[Nix-dev] Python: getting rid of PYTHONPATH in Nixpkgs

[Thomas Hunger]

I played with `withPackages` and it's rather nice. The only problem I had
was "collision between A and B" style errors. Is there any plan to allow
for collisions?

On 2 November 2016 at 06:05, Dmitry Kalinkin <dmitry.kalinkin at gmail.com>
wrote:

>
> > On 01 Nov 2016, at 11:56, Freddy Rietdijk <freddyrietdijk at fridh.nl>
> wrote:
> >
> >
> > > To solve 2) you could also make a wrapper that combines selected
> outputs of the multiple output package without python itself. I think, this
> is what texlive does. That way all the burden of wrapping will be limited
> only to the multiple output packages.
> >
> > I'm not sure whether I understand what you mean. What kind of wrapper do
> you mean? What does it wrap, and what does it set? The problem with the
> multiple outputs is that when you would add each to PYTHONPATH, the extra
> outputs won't be importable unless the 'main' output is first. I don't see
> how that can be solved. One that sets PYTHONPATH? Or some kind of package
> I had in mind a simple ```combine [blah.foo blah.bar]``` that would
> produce a derivation with symlinks to both $foo/lib/python2.7/blah/foo and
> $foo/lib/python2.7/blah/bar, so multiple outputs are combined back and can
> be importable with the current envHook setting the PYTHONPATH. This is
> similar to the current python wrapper that additionally throws in the
> python interpreter itself along with wrapper to set PYTHONHOME.
> >
> > > As for the first point, I couldn’t very well understand what the
> problem is. Are you talking about conflict between instances of a package
> that were built against different version of the python? Then shouldn’t
> `--prefix` override the incompatible version, since, as you said, order
> matters?
> >
> > Example with discussion: https://github.com/NixOS/nixpkgs/pull/17749
> >
> >
> > On Tue, Nov 1, 2016 at 4:12 PM, Dmitry Kalinkin <
> dmitry.kalinkin at gmail.com> wrote:
> >
> > > On 01 Nov 2016, at 06:22, Freddy Rietdijk <freddyrietdijk at fridh.nl>
> wrote:
> > >
> > > Hi,
> > >
> > > Currently we use PYTHONPATH a lot in Nixpkgs to let applications and
> the interpreter find Python modules. This typically works fine, but there
> are problems with this method and so I would like to get rid of it and use
> only `python.withPackages` which uses `python.buildEnv`.
> > >
> > > The two main issues with the use of PYTHONPATH:
> > >       • Before we used `--prefix $PYTHONPATH`, but this would leak
> PYTHONPATH into subprocesses, which is especially problematic when both
> parent and child depend on Python but of different versions.  `--set
> $PYTHONPATH` would solve that issue, but that makes it impossible to add
> other (impure) paths, which is an important feature. This also breaks
> alternative shells like `ipython`. This issue is currently solved by
> extending `sys.path` in the Python applications.
> > >       • Limits the use of multiple outputs. When moving a module of a
> package into a separate output it becomes problematic to load this again,
> since just adding the module to PYTHONPATH typically doesn't work because
> the order matters. While Python modules are typically small, and build
> fast, rebuilding can take a lot of time in cases like `matplotlib` which
> supports multiple backends and is depended on by quite some packages.
> > > A method that is more reliable is building an environment with
> symbolic links to all the modules, and wrapping the applications with a
> wrapper that sets PYTHONHOME. This is exactly what `python.buildEnv` does,
> and it solves both 1) and 2).
> > >
> > > `PYTHONPATH` is mainly constructed with the Python interpreter
> setupHook. It is used in `buildPythonPackage` for building the package, and
> after installing it is extended so the tests can run. Furthermore,
> `PYTHONPATH` is set by the `setupHook` when using `nix-shell` like
> `nix-shell -p pythonPackages.numpy`.
> > >
> > > I think we can get rid of the setupHook. For the building we can
> create an env. This would be able to support multiple outputs as inputs,
> but will be more expensive than setting PYTHONPATH. For the tests we do add
> the newly constructed package to PYTHONPATH; there's no way around it and
> it doesn't cause any problems either.
> > >
> > > The biggest impact will be on how `nix-shell` is used. It won't be
> possible anymore to use it as shown before, instead `nix-shell -p
> 'python.withPackages(ps:[ps.numpy])'` would have to be used. While it is
> possible to keep the setupHook (or use it as a shellHook) it will be
> unreliable in the case of multiple outputs.
> > >
> > > What do you think about this change? Do you see any problems with it?
> Any other suggestions?
> > >
> > > Freddy
> > > _______________________________________________
> > > nix-dev mailing list
> > > nix-dev at lists.science.uu.nl
> > > http://lists.science.uu.nl/mailman/listinfo/nix-dev
> >
> > Hi,
> >
> > I’m far from being an expert in python, but here are my thoughts:
> >
> > To solve 2) you could also make a wrapper that combines selected outputs
> of the multiple output package without python itself. I think, this is what
> texlive does. That way all the burden of wrapping will be limited only to
> the multiple output packages. As for the first point, I couldn’t very well
> understand what the problem is. Are you talking about conflict between
> instances of a package that were built against different version of the
> python? Then shouldn’t `--prefix` override the incompatible version, since,
> as you said, order matters?
> >
> > Also, if you remove setupHook, won’t you also lose ability to specify
> python dependencies directly in buildInputs? Put that inside mkDerivation?
> >
> > Dmitry
> >
>
> _______________________________________________
> nix-dev mailing list
> nix-dev at lists.science.uu.nl
> http://lists.science.uu.nl/mailman/listinfo/nix-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.science.uu.nl/pipermail/nix-dev/attachments/20161102/486c06e1/attachment.html>