speed-up the function to add country borders and coastlines #1073
Conversation
|
As an example of the grid spacing, with centroids as defined in the main climada tutorial (1_main_climada.ipynb): Of course on the other hand, the default 'filled borders' in e.g. a plot with France and it's colonies are a bit larger. |
peanutfun
left a comment
There was a problem hiding this comment.
Thank you for the quick PR! 🙌
Please add two entries into the changelog, one for the change to add_shapes and the speedup of the plotting, and one for the change in mask_distance. Also, please make sure the pre-commit hooks run through. The code is not properly formatted by black, see the failing test.
@ValentinGebhart, please check if you concur with the changes to mask_distance.
|
Thanks @timschmi95! Very much agree with the new default value of |
|
Changelog is updated and tests passing. Is it ready to merge then @peanutfun ? |
peanutfun
left a comment
There was a problem hiding this comment.
Huh? I must have been confusing by you closing the issue already, @timschmi95. I thought it was already merged. Good to go! 🚀
Changes proposed in this PR:
u_plot.add_shapesfunction. I've tested the new function in multiple tutorials and my own data (both with standard and projected coordinates) and the plots look equivalent.This PR fixes #1071
Substantially speeds up hazard.plot_intensity (and other functions that use
u_plot.add_shapes). Before the update, adding the country borders took up to 75% of the overall time, which is reduced to <1%, significantly speeding up the overall function.The change of the default mask_distance is somewhat independent. While testing many plots in the tutorial, we noticed in multiple instances the haz.plot_intensity() with the recently updated raster plotting functionality (PR #1047) showed small white gaps within gridded data, if there are fewer than ~80 regularly spaced gridpoints in one direction. Changing the default from 0.01 to 0.03 allows smooth plots for data with fewer (down to ~30) gridpoints per side, while keeping the default interpolation distance for data with gaps reasonably small.
Of course, users can always adjust the parameter depending on their data, but increasing the default will lead to fewer user seeing gaps in their gridded hazard data plot and needing to dig into the keyword arguments of the haz.plot_intensity() function.
PR Author Checklist
develop)PR Reviewer Checklist