For questions, use https://discourse.holoviz.org/ instead! This chat is for informal discussions and debugging sessions with developers of the HoloViz.org tools (HoloViews, Datashader, Panel, hvPlot, GeoViews, Param, etc.), but user questions should go to the HoloViz Discourse site so that others won't have to ask them again.
That's the same as @MarcSkovMadsen's question about hvPlot -- in no way does hvPlot take away from HoloViews development; hvPlot is a user-facing way to exploit functionality provided by HoloViews, with a different userbase and use case than HoloViews itself, but nearly everything it does is implemented by HoloViews and not by hvPlot directly. Simiarly, Lumen is a user-facing way to exploit functionality in Panel (and partly HoloViews), with a potentially different userbase (not even necessarily Python programmers) and different use cases ("standard" dashboards, not fully custom), but again what it provides is fundamentally implemented by Panel, not Lumen. Neither Lumen nor hvPlot have any way to exist without Panel or HoloViews, respectively.
I think that there are so many packages names in the holoviz ecosystem, so noob people like me get confused about the functionalities, and which have to use. In matplotlib for example, you only have matplotlib, and pyplot is only a subpackage. Then most people talk about matplotlib, but they only use pyplot without OOP. Something similar in holoviz would be useful, in order to increase its popularity and widespread usage.
I got it running: two
sensor processes emitting data over websockets, and a notebook that creates a dashboard displaying the data with a button that calls a remote procedure in one of the sensor processes! More info (and a link to the code) is at https://discourse.holoviz.org/t/panel-meets-d3-js-and-websocket/1635/13
@MarcSkovMadsen
Any one who could tell me how to add a bokeh Plot when developing a statscard as a custom bokeh extension? https://discourse.bokeh.org/t/how-do-i-create-bokeh-extension-which-includes-a-stackline/6991
[
](https:/
Any one who could tell me how to add a bokeh Plot when developing a statscard as a custom bokeh extension? https://discourse.bokeh.org/t/how-do-i-create-bokeh-extension-which-includes-a-stackline/6991
[
A new Panel competitor IDOM https://github.com/idom-team/idom. https://github.com/idom-team/idom-plotly-dash. https://github.com/idom-team/idom-streamlit. Quite interesting approach.
You can just import any React library and use it. I.e. no spending time on wrapping those elements https://idom-docs.herokuapp.com/docs/javascript-modules.html
i have abhypothesis that there is room for one more analytics app framework. Something in the spirit of fast api. That integrates with it, uses modern python in the same way and can associate it self with fast api and ride that wave.
It would also be nice to have one React and one Vue example. That might lead to more contributions from the community if there where some examples that could be used as a starter.
E.g. before HoloViews had Sankey, wrapping D3 Sankey would have been good. Now, anything else like that that's available in D3 but not in Bokeh or HoloViews.
You could play around with the preliminary PR holoviz/panel#1894 if you want @ea42gh . That could help shape the api and provides some examples for the gallery. It might take a undocumented commands to be able to develop including some commands like
panel build panel or python -c "from panel.compiler import bundle_resources;bundle_resources()".
@philippjfr has started adding documentation to the PR and I've started working on some examples
@philippjfr , might you be able to take a quick look at: https://discourse.holoviz.org/t/hvplot-reporting-inconsistent-chunks-on-xarray-dataarray/1690. I have no idea what is going on there.
are the jscallback function and the names of the Spinner and Select widgets on https://panel.holoviz.org/user_guide/Links.html inconsistent?
This makes it so much clear that the full functionality is only available in the Notebook. And thus that this is very much a vibrant and alive framework. Congrats community. Thanks Jean-Luc Stevens.
Yep, I think it's much clearer now. If anyone notices that the problem is displayed on a page that really doesn't need a Python server (if there's no interactivity, or if all interactivity is JavaScript based), than please file a bug report, because we can easily turn off the blanket warning for specific pages.
Usually HoloViews itself doesn't do much itself, since it defers the rendering to an underlying plotting library. If you use a Dask dataframe and then call
datashade or rasterize before display, rendering the individual frames of the holomap will be parallelized. But maybe file an issue on HoloViews to see if there's anything about the entire data structure that can be done in parallel when you have a HoloMap.
holoviz/holoviews#3523 "There will likely never be a solution that works in parallel here. It has to evaluate the state space which means modifying the model and then recording the changes applied to it. It isn't possible to evaluate multiple state changes at the same time. Therefore I'll regretfully have to close this as wontfix."
But even if it was, it all works by modifying some object whether that's a mpl figure or a Bokeh model. We'd have to copy the figure to modify it in parallel and at least for mpl that's definitely not possible.
But anyway, yes, whether even a HoloMap could be done in parallel is an open question; seems like if it's rendering to PNG for Matplotlib it's possible at least in principle.
@dask.delayed
def plot(ds, time):
ds.sel(time=time)['air'].plot()
plt.savefig(time)
dask.compute([
plot(ds, time) for time in ds['time'].values],
scheduler='processes', num_workers=4)