
> make sense to use outside of a Jupyter notebook. We should make sure that 
> use-cases like this are possible, and we only require a notebook if it's 
> actually necessary. In my mind that should just be features that are 
> _explicitly_ integrating with Jupyter, like side panels, custom magics, etc...
> Things like caching and plotting tools even make sense to use outside of 
> Jupyter (a plotting tool can return a handle that the user could use to save 
> the plot as an image).



--
This message was sent by Atlassian Jira
(v8.20.7#820007)

Reply via email to