Replace Stamen tiles + fix other deprecations

[AmitAronovitch]

Resolves #329

This PR is for making telluric work with more up-to-date versions python and packages.

Changes included so far:

1. Use CartoDB refernce maps instead of Stamen (which are not available any more - see folium #1803
2. Remove dependency on deprecated pkg_resources (not required for Python>=3.7)
3. Fix some failing tests and linters

[matiasg]

Hi, @AmitAronovitch !!
Are you still using telluric? Do you want me to try find somebody to review the PR?

[AmitAronovitch]

Hi, @AmitAronovitch !! Are you still using telluric? Do you want me to try find somebody to review the PR?

Hi @matiasg , yes that would be great :-)

Note however that releasing a new version of telluric still requires some work.

This is because the dependencies (mainly rasterio) had moved forwards and we need to change stuff if we want to telluric to use supported versions of its dependencies...

My branch fixes some issues, but there are some failing tests which I am not sure what to do with (it seems to be some regression test related to fixup of rounding issues at the edges of windows. It breaks because rasterio had changed the way it rounds up its images, but I do not have the rasters that the old telluric issue was trying to fix, so cannot verify that fixing the code does not re-break them).
I will add more info on this later.

[enomis-dev]

Thanks for sharing @AmitAronovitch , i'll take a look asap

[AmitAronovitch]

Hi @enomis-dev ,

Adding some info regarding the failing test (there might be some more...)

The failing test is tests/test_georaster.py::test_crop_respects_rounding_precision (dump below),

This test seems to be a regression test, that @drnextgis had added after fixing some issue (probably gaps that used to happen at the edges of tiles). The comment refers to #311.
That PR had added precision=3 to the round_offsets() call. However, this parameter is ignored in more recent versions of rasterio. See this PR: rasterio/rasterio#2311, which was merged in rasterio 1.3 (around the same time of #311 - it seems like rasterio devs were dealing with similar issues to ours, but 1.3 was not released yet).

I do not know if running with newer rasterio will re-introduce the original problem of #311 or not, because the test does not test it directly (it uses dummy rasters, and only compares window sizes - not checking effects at the edges of tiled rasters, which I believe the original problem was).

>       assert r1_cropped.shape == r2_cropped.shape
E       assert (1, 2663, 2180) == (1, 2663, 2179)
E         
E         At index 2 diff: 2180 != 2179
E         
E         Full diff:
E           (
E               1,
E               2663,
E         -     2179,
E         ?       ^^
E         +     2180,
E         ?       ^^
E           )

And here are some details from the debugger:

(Pdb) p roi
GeoVector(shape=POLYGON ((271806.0179640717 1438839.4164977344, 271806.0179640717 1519170.2272236191, 337216.40041283204 1519170.2272236191, 337216.40041283204 1438839.4164977344, 271806.0179640717 1438839.4164977344)), crs=EPSG:32628)
(Pdb) p r1.image.shape
(1, 4289, 3517)
(Pdb) p r1.affine
Affine(30.0, 0.0, 231681.0,
       0.0, -30.0, 1567951.0)
(Pdb) p r2.image.shape
(1, 4289, 3517)
(Pdb) p r2.affine
Affine(30.0, 0.0, 231669.0,
       0.0, -30.0, 1567970.0)

first case in crop:

(Pdb) p bounds, window
((1337, 1626, 3517, 4289), Window(col_off=1337, row_off=1626, width=2180, height=2663))

second case:

((1338, 1626, 3517, 4289), Window(col_off=1338, row_off=1626, width=2179, height=2663))