Hi, I am finally back home and did the timing on my Windows PC (compared to my Linux Laptop). It seams that, the loading of the geojson compares better in the new set up compared to the Linux one.
Compare Execution time Python / Julia for Raster processing / zonal stat
using btime from BenchmarkTools for Julia and time in Python using the medium time of three runs.
I only see the speed-up while loading the vector on Windows, on Linux it seams to be slower.
In Rasters Python outperforms Julia in up to 4 orders. As shown of the Zonal stats before.
| Python | Julia | Speed-up | |
|---|---|---|---|
| unit | s | s | |
| Loading vector (geojson) | 18.15 | 11.07 | 1.55 |
| Loading raster (tif) | 0.0330 | 0.0915 | 0.36 |
| Merging 2 raster | 0.176 | 18.71 | 0.009 |
| Zonal stat | 0.1070 | 44.22 | 0.0024 |
Although speed is normally defined for multi threading/-processing. 
Are there other things, I shall try the timing for? 
Thanks at @evetion for the link for the documentation. I tried it out on Linux, Julia 1.7.0 packages newly updated. Some functioned referenced in the source, other did not. Don’t know why. I did not try every function, but most of them.
By the way, how can I get the union of all geometries in a GeoDataFrame? Is it the unary union?
Functions on Polygons
using ArchGDAL as reference.
Unary
Why union here?
| attributes | predicates | Immutable Operations | mutable Operations | |
|---|---|---|---|---|
| found | geomlength, geomarea | issimple, isring | boundary, convexhull , buffer, centroid | |
| not found | geomdim, getcoorddim, envelope | isempty, isvalid | simplify, union | closerings! |
Binary
| predictors | Immutable Operations | |
|---|---|---|
| found | intersects, equals, disjoint, touches, crosses, within, contains, overlaps | intersection, difference , symdifference |
| not found | union |
I don’t know, why some could not be found. Btw. are there some anual (virtuell) meetings/ discussions in the Julia gis community?