Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • 17:20
    codecov[bot] commented #2306
  • 17:19
    codecov[bot] commented #2306
  • 17:19
    codecov[bot] commented #2306
  • 17:18
    codecov[bot] commented #2306
  • 17:14
    codecov[bot] commented #2306
  • 17:09
    codecov[bot] commented #2306
  • 17:09
    codecov[bot] commented #2306
  • 16:51
    Hoxbro edited #2306
  • 16:51
    codecov[bot] commented #2306
  • 16:51
    Hoxbro synchronize #2306
  • 15:17
    codecov[bot] commented #2306
  • 15:17
    MarcSkovMadsen commented #2320
  • 15:17
    codecov[bot] commented #2306
  • 15:16
    MarcSkovMadsen commented #2320
  • 15:16
    codecov[bot] commented #2306
  • 15:15
    MarcSkovMadsen labeled #2320
  • 15:14
    codecov[bot] commented #2306
  • 15:10
    MarcSkovMadsen opened #2320
  • 15:07
    codecov[bot] commented #2306
  • 15:07
    codecov[bot] commented #2306
gbonx
@gbonx
Hello, is it possible to display contour ticks values as on matplotlib?
As in following screenshot :
image.png
I would like to do the same with hvplot. Can anyone help me please?
Andrew
@ahuang11
I don't think it's possible quite yet
holoviz/holoviews#4494
unless you hack together .hvplot.contours + manually .hvplot.labels
gbonx
@gbonx
Ok thank you @ahuang11. I may try to do something with hvplot.labels. Otherwise integrating this option would be great for people like me who are colorblinds ^^
James A. Bednar
@jbednar
Seems like that should be possible with HoloViews with the Matplotlib backend, using a plotting hook if it's not currently exposed as an option. But doing it with Bokeh would require more work.
gbonx
@gbonx
Ok thank you @jbednar. I thought also about doing it with holoviews and matplotlib backend but it does not seems straightforward and we lose bokeh features.
Marc Skov Madsen
@MarcSkovMadsen
I am currently promoting Panel 0.11 on twitter and linkedin. Feel free to retweet or reshare if you like and think Panel could deserve some more attention.
Robin B
@code-inflation
Hi, I have an issue with bokeh. I would like to use it in a Jupyter notebook and load the JS using the output_notebook function.
However this needs access to a CDN. The documentation mentions, that you can use the resources parameter to set a custom location for the BokehJS. However I don't understand how I have to set that parameter. https://docs.bokeh.org/en/latest/docs/reference/io.html#bokeh.io.output_notebook
Has anyone used this or knows how it works?
Philipp Rudiger
@philippjfr
This room is for dev discussion of HoloViz tools. For bokeh questions refer to the bokeh discourse.
But from memory you import INLINE from bokeh.io.resources and pass that to output_noteboik
Robin B
@code-inflation
oh I see, thanks for the information
narner90
@narner90

Hi all, I'm attempting to use a value_format for datetime column in a holoviews table (bokeh backend) with the goal of showing nanoseconds rather than the default format of milliseconds. If this question is more appropriate for Github please let me know!

Here's a snippet of what I'm trying to do. Neither my format function or pprint_value in dimension.py are called:

    def datetime_format_function(x) -> str:
        return x.strftime('TEST: %s')

    time_dimension = hv.Dimension(('sending_time_dt', 'sending_time'),
                                  value_format=datetime_format_function,
                                  type=str)

    return hv.Table(data, kdims=[time_dimension], values=['x', 'y'])
narner90
@narner90
Additionally, if I change the DateFormatter in bokeh/tabular.py from '%Y-%m-%d %H:%M:%S.%3N' to '%Y-%m-%d %H:%M:%S.%9N', the displayed value shows 6 trailing 0's rather than the microseconds and nanoseconds, which I can confirm are present in the underlying data
James A. Bednar
@jbednar
As a usage question, that's more appropriate for https://discourse.holoviz.org/c/holoviews than for either here (dev chat) or github (specific bugs and feature requests).
Marc Skov Madsen
@MarcSkovMadsen
Most easy things to improve 1) More info about support options 2) Deployment guides
Marc Skov Madsen
@MarcSkovMadsen
Most difficult things to improve 1) The api - gets some feedback on being hard to learn.
Irv Lustig
@Dr-Irv
This issue holoviz/holoviews#4875 is being seen my me, but also a colleague at my firm, plus the other folks who posted about the issue. We can't render anything with holoviews using bokeh right now. Hope you can look into it.
James A. Bednar
@jbednar
Scary! @philippjfr ?
Philipp Rudiger
@philippjfr
The main problem is that I cannot reproduce under any conditions I've tried.
Irv Lustig
@Dr-Irv

