Open ahuang11 opened 10 months ago
For reference, I was able to demonstrate each of these issues using this code on a machine without Cartopy installed (i.e., an environment with hvPlot but without GeoViews):
import hvplot.xarray, xarray as xr, holoviews as hv
hv.extension("bokeh", width=100)
ds = xr.tutorial.open_dataset('air_temperature').load()
hvplot.explorer(ds, x="lon", y="lat")
Notice the small size of the plot relative to the page width, which is why the horizontal placement of the HoloMap/DynamicMap widgets is a problem. For different zoom levels and browser widths the issue gets much worse; I think we really need to be reserving much more of the screen real estate for the plot:
General comment on naming things, I definitely want the name of the parameters passed to hvplot.<method>
appear in the explorer as I strongly believe the explorer is a good way to discover and learn hvPlot's (and Pandas' to some extent) API. It's for example the best way to understand the difference between by
and groupby
, just by playing around with the options. So their actual names must be shown in the UI.
Two time(s) in the fields list: 'time' and "['time']"
I've seen this issue with a dev release of Param, just make sure to reproduce it with Param 2.0.
So their actual names must be shown in the UI.
I believe there is an open issue about the name of "by" vs "groupby", and I agree that changes in the Explorer should be coordinated with changes in the API. I continue to find the naming of those two options highly confusing.
I think this is the issue you're referring to: https://github.com/holoviz/hvplot/issues/157.
Note that I just referred to by
and groupby
as an example, my comment applies to all the parameters exposed by the explorer.
I definitely want the name of the parameters passed to hvplot.
appear in the explorer as I strongly believe the explorer is a good way to discover and learn hvPlot's
Perhaps we can keep the original parameter names by using label
like:
"Colorbar Limit (clim
)"
Yes that works for me!
After trying to add framewise
, I realize it may not be ideal for explorer because it requires dynamic=False
and if there's 2320 frames in the plot, it takes a while to load.
I also discovered that axiswise
is not implemented as a kwarg i.e. .hvplot(axiswise=True)
is invalid.
Regarding:
- [ ] Infer color label from dataset (right now colorbar has no text label even though the dataset declares that dimension; explicitly filling in "Temp" works on clabel)
It seems that HoloViews doesn't automatically label colorbar so I think that's an issue for HoloViews: https://github.com/holoviz/holoviews/issues/5982
import hvplot.xarray
import xarray as xr
import holoviews as hv
hv.extension("bokeh")
ds = xr.tutorial.open_dataset("air_temperature").isel(time=0)
hv.Image(ds, ["lon", "lat"], ["air"]).opts(colorbar=True).redim.label(lat="Latitude", lon="Longitude", air="Air Temperature")
- [ ] Bokeh invalid tooltip (e.g. for "HoloViews .opts()" under Advanced, the tooltip disappears permanently after entering invalid input
I believe this is a Panel issue: https://github.com/holoviz/panel/issues/5857
import param
import panel as pn
pn.extension()
class Test(param.Parameterized):
d = param.Dict(doc="This should not disappear")
pn.Param(Test)
These things make the explorer unintuitive to use:
Minor
Medium
Major
TBD
Discussion needed
Out of scope in hvPlot
Credits to @jbednar for discovering these.