I'm doing a lot of packaging for Python packages right now. A lot of
the code looks like this:
> (define-public python-execnet
> (package
> (name "python-execnet")
> (version "1.4.1")
> (source (origin
> (method url-fetch)
> (uri (pypi-uri "execnet" version))
> (sha256
> (base32
> "1rpk1vyclhg911p3hql0m0nrpq7q7mysxnaaw6vs29cpa6kx8vgn"))))
> (build-system python-build-system)
> (native-inputs
> `(("python-setuptools-scm" ,python-setuptools-scm)))
> (propagated-inputs
> `(("python-apipkg" ,python-apipkg)))
> (synopsis "Rapid multi-Python deployment")
> (description "Execnet provides a share-nothing model with
> channel-send/receive communication for distributing execution across many
> Python interpreters across version, platform and network barriers. It has a
> minimal and fast API targetting the following uses:
> @enumerate
> @item distribute tasks to (many) local or remote CPUs
> @item write and deploy hybrid multi-process applications
> @item write scripts to administer multiple environments
> @end enumerate")
> (home-page "http://codespeak.net/execnet/";)
> (license license:expat)
> (properties `((python2-variant . ,(delay python2-execnet))))))
>
> (define-public python2-execnet
> (package
> (inherit (package-with-python2
> (strip-python2-variant python-execnet)))
> (inputs
> `(("python2-setuptools" ,python2-setuptools)))))
As you can see, there is no inputs on python-execnet, so it's not
necessary to include the inputs. However, if (inputs) were added in the
future, this could lead to a developer mistakenly forgetting to change
the python2 variant.
What's the better approach?
- Chris
