Support figures that are larger than Chromium's maximum message size

[bhachauk]

Hi,

I am trying to export the fig using the method

fig.to_image(format="png", engine="kaleido")

I got the exception like:

Transform failed. Error stream:

[0723/161014.534602:FATAL:ipc_mojo_bootstrap.cc(729)] Check failed: message->data_num_bytes() <= Channel::kMaximumMessageSize  (366505976 vs. 134217728)

How to handle this check ?

[nicolaskruchten]

@jonmmease transferred from Plotly.py

[jonmmease]

Hi @Bhanuchander210, this message indicates that the size of the figure JSON bundle exceeded the maximum chromium message size. This is generally a pretty large threshold (e.g. somewhere around an 8 million point scatter plot).

Could you describe the figure you are trying to export? Are you able to display it interactively? Did you ever try exporting it using the Orca? Thanks!

[bhachauk]

Hi @jonmmease, As you mentioned, yes it is pretty large scatter plot approx (6 columns * 1696900 rows : may have nan, so number of points will vary).

Are you able to display it interactively?

• Yes i can display the plot interactively

Did you ever try exporting it using the Orca?

• Yes i have tried that also, but seems not working...

It shows the following exception :

For some reason plotly.py was unable to communicate with the
local orca server process, even though the server process seems to be running.

Please review the process and connection information below:

orca status
-----------
    state: running
    executable: /home/bhanuchander/miniconda3/bin/orca
    version: 1.2.1
    port: 46368
    pid: 11821
    command: ['/home/bhanuchander/miniconda3/bin/orca', 'serve', '-p', '46368', '--plotly', '/home/bhanuchander/miniconda3/lib/python3.7/site-packages/plotly/package_data/plotly.min.js', '--graph-only', '--mathjax', 'https://cdnjs.cloudflare.com/ajax/libs/mathjax/2.7.5/MathJax.js']

[jonmmease]

Thanks for the details. As it does look like a figure size issue, it's not surprising that orca fails as well since Electron is using the same Chromium engine underneath.

It should be possible to work around this limitation, but it will take some redesign work to support streaming the figure specification from the C++ to the JavaScript side.

Right now, we parse the input JSON in C++ and then let chromium transfer the resulting dictionary from C++ to JavaScript. We may need to transfer the figure specification in string form in chunks and then parse it into a JavaScript object at the end.

[jayceslesar]

I also get this issue with large datasets

[rskabelund]

I am also getting this error on large plots.

figure.to_image( 
File "/opt/conda/lib/python3.8/site-packages/plotly/basedatatypes.py", line 3495, in to_image return pio.to_image(self, *args, **kwargs) 
File "/opt/conda/lib/python3.8/site-packages/plotly/io/_kaleido.py", line 131, in to_image img_bytes = scope.transform( 
File "/opt/conda/lib/python3.8/site-packages/kaleido/scopes/plotly.py", line 103, in transform 
response = self._perform_transform( File "/opt/conda/lib/python3.8/site-packages/kaleido/scopes/base.py", line 305, in _perform_transform raise ValueError(message) 
ValueError: Transform failed. 

Error stream: [0513/215523.654363:FATAL:ipc_mojo_bootstrap.cc(729)] 
Check failed: message->data_num_bytes() <= Channel::kMaximumMessageSize (400041304 vs. 134217728) #0 0x55f02719ef89 

Has there been any traction in chunking messages from C++ to JavaScript? Been searching all over for a workaround, but that might be the only solution for me.

Are there any alternatives to get base64 data from a plotly plot for large datasets? Is there a way to chunk it on the python side? I tried pushing back and produce aggregate data for these plots, but they need to have all the data and I am stuck as we need to save to disk.

Any suggestions would be greatly appreciated.

[JesseFarebro]

Also experiencing this issue. I have a large figure with ~32 subplots, hopefully, there's a solution on the horizon.

[artizzq]

Same problem, please help

[Rob-Hulley]

I just updated kaleido.py to v 0.2.1 and it has resolved the issue for me.