Call .compute() in all plot methods?

[johnomotani]

I noticed what I think might be a performance bug: should .compute() be called on the input data in all the plotting methods (e.g. plot.pcolormesh()) like it is in .plot() here

xarray/xarray/plot/plot.py

Line 166 in bf27e2c

darray = darray.squeeze().compute()

See also discussion in #2390.

I was making plots from a large dataset of a quantity that is the output of quite a bit of computation. A script which made an animation of the full time-series (a couple of thousand time points) actually ran significantly faster than a script that made pcolormesh plots of just 3 time points (~2hrs compared to ~5hrs). The difference I can think of is that the animation script called .values before the animation function, but the plotting script called xarray.plot.pcolormesh() without calling .values/.load()/.compute() first. A modified version of the script that calls .load() before any plot calls reduced the run time to 30 mins even though I plotted 18 time points, not just 3.

2d plots might all be covered by adding a darray = darray.compute() call in newplotfunc()?

xarray/xarray/plot/plot.py

Line 609 in bf27e2c

def newplotfunc(

I guess the 1d plot functions would all need to be modified individually.

[TomNicholas]

We should be calling .compute() before passing anything to matplotlib - if we're not then that's a bug. I've also noticed that not doing so can cause huge slowdown.

I think that #5561 might (or really should) fix this as a side effect though.

[TomNicholas]

But let me know if you spot anywhere that #5561 isn't going to cover with to_numpy() calls!

I guess the 1d plot functions would all need to be modified individually.

@Illviljan was talking about adding a @_plot1d decorator like the 2d one, which might be a good place to enforce this.

[dcherian]

We should create a DataArray with just the data that is needed based on x, y etc. and then call compute on the whole thing, rather than computing x separately and y separately.