Command to add a dependency

[sonotley]

Heavier-weight alternatives to hatch such as Poetry and PDM have an add or install command for dependencies which:

1. Performs dependency resolution
2. Installs the dependency and it's dependencies into the environment
3. Adds the dependency to pyproject.toml with version specifiers based on step 1

For the tools cited above, this command is necessary because it invokes the built-in dependency handling etc. However, I think it would be useful if hatch provided an equivalent that simply wrapped pip install for steps 1 and 2 then performed step 3 based on the version selected by pip (or equivalent logic).

In my opinion, this offers a more elegant workflow than having the user manually complete pyproject.toml and either research appropriate minimum versions or initially specify * then go back and copy and paste the version that is actually installed.

[ofek]

I might consider this when lock file support lands, if and only if tomlkit is no longer buggy for me.

[merwok]

could you share specifics about tomlkit bugs?

[ofek]

I've encountered:

• Critical data loss python-poetry/tomlkit#113
• v0.7.1 breaks mappings python-poetry/tomlkit#124

Also, I dislike its defaults like single-line arrays and forced parent tables like [project.entry-points] below:

[frostming]

Also, I dislike its defaults like single-line arrays and forced parent tables like [project.entry-points] below:

It only becomes a problem if you are going to build the whole pyproject.toml programmatically, like via CLI. IMO this will make the interface too complex and also difficult to design. I recommend most of the fields should still be hand-written.

[timotk]

Hi @ofek, are these issues still a problem for you? I tried your example in python-poetry/tomlkit#124 and it's working for me.

>>> list(options.items())
[('option', 'foo')]

[wdonofrio-micron]

Probably not recommended, and I'm new to both hatch & uv (coming from poetry), but the following seems to work:

Alter default env path location
config.toml

[dirs.env]
virtual = ".venv"

Alter env default to use uv and path .venv (since uv expects .venv)
pyproject.toml

[tool.hatch.envs.default]
path = ".venv"
install= "uv"

Execute hatch run uv add numpy
Results in the pyproject.toml update as expected.
hatch run python -m numpy --version runs as expected.

You could also probably configure uv to support the hatch env path instead.