Where communities thrive


  • Join over 1.5M+ people
  • Join over 100K+ communities
  • Free without limits
  • Create your own community
People
Activity
    Rich Signell
    @rsignell-usgs
    Guys, I'm back from vacation and digging out from email pile. Just wanted to let you know I'm alive. :-)
    Harman Deep Singh
    @hdsingh
    :smile:
    Harman Deep Singh
    @hdsingh
    @martindurant I had a look at interface with intake. It is extremely slow. For me it took about 4 seconds just to load in new tab and 11 seconds in the notebook. Also after selecting a variable it takes about 120 seconds just to enable plot button. During this time CPU usage is 100% for 1 core. This is really strange!
    Philipp Rudiger
    @philippjfr
    Is it the browser process or the python process that's using that much CPU?
    Harman Deep Singh
    @hdsingh
    It is browser process.
    Philipp Rudiger
    @philippjfr
    Quite worrying it's likely bokeh's layout solver then. I'll have to take a look.
    Rich Signell
    @rsignell-usgs
    Rich Signell
    @rsignell-usgs
    Very nice demo for the Pangeo group @hdsingh !
    I like the command line "ncview replacement" idea
    Harman Deep Singh
    @hdsingh
    Thank You! Just the audio caused a bit trouble. I think argparse with a script to launch dashboard can become "nc replacement".
    Rich Signell
    @rsignell-usgs
    The audio improved somewhat near the end. I'm guessing it was just a period of limited bandwidth
    Martin Durant
    @martindurant
    I find clickmuch easier to use for CLIs, although the choice is up to you. We probably won’t expose too many options, so it probably doesn’t matter; even simple sys.argv may be enough.
    Harman Deep Singh
    @hdsingh
    I will look into click and create a PR for this after 23rd Aug.
    Rich Signell
    @rsignell-usgs
    I see this updated https://github.com/intake/xrviz/blob/master/docs/source/index.rst but the read the docs page https://xrviz.readthedocs.io/en/latest/ didn't update. I guess this doesn't build via CI?
    Martin Durant
    @martindurant
    Triggered doc build now, don’t know why it’s not doing that automatically
    Harman Deep Singh
    @hdsingh
    I have filled the final evaluation form and submitted https://github.com/intake/intake-gsoc-gui/blob/master/Report%20for%20GSOC%202019.md as link to work.
    Rich Signell
    @rsignell-usgs
    :thumbsup:
    I can submit the required student evaluation on August 26
    Harman Deep Singh
    @hdsingh
    @rsignell-usgs @martindurant My exams are over now. Let me know when you would be available for the meeting.
    Rich Signell
    @rsignell-usgs
    Harman Deep, congrats!
    @martindurant are you available for one last Wednesday-at-10AM-Eastern Meeting tomorrow?
    Harman Deep Singh
    @hdsingh
    Thank You! :tada:
    Martin Durant
    @martindurant
    I believe I can be
    Rich Signell
    @rsignell-usgs
    Okay, let's try. We can use https://appear.in/rsignell
    Rich Signell
    @rsignell-usgs
    Harman Deep Singh
    @hdsingh
    Lets go with https://appear.in/rsignell as mentioned earlier.
    Martin Durant
    @martindurant
    coming in a moment, and not for too long...
    Martin Durant
    @martindurant
    One example of astronomy using xarrays directly: https://github.com/ska-sa/dask-ms
    Rich Signell
    @rsignell-usgs
    ds_lonlat = xr.open_dataset(fsspec.open('s3://esip-pangeo-uswest2/pangeo/NWM/nwm-v1.2-channel_spatial_index.nc'))
    can't quite figure out how to use fsspec here
    Martin Durant
    @martindurant
    with fsspec.open(‘s3:…’) as f:
        ds_lonlat = xr.open_dataset(f)
    ?
    or
    fs = fsspec.filesystem(‘s3’, …)
    f = fs.open(…)
    ds_lonlat = xr.open_dataset(f)
    Rich Signell
    @rsignell-usgs
    yup. Thanks!
    Martin Durant
    @martindurant
    To clarify: fsspec.open returns an OpenFile, which defers actual intraction with the storage backend until within a “with” context, so that it can, for example, be safely serialised between Dask workers.
    Rich Signell
    @rsignell-usgs
    ah, thanks ! I was just trying to figure that out!
    but this is okay, right? ds = xr.open_zarr(fsspec.get_mapper('s3://esip-pangeo-uswest2/pangeo/NWM/2010'))
    because that's a map
    Martin Durant
    @martindurant
    Yes, that’s right
    Harman Deep Singh
    @hdsingh
    Since pyviz/geoviews#347 has been closed now I will test the dashboard with latest version of holoviews and geoviews and see if it shows correct behaviour.
    Rich Signell
    @rsignell-usgs
    Excellent. Fingers crossed!
    Harman Deep Singh
    @hdsingh

