Hi all,
Apologies in advanced, I don’t know how to recreate the error exactly- the behavior is seemingly random- but I’ll give as much code as I can.
I have an app of type pn.viewable.Viewer with many parameters, including tap_x = param.Number(np.nan) and tap_y = param.Number(np.nan). I also have a DoubleTap stream initiallized as
self._dtap = hv.streams.DoubleTap(
x=np.nan,
y=np.nan,
source=self._tile
)
where self._tile is a geoviews tile (geoviews.tile_sources.EsriImagery()). This stream is used in a pn.bind function as follows:
timeseries_plot = pn.bind(
self._timeseries_plot,
self._dtap.param.x,
self._dtap.param.y,
self.param.time,
self.param.depth
)
def _timeseries_plot(self,x,y,time,depth):
if (not np.isnan(x)): # at app initialization, don't return any plots
# if a point is selected, transform from web_mercator to PlateCarree
lon,lat = self._plot_proj.transform_point(x, y,self._tile_proj)
da = self._roms_ds.isel(
s_rho=depth
)
da = da.sel(lat_rho=lat, lon_rho=lon)
if (da.mask_rho == 1)): # check if selected point is in valid domain
# loop through variables, create time series of each at given point
time_x = self._roms_ds.ocean_time.values[int(time/3)]
overlay = da[list(VAR_OPTIONS.values())[0]].hvplot(
x='ocean_time',
dynamic=False,
title=list(VAR_OPTIONS.keys())[0]
).opts(labelled=[],active_tools=[],max_height=150,width=250,)
overlay *= hv.VLine(time_x).opts(color='r')
for var in list(VAR_OPTIONS.values())[1:-1]:
var_name=list(VAR_OPTIONS.keys())[list(VAR_OPTIONS.values()).index(var)]
plot = da[var].hvplot(
x='ocean_time',
dynamic=False,
).opts(labelled=[],active_tools=[],max_height=150,width=250,title=var_name,)
plot *= hv.VLine(time_x).opts(color='r')
overlay += plot
self._text.object = f'Timeseries at location:\n {lat:.2f}°N, {lon:.2f}°E'
return overlay
else: self._text.object = ''
else: self._text.object = ''
I know there is a lot that is unexplained above, but the idea is that a map is double clicked and a set of timeseries are returned at that point; however, if a point outside the valid domain is clicked then no plots are returned. Sometimes this works perfectly, while other times I get an import error about numpy. I can fix this by adding import numpy as np in the functions, but then I sometimes get an import error about xarray. I can import xarray in the function as well with no problems, but it seems silly to have to import the same packages every time a function is called. Is there something going on with the threadpool workers? I know in some mpi and threading packages you must manually copy your active environment to each worker- perhaps there’s a way for me to do that?