1

I try to plot a large dataseries with holoviews, bokey and datashader. The plot shows without any issues, but with zooming in, the plot doesn't change the resolution, so the dataseries is very pixelated. I did it before on another PC and there it worked without problems. Also on the HV-Website, the examples don't get a better resolution by zooming in, although it worked on the other PC. Have imported this:

import holoviews as hv
import datashader
import holoviews.operation.datashader as hd
hv.extension('bokeh')

hd.datashade(hv.Curve((x,y)))

In the documentation is described, that a jupyther or bokeh server hast to be run. Didn't i done this with the hv.extension?

Maegges04
  • 13
  • 5

2 Answers2

0

hv.extension() loads the JavaScript code that's necessary for bidirectional communication in Jupyter, along with some miscellaneous convenience things like letting the user select which backend to use. It doesn't affect whether you have a server running, which is up to you to ensure. You'll have a server running if you run bokeh serve file.py, panel serve file.py, or if the above code is executed in a Jupyter notebook cell with a live Jupyter kernel. You will not have a server running if you export a jupyter notebook to .html or do bokeh html file.py or python file.py and then load a resulting .html file into your browser, or if you save a Jupyter notebook and then reload that .ipynb into another Jupyter without starting a kernel. There is no Python server running for the HoloViews website, and so you should not expect any of those examples to dynamically update on zoom either.

In your case, if you are running in Jupyter with a live kernel, try looking at your browser's JavaScript console output (Cmd-Opt-J in Mac Chrome, or in the menus for other browsers/OSs). Typically there will be an error message there saying what went wrong.

James A. Bednar
  • 3,195
  • 1
  • 9
  • 13
  • Yes, i'm working with Jupyter Notebook / Lab. I'm really sorry for bother you, but i didn't get any warnings or errors in the JS console. Two Kernel sessions are running in Jupyer lab, so a live kernel? I also have problems with troubleshooting the fact, that Jupyter lab doesn't open automatically by starting jupyter lab in the anaconda prompt. Therefore I have to copy the url **http://localhose:888/?token=[...]** in the browser. Does this affect the problem? – Maegges04 Apr 28 '20 at 12:45
  • 2
    You do have a kernel if you are running in JupyterLab and can execute Python code dynamically, so that part should be fine. Have you installed the JupyterLab extension as described at http://holoviz.org/installation.html, i.e. "jupyter labextension install @pyviz/jupyterlab_pyviz"? – James A. Bednar Apr 29 '20 at 15:14
  • I installed everything like given in the instructions... But nothing helped. Multiple reinstalled anaconda and created different environments. The original problem has not been solved yet. – Maegges04 May 05 '20 at 08:48
  • Please join us on our Discourse (https://discourse.holoviz.org/) and provide a full breakdown of your environment. – philippjfr May 06 '20 at 15:30
  • New status: After installing jupyter labextension multiple times, it now works... Don't know why, but it works now. Computer is having one of its good days... :D Thank u for helping. – Maegges04 May 07 '20 at 14:53
0

I managed to pin down the same issue with jupyterlab 4.0.X to this:

The solution was to reinstall the pyviz_comms with:

conda install pyviz_comms --force-reinstall

and then rebuilding jupyter lab with:

jupyter lab build

without rebuild it did not work (and of course refresh your browser/restart jupyter lab)

n4321d
  • 1,059
  • 2
  • 12
  • 31