GeezerVison

coverage: 94.784%. remained the same
when pulling ac5a4d1 on KybernetikJo:add_pandas_to_pyproject
into a8a54d1 on python-control:main.

I don't really see a big enough benefit of adding a single optional well known package with the same name as the package.

What's the effective difference between pip install control pandas and pip install 'control[pandas]'?

FTR, I am guilty of starting that kind of extra with Slycot in #678. Probably would not do it again.

What's the effective difference between pip install control pandas and pip install 'control[pandas]'?

You are right. There is no real difference. It was just for the sake of completeness.

What do you think about a pip install[full] or pip install[all]?

[project.optional-dependencies]
test = ["pytest", "pytest-timeout"]
slycot = [ "slycot>=0.4.0" ]
cvxopt = [ "cvxopt>=1.2.0" ]
pandas = [ "pandas>=1.5.0" ]
all = [
    "control[test]",
    "control[slycot]",
    "control[cvxopt]",
    "control[pandas]",
    ]

or

[project.optional-dependencies]
test = ["pytest", "pytest-timeout"]
all = [
    "control[test]",
    "slycot>=0.4.0" ]
    "cvxopt>=1.2.0" ,
    "pandas>=1.5.0" ,
    ]

https://hynek.me/articles/python-recursive-optional-dependencies/

I do not have strong feelings about it. I am OK not merging it.

I don't really see a big enough benefit of adding a single optional well known package with the same name as the package.

What's the effective difference between pip install control pandas and pip install 'control[pandas]'?

These are valid points, so I looked for other projects that use the pattern. I was able to find two well known packages:

• polars https://github.com/pola-rs/polars/blob/main/py-polars/pyproject.toml
• cvxpy https://github.com/cvxpy/cvxpy/blob/master/pyproject.toml

The reason could be

• to document all dependencies (also optional) in pyproject.toml
• to indicate that a single option is a valid installation pip install control[pandas]
• to ensure the right version of optional packages
  • pip install control[slycot] is actually pip install "slycot>=0.4.0" control
  • pip install control[pandas] would be pip install "pandas>=1.5.0" control

Remark:
Well, the benefits don't seem to be big enough. I will keep the PR open for a few weeks, and then I will close it.

Well, the benefits were not big enough.