pyproject.toml file is missing version constraints of dependencies

[bguillod]

Describe the bug
When installing climada with anything other than mamba or conda (e.g., uv in my case), the version constraints of climada's dependencies (e.g. geopandas>0.14,<1.0) are not considered and the built environment is hence not climada-compatible. Reason is, those are not aware of constraints put in requirements/env_climada.yml...
I am aware that conda is the main supported dependency management framework. However, wouldn't it make things easier to simply include not only dependencies but also their version constraints in pyproject.toml, thereby making it easier for people working with pip, uv or such alternatives that are faster than conda/mamba?

Expected behavior
Standard python package managers can install environments that include climada relatively easily.

Climada Version: 6.0.1, but this looks still applicable to the current develop branch.

System Information (please complete the following information):

• Operating system and version: MacOS 15.6.1
• Python version: not relevant
  (to obtain this information execute > import sys >print(sys.version))

[peanutfun]

@bguillod Thanks for bringing that up. We tried to avoid version constraints in the pyproject.toml because we definitely need them for Conda, and keeping both up-to-date is a bit of an effort. We also tried to have a single requirements file, that can be integrated in the pyproject.toml and be used by conda, but this failed, see the discussion at #1067. As an installation via uv or pip alone is not feasible with Climada, we currently don't want to use these tools for dependency version checks (in addition to Conda).

If you have a suitable environment ready for Climada, and want to install Climada without upgrading the dependencies to incompatible versions, you should be able to execute pip install --no-deps climada (uv probably as a similar option).

Do you have a suggestion on how to keep both pyproject.toml and env_climada.yml aligned without too much maintenance overhead?

[peanutfun]

@bguillod Could you give us some feedback?

Do you have a suggestion on how to keep both pyproject.toml and env_climada.yml aligned without too much maintenance overhead?

In particular, does this solve your immediate problem?

If you have a suitable environment ready for Climada [...] you should be able to execute pip install --no-deps climada (uv probably as a similar option)