@philippjfr I can reproduce as follows:

conda create -n hvtest0402 python=3.8
conda activate hvtest0402
conda install pandas holoviews
conda install jupyter
jupyter-notebook

Then I ran the file https://holoviews.org/user_guide/Applying_Customizations.html but changing the hv.extension call to only use bokeh and cannot get any visualization.

This is on Windows.

Hoxbro
@Hoxbro
Could recreate the problem with a clean install in WIndows. It seems to be related to packages in thedefaults channel if I run conda update --all -c pyviz -c conda-forge -y in the environment I could get it up and running. Though it creates another problem where I could not start a jupyter notebook, so I used jupyter lab instead of.
Philipp Rudiger
@philippjfr
Defaults doesn't yet have latest panel so that won't work.
epifanio
@epifanio
Hi, attemting to run the following example https://geoviews.org/user_guide/Resampling_Grids.html will trwo an errror: AttributeError: 'DataArray' object has no attribute '_file_obj'
James A. Bednar
@jbednar
That sounds like a bug report! Please file it at the geoviews github repository.
Thanks!
epifanio
@epifanio
yes, I was wondering of where to report - I will fill an issue with the version of sw I am using
James A. Bednar
@jbednar
Perfect.
Philipp Rudiger
@philippjfr
@jbednar You should have recognized that as the datashader bug that we fixed recently :)
So @epifanio just update to Datashader 0.11.1
Not even we, you fixed it 😉
epifanio
@epifanio
@philippjfr before filling the issue, I was attemting a rebuild making sure the packages are all installed with a single tool (mamba in my case) and double check I do not have any of the pyviz libraries installed via pip - I will wait the end of docker build and test it.
@philippjfr you meant datashader v0.12.1 , right?
Philipp Rudiger
@philippjfr
I did sorry
epifanio
@epifanio

no prob!
I checked the latest tagged releases to make sure the problem was on my side, so I ended up runing (on the running server):

 pip install geoviews==1.9.1
 pip install holoviews==1.14.2
 pip install param==1.10.0
 pip install datashader==0.12.1

and I saw the bug was no longer there .. bu this also messed-up my jupyter widgets .. hence a rebuild
Now I want to check if mamba is doing the right things otherwise I will try pip.

James A. Bednar
@jbednar
I don't keep knowledge of fixed bugs in my memory; always onwards and upwards! :-)
But I have noticed that I really had to force conda to update datashader to anything past 0.10 even specifying the pyviz channel, which may be why people are still running into this.
epifanio
@epifanio
I confirm that, mamba installed the older version - manually upgrading with pip fixes the problem
Michael Aye
@michaelaye
is using opts.defaults to set default for the whole notebook supposed to work for hvplot? If yes, I'll submit a new issue.
Michael Aye
@michaelaye
is this sentence in the hvplot docs potentially incomplete?
x_sampling/y_sampling (default=None):
    Declares a minimum sampling density beyond.
James A. Bednar
@jbednar

is using opts.defaults to set default for the whole notebook supposed to work for hvplot? If yes, I'll submit a new issue.

My guess is that opts.defaults would work only for options not processed as hvplot command-line arguments, except maybe for some that are just passed through. Basically, HoloViews's option system lets you use opts.defaults to set global defaults, but then those defaults get overridden by any values specified explicitly. Due to how hvPlot works, I would assume that it sets various options explicitly even when the user doesn't provide values, in part because hvPlot defaults need to match Pandas defaults, not HoloViews defaults. So no, I don't think hvPlot not being responsive to changes in opts.defaults counts as a bug.

And yes, that sentence is incomplete. The corresponding HoloViews docstring for x_sampling is "Specifies the smallest allowed sampling interval along the x axis.".
A PR to fix that sentence would be welcome!
Marc Skov Madsen
@MarcSkovMadsen
FYI. My talk "Awesome Analytics Apps Using HoloViz Panel" has been accepted for SciPy 2021. See you there?
Philipp Rudiger
@philippjfr
Congrats @MarcSkovMadsen thats awesome ;)