    This works(old setup)

    holoviews.__version__ : '1.12.1'
    geoviews.__version__: '1.6.3'
    xarray.__version__ :'0.12.3'

    The latest versions do not work.

    holoviews.__version__ : '1.12.5'
    geoviews.__version__: '1.6.3'
    xarray.__version__ :'0.13.0'

    With the latest hv, gv, xr the quadmesh is not displayed.
    I have also tried installing these 3 directly from source but the behaviour still remains same.

    Screenshot from 2019-09-22 11-28-35.png
    Philipp Rudiger
    @philippjfr
    There is a known bug with the versions on master I'll push a fix for. Not sure what's up with the latest released versions though.
    Harman Deep Singh
    @hdsingh

    @philippjfr
    There is inconsistency between holoviews installed using two different methods.

    1. Upon installation using conda install xarray cartopy panel hvplot netcdf4 datashader the hv installed is:

    https://repo.anaconda.com/pkgs/main/noarch/holoviews-1.12.6-py_0.tar.bz2

    Output of print(graph):

    Row
        [0] Row
            [0] HoloViews(DynamicMap)
            [1] Column
                [0] DiscreteSlider(name='Sigma Stretched V..., options=OrderedDict([('0', ...]), value=0.0)
                [1] DiscreteSlider(name='Time', options=OrderedDict([('2019-03-20 ...]), value=numpy.datetime64('2019-03-...)

    2 . Upon installation using env file hv installed is:
    https://conda.anaconda.org/conda-forge/noarch/holoviews-1.12.6-py_0.tar.bz2

    Output of print(graph):

    Row
        [0] Row
            [0] HoloViews(DynamicMap)
            [1] Column
                [0] WidgetBox(css_classes=['widget-box'])
                    [0] DiscreteSlider(margin=(20, 20, 5, 20), name='Sigma Stretched V..., options=OrderedDict([('0', ...]), value=0.0, width=250)
                    [1] DiscreteSlider(margin=(5, 20, 20, 20), name='Time', options=OrderedDict([('2019-03-20 ...]), value=numpy.datetime64('2019-03-..., width=250)
                [1] VSpacer()
    For same versions first output does not have WidgetBox while second has, which is resulting in errors.
    Philipp Rudiger
    @philippjfr
    @hdsingh I'd have to see the full set of version. The latter looks like the output I'd expect when using HoloViews 1.13.0devX
    Rich Signell
    @rsignell-usgs
    @hdsingh maybe try this again as there have been new releases I think
    @hdsingh, can you remind me, you are now a junior in CS at which university?
    Martin Durant
    @martindurant
    ^ this means: which year of your degree are you in (e.g., “in year three out of four”). I don’t know what a “junior” is either :)
    Harman Deep Singh
    @hdsingh
    @rsignell-usgs As of now I am a third year undergraduate student pursuing B.Tech. Computer Science from Guru Nanak Dev University, Amritsar. As of now in 5th out of total 8 semesters. Sorry for the late reply.
    Since everything works fine with conda-forge version, which is where xrviz is too, this issue has been resolved (as explained here).