consider using poetry in pytest

[veenstrajelmer]

• requirements.txt was already deprecated in deprecate requirements.txt #676
• export requirements.txt from poetry package makes it possible to only install certain dev/docs dependency groups and not the complete package: https://github.com/python-poetry/poetry-plugin-export
• Min reqs from pyproject.toml: Support lock dependencies with lowest solvable versions python-poetry/poetry#3527 (comment)
• check of minimal testbank indeed installs minimal versions
• check twine build process locally with: python -m build and twine check dist/*
• check pip install -e .

poetry deps format is completely different:

[tool.poetry.dependencies]
python = "^3.8"
#scipy<1.6.0 pip install fails in py39, is also 3 years old
scipy = ">=1.6.0"
#numpy 1.21 is EOL since june 2023
numpy = ">=1.22"
#contour colorbar on uniform values fails with matplotlib<3.7.0, several other features with matplotlib<3.6.0
matplotlib = ">=3.7.0"
#pandas<2.0.0 in case of py38 to avoid conflict with xarray<2023.3.0: https://github.com/Deltares/xugrid/issues/78#issuecomment-1597723955 (xarray 2023.1.0 is the latest py38 release, and py38 is still supported by dfm_tools)
#pandas>=1.4 is required by xarray>=2023.3.0
pandas = [
	{version = ">=1.4.0"},
	{version = "<2.0.0", python = "<=3.8"},
	]
#shapely<2.0.0 give "AttributeError: module 'shapely' has no attribute 'GeometryType'"
shapely = ">=2.0.0"
#geopandas<0.13.0 require shapely<2.0.0
geopandas = ">=0.13.0"
#fiona<1.9 does not contain sql where filter
fiona = ">=1.9"
#contextily<1.0.0 is from April 2020
contextily = ">=1.0.0"
#xarray<2023.4.0 conflicts with pandas<2.0.0 for resampling, only available for py39
#xarray<2022.6.0 did not find certain variable in dataset in test_data_map testcase. TODO: lower xarray/dask to 2022.6.0 when xugrid>0.6.5 is available
xarray = [
	{version = ">=2022.11.0"},
	{version = ">=2023.4.0", python = ">=3.9"},
	]
#dask is aligned with xarray, no particular reason
dask = [
	{version = ">=2022.11.0"},
	{version = ">=2023.4.0", python = ">=3.9"},
	]
#netcdf4<1.5.4 pip install fails in py39
netcdf4 = ">=1.5.4"
#bottleneck<1.3.3 pip install fails in py39
bottleneck = ">=1.3.3"
#xugrid 0.6.5 supports wider xarray version range. TODO: xugrid>=0.6.5 requires xarray>=2022.11.0 without explicitly stating this, will be solved in next version (https://github.com/Deltares/xugrid/issues/167)
xugrid = ">=0.6.5"
#cdsapi<0.6.1 is from Feb 2021 and does not support newer python versions
cdsapi = ">=0.6.1"
#pydap<3.4.0 is from May 2017 and does not support newer python versions
pydap = ">=3.4.0"
#pooch<1.1.0 do not have attribute retrieve
pooch = ">=1.1.0"
#hydrolib-core<0.5.1 does not yet contain 3D z-sigma mdu settings and many other essentials
hydrolib-core = ">=0.5.1"
#meshkernel<2.1.0 does not yet contain mesh2d_refine_based_on_gridded_samples and other essentials
meshkernel = ">=2.1.0"

Also, it seems to mess up pip install -e . but not thoroughly checked

groups of extras can be defined via extras: https://python-poetry.org/docs/pyproject/#extras
dev dependencies via: https://github.com/Deltares/HYDROLIB-core/blob/bf75f024e48b515d4cd7c93cd3f997de4fc5b1ef/pyproject.toml#L26

[veenstrajelmer]

Not for now, process was already cleaned and pyproject.toml is now the single source of truth for dependencies and build. Poetry would mainly reduce flexibility now. Maybe consider pixi in the future instead, after package manager options have consilidated a bit.