Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
  • Dec 12 21:52
    jbednar closed #132
  • Dec 12 21:52
    jbednar commented #132
  • Dec 12 21:26
    philippjfr labeled #405
  • Dec 12 21:26
    philippjfr unlabeled #405
  • Dec 12 21:26
    philippjfr labeled #406
  • Dec 12 21:26
    philippjfr unlabeled #406
  • Dec 12 21:26
    philippjfr commented #4126
  • Dec 12 21:26

    philippjfr on master

    fix XArrayInterface.aggregate h… (compare)

  • Dec 12 21:26
    philippjfr closed #4126
  • Dec 12 21:25
    philippjfr labeled #4126
  • Dec 12 19:52
    jbednar reopened #846
  • Dec 12 19:52
    jbednar closed #846
  • Dec 12 19:52
    jbednar commented #846
  • Dec 12 19:47
    jbednar commented #846
  • Dec 12 19:44
    jbednar commented #846
  • Dec 12 19:40
    nluetts opened #4126
  • Dec 12 18:24
    brendancol edited #847
  • Dec 12 18:23
    brendancol edited #847
  • Dec 12 17:47
    jbednar commented #846
  • Dec 12 17:47
    jbednar commented #846
Philipp Rudiger
@philippjfr
I'd love to have a solution for this but all that I can think of would be either exceedingly hacky or very expensive.
JieLin
@JieLin69105075_twitter
hello, have you ever used cartoframes(a geographic visualization library)? How about it compare to geoviews?
James A. Bednar
@jbednar
I haven't tried it; can't say!
lancelot1969
@lancelot1969
When I create voronoi map from 2d points. number of my cells are equal to the number of points as expected. But when I am making them interactive by using geoviews polygons I am loosing some of the points in the process.
I mean len(cells.data) != len(gvpoints)
JieLin
@JieLin69105075_twitter
@jbednar OK! Thanks.
Julia Signell
@jsignell
@bryevdv they are built daily, but for the last month. I think I did that because the newest data isn’t made available at a predictable time of the month. I am just refreshing the whole cache daily since the page has too many badges to fetch them all on each hit
Bryan Van de Ven
@bryevdv
@jsignell thanks for the explanation!
lancelot1969
@lancelot1969
I figured out what was happening. The way convex hull buffer works it just merges some bordering polygons together so best way of treating this type of problems is considering boundary points as a separate object
James A. Bednar
@jbednar
@JieLin69105075_twitter , I looked briefly at cartoframes, and the immediate differences I can see are that GeoViews is completely open source, with no proprietary or paid features, and that GeoViews offers full support for multidimensional array data (satellite images across bands, climate simulations, etc.), while the examples I found when clicking around on Carto were much lighter weight (a few thousand points) compared to what GeoViews+Datashader can do (arbitrarily many billions of points, huge rasters, n-D arrays, etc.) There could be such support hidden in there, but given that it's "cartoframes" (i.e. dataframe-based) it's probably not likely to have n-D support.
Tobias Bartsch
@tobiasbartsch
Hi, I am trying to download the bokeh sample data using bokeh.sampledata.download(). This, however results in urllib.error.HTTPError: HTTP Error 403: Forbidden. Is this a bug or am I doing something wrong?
Jean-Luc Stevens
@jlstevens
@tobiasbartsch That sounds like a bug to report to bokeh to me.
Timothy W. Hilton
@Timothy-W-Hilton

howdy - is it possible to connect a RangeToolLink to a dimensioned container? I have some overlaid curves setup like this:

:HoloMap   [WRFrun]
   :NdOverlay   [area]
      :Curve   [time]   (HFX)

I'd like to connect them like in the timeseries range tool example. I'm finding that if I remove the overlay the RangeToolLink works as expected, but it disappears if I activate the overlay.

