slow flopy.plot.PlotMapView.plot_array

[wkitlasten]

Updated to flopy version 3.3.4 and existing calls that took seconds are now taking minutes. Admittedly my grid is large, and the method works find for smaller grids, but thought it might be worth mentioning.

A quick example with 10 million cells that takes over 10 minutes to generate the plot (which is longer than the model run and only took a couple seconds with 3.3.3):

sim = flopy.mf6.MFSimulation(sim_name='big_grid',exe_name="mf6",
                             version="mf6",sim_ws='big_grid')
tdis = flopy.mf6.ModflowTdis(sim,pname="tdis",time_units="DAYS",
                             nper=1,perioddata=[1.0,1,1.0])

gwf = flopy.mf6.ModflowGwf(sim, modelname='big_grid')
dis = flopy.mf6.ModflowGwfdis(gwf,nlay=1,nrow=1000,ncol=10000,delr=1,delc=1,top=10,botm=0)
pmv=flopy.plot.PlotMapView(modelgrid=gwf.modelgrid)
pmv.plot_array(dis.top.array)

[jlarsen-usgs]

@wkitlasten

It looks like the slow down has to do with building patches. I moved everything over to PatchCollections to simplify the plotting code and make it uniform between modelgrid types in version 3.3.4. I'm looking for solutions for redering polygons and polygon collections quickly, to speed things up. Thanks for reporting this issue!

[jlarsen-usgs]

@wkitlasten

I found a solution that will work for now to speed up the plotting routines. 1) for structured grids we will revert back to using a QuadMesh instance instead of a PatchCollection. As you pointed out, QuadMesh plots 10 million cells in ~2 seconds, PatchCollection was taking ~10 minutes.

For Unstructured and Vertex based modelgrids, I've updated the PlotMapView code to use PathCollection. This shows quite a bit of spead improvement over PatchCollection. Timings here for 10 million cells are ~35 seconds for PathCollection which is a great improvement over PatchCollection's 10 minutes.

I'll be putting in a PR sometime today with these changes.