Shapely 2.0 compatibility

[knthls]

I have made some modifications to replace PyGEOS with shapely 2.0

PyGEOS was merged with shapely, which is why shapely offers exactly the same API as PyGEOS. I have basically renamed all imports from pygeos to imports from shapely and I have eliminated the additional parts that deal with shapely/pygeos compatibility. My modifications address issues #213 and #208.
I have tested my modifications an verified that they work, given that the environment variable USE_PYGEOS is set to 0.

Please Note that my modifications introduce a hard requirement for shapely 2.0 so it may not be ready to merge, depending on your policy regarding support of older libraries. I still hope that this contribution saves you some time.

[knthls]

Please let me know if I can do anything to help this getting merged. I guess the CI pipeline is broken.

python: /lib/x86_64-linux-gnu/libc.so.6: version `GLIBC_2.34' not found (required by /__t/Python/3.10.11/x64/lib/libpython3.10.so.1.0)
Error: Process completed with exit code 1.

I have applied black and flake8, but I think that it actually ignores the cython stuff.

[HTenkanen]

@knthls Thanks a lot for helping out!! 🙂 I've been frustratingly busy with my day job, and getting back to pyrosm seem to continuesly get pushed further on my todo-list. 🤷🏻‍♂️ Hence, all the help with maintaining and developing this further are highly appreciated! 🙏🏻

Indeed it seems that we need to update the tests.yaml workflow which is not up-to-date anymore. I'm quite busy the coming days, so don't have time to fix this until sometime next week, but in case you want to take a look at the CI file, I'd be happy to merge a PR that fixes the pipeline, and after that we can take a closer look at this PR.

[hbruch]

Thanks @knthls for your work! To get the CI fixed, #215 should solve the black/linting issues.
Running USE_PYGEOS=0 pytest -v on my Mac, I still had some test issues (some of them might be related to me not using conda, i.e. I have install issues with pandana/tables):

FAILED tests/test_graph_exports.py::test_igraph_export_by_driving - assert 51420 == 44296
FAILED tests/test_graph_exports.py::test_pdgraph_connectivity - ImportError: 'pandana' needs to be installed in order to export the network for it.
FAILED tests/test_graph_exports.py::test_to_graph_api - ModuleNotFoundError: No module named 'pandana'
FAILED tests/test_network_parsing.py::test_saving_network_to_shapefile - AssertionError: assert [None, None, ...ne, None, ...] == [nan, nan, na...nan, nan, ...]
FAILED tests/test_utils.py::test_timestamp_integer - TypeError: Cannot localize tz-aware Timestamp, use tz_convert for conversions
FAILED tests/test_utils.py::test_OSH_file_without_timestamp - TypeError: Cannot localize tz-aware Timestamp, use tz_convert for conversions
====================================================================================== 6 failed, 99 passed, 2 skipped, 37 warnings in 127.89s (0:02:07) ==

The tz-aware Timestamp issue could be solved by fixing unix_time_to_datetime in utils/__init__.py:

def unix_time_to_datetime(unix_time):
    return pd.Timestamp.utcfromtimestamp(unix_time)

Regarding test_igraph_export_by_driving I wonder, if the used dataset is a fixed dataset or a live dataset which surely will change from time to time.

The pandanas errors are due to pandana install issues, probably only on my side.

[knthls]

Thanks @hbruch. I didn't get into this, since I am really busy too.
The only issue that I could confirm is this one.

FAILED tests/test_graph_exports.py::test_igraph_export_by_driving - assert 51420 == 44296

I will try to get into this as soon as possible. It could really could be that I introduced some errors without noticing, since the only thing that I really changes were the imports.

[hbruch]

I don't think, these are introduced errors. The osm dataset for test_igraph_export_by_driving is not static and hence no assertions on exact number of features should be expected (or instead one of the static Helsinki datasets be used). @HTenkanen wsa there a specific reason, why ulanbator is used instead?

[tfardet]

I've been using this branch to make sure it worked and, as mentioned by @hbruch, the errors seem entirely unrelated to that PR.
Would it be possible to merge it?

[HTenkanen]

A big thanks for these 🙂

[HTenkanen]

The test errors are not related to this PR, so will fix them separately:

Thanks @knthls for your work! To get the CI fixed, #215 should solve the black/linting issues. Running USE_PYGEOS=0 pytest -v on my Mac, I still had some test issues (some of them might be related to me not using conda, i.e. I have install issues with pandana/tables):

FAILED tests/test_graph_exports.py::test_igraph_export_by_driving - assert 51420 == 44296
FAILED tests/test_graph_exports.py::test_pdgraph_connectivity - ImportError: 'pandana' needs to be installed in order to export the network for it.
FAILED tests/test_graph_exports.py::test_to_graph_api - ModuleNotFoundError: No module named 'pandana'
FAILED tests/test_network_parsing.py::test_saving_network_to_shapefile - AssertionError: assert [None, None, ...ne, None, ...] == [nan, nan, na...nan, nan, ...]
FAILED tests/test_utils.py::test_timestamp_integer - TypeError: Cannot localize tz-aware Timestamp, use tz_convert for conversions
FAILED tests/test_utils.py::test_OSH_file_without_timestamp - TypeError: Cannot localize tz-aware Timestamp, use tz_convert for conversions
====================================================================================== 6 failed, 99 passed, 2 skipped, 37 warnings in 127.89s (0:02:07) ==

The tz-aware Timestamp issue could be solved by fixing unix_time_to_datetime in utils/__init__.py:

def unix_time_to_datetime(unix_time):
    return pd.Timestamp.utcfromtimestamp(unix_time)

Regarding test_igraph_export_by_driving I wonder, if the used dataset is a fixed dataset or a live dataset which surely will change from time to time.

The pandanas errors are due to pandana install issues, probably only on my side.