Tobias Bartsch
@tobiasbartsch
Ok, will do! Thank you!
lancelot1969
@lancelot1969
When I am running DataLink(table, table) and then (table+table).opts() I am getting error which looks like a bug: AttributeError: 'DataFrame' object has no attribute '_plot_id'
lancelot1969
@lancelot1969
DataLink only works between hv.table and hv.Polygons or other holoviews objects too?
Hayley Song
@cocoaaa
Another shoutout for Holoviews! It makes checking the segmentation mask so much easier! https://recordit.co/liiE1tAHyJ
JieLin
@JieLin69105075_twitter
@jbednar Thank you so much! Your answer means geoviews and holoviews are the best geo visual tools in Pyhton.
James A. Bednar
@jbednar
@limx0 , if you search for progress bar and panel you should be able to find various issues that people have raised and the somewhat hacky solutions that we have proposed for them. Yes, it's possible, no, it's not as easy and straightforward as it should be. Definitely future work and an active area of need.
Andrei Berceanu
@berceanu
If I take the classical holoviews mandelbrot example and add a vertical slice for x (besides the one for y already present), the holomap already becomes too heavy to display in the browser.
How so?
Andrei Berceanu
@berceanu
I understand it has to generate all the combinations of 21 x 21 values, but if I multiply that by the array size I get about 35 MB so not that big.
James A. Bednar
@jbednar
The size depends on the backend; with Matplotlib backend it stores the full PNG plot image for each combination, not just the original array size. You can replace the HoloMap with a DynamicMap to keep the size down, but then it won't export to static HTML.
Andrei Berceanu
@berceanu
I'm using the bokeh backend.
But I'm surprised the limitation is so low with HoloMap.
James A. Bednar
@jbednar
With Bokeh I'm not sure what it stores per combination.
Bryan Van de Ven
@bryevdv
If it's static HTML, you could take a look
James A. Bednar
@jbednar
Right, it's static HTML...
JieLin
@JieLin69105075_twitter
geoviews can plot large points. Does it can plot large curves and polygons??
Andrei Berceanu
@berceanu
Should I use heatmap or image?
(in holoviews)
James A. Bednar
@jbednar
Image is good for continuous axes; heatmap for categoricals. If you have one of each type of axis in the same plot, you'll have to hack one or the other.
Andrei Berceanu
@berceanu
Oh I see. I have continuous in both axes, so hv.Image it is :) tnx!
James A. Bednar
@jbednar
@JieLin69105075_twitter , GeoViews supports whatever data types HoloViews does, and HoloViews supports large data of many different types via Datashader, including points and curves: http://holoviews.org/user_guide/Large_Data.html
Datashader polygon support is being implemented as we speak; see pyviz/datashader#181 ; I expect it to be in master in a couple of weeks and then available in the following HoloViews and Datashader releases.
For now, large polygons are supported only as outlines, not filled shapes, and they may require converting into a Datashader-friendly storage format first (can't recall just what's been implemented in HoloViews for e.g. Shapely objects).
JieLin
@JieLin69105075_twitter
@jbednar Thank you
Andrei Berceanu
@berceanu
How come I can't add a group to an image sample?
x_sample = hv.DynamicMap(lambda x, y: image.sample(x=np.clip(x,bounds[0],bounds[2]), group='xsample'), streams=[pointer])
Dimension group not found in Image.
James A. Bednar
@jbednar
I think it would be something like x_sample = hv.DynamicMap(lambda x, y: image.sample(x=np.clip(x,bounds[0],bounds[2])), streams=[pointer]).relabel(group='xsample')
(Sample accepts ranges on user-defined dimensions, which can have arbitrary names and so we don't want to have a lot of keywords supported, so you have to do that sort of thing separately.)
Andrei Berceanu
@berceanu
I see, thanks!
Can I get the overall min and max values of an Image?
James A. Bednar
@jbednar
i = hv.Image(...) ; i.data will give you the underlying array, with which you can do whatever you like if it's not already supported. (i.e. i.max() may work, but e.g. np.amax(i) will if that doesn't, for an image backed by a Numpy array).
Andrei Berceanu
@berceanu
gotcha!
sameerCoder
@sameerCoder

@jbednar
As suggested i have installed pip install --upgrade git+https://github.com/pyviz/panel.git#egg=panel
My panel version also got upgraded to 0.7 as it is showing in conda list .
conda list
panel 0.7.0a14.post21+gb55b21c pypi_0 pypi

Still i am getting Div Error
ValueError: Div type not found, could not apply options.
I have imported panel also and i am using pn.Column for my finalplot.
My little code as below :-

hv.extension("bokeh")
hv.extension("matplotlib")

allplot={(k.strftime("%Y-%m-%d %H:%M:%S"),r):plotthis(k,r)for k in perdelta(strt, strt + timedelta(days=1), timedelta(hours=18)) for r in ['O','A','Wh','T_N']}



df_div = hv.Div("""
    <figure>
    <img src="https://i.ibb.co/imglogo.png" height='80' width='90' vspace='-10'>

    """)

df_div1 = hv.Div("""
    &nbsp<center><b><p style="color:#B22222";font-size:80px;font-family:Times new roman><h1 style=font-size:20px;margin-left:2.5em;margin-top:-1em;color:#B22222>I Services<br />
    (IN)</h1></p></b></center>

    """)

tiles = gv.tile_sources.Wikipedia
hmap1 = hv.HoloMap(allplot, kdims=['Date and Time :','region'])
#hmap2 = hv.HoloMap(allplot2, kdims=['Date and Time :','region'])


dd=df_div.opts(width=70, height=70)
dd1=df_div1.opts(width=600, height=90)
dd2=df_div2.opts(width=100,height=10)
finalplot=pn.Column(pn.Row(dd, dd1, dd2), tiles*rasterize(hmap1).options(**opts)*logo1.opts(hooks=[absolute_position], apply_ranges=False)).servable()

#finalplot
hv.output(finalplot, holomap='gif', fps=3, backend='matplotlib')
Div_Errror_msg.png
@philippjfr @jbednar
Kindly let me know where i am doing wrong.
James A. Bednar
@jbednar
Remove hv.extension("matplotlib") and backend='matplotlib'; you can now ignore matplotlib and just keep using Bokeh as you had been before.