conda-forge package

[VolkerH]

It would be nice to be able to install this library and the python-bindings via conda-forge.

[cheesema]

This indeed would be nice, and is something we will look at trying to do in the future.

[skalarproduktraum]

Probably @maweigert knows how to do that 👍

[cheesema]

That would be awesome?, @maweigert do you hold the magic keys?

[maweigert]

Hi,

Sadly, I'm not really experienced with conda-forge - I'll mostly distribute python packages directly via setuptools (pip), which might be doable in your case, too, e.g.

http://www.benjack.io/2017/06/12/python-cpp-tests.html

(Disclaimer: Haven't tried it though)

[MSusik]

If you manage to create a pip installable package, it would have a large advantage over conda-forge as it would be independent of anaconda. Some of the very C/C++ heavy packages like numpy did that, so it should be possible.

[cheesema]

So pip seems the correct direction, we should look into this.

[VolkerH]

pip installable would definitely simplify the python installation. I'd appreciate any effort to make it easy to install, be it pip or conda.

However, conda has the advantage that it is not limited to Python. I think @ctrueden has started some work to make ImageJ / Fiji installable via conda-forge.

[ctrueden]

I think @ctrueden has started some work to make ImageJ / Fiji installable via conda-forge.

We started a conversation about packaging Java components at conda-forge/conda-forge.github.io#590. But I doubt it is going anywhere any time soon. We now have jgo for downloading Java packages into Python on demand, and I don't have any other concrete use cases for making ImageJ or Fiji installable as an end-user application via conda. If you want to make a case for it though, then feel free to file an issue at https://github.com/imagej/imagej/issues/new highlighting some reasons it would help you.

[VolkerH]

Thanks for weighing in. Maybe that was the discussion that gave me ideas imageJ might be distributed this way. I guess I misinterpreted this. Sorry for the distration. In that case I don't have a preference either way.

[ctrueden]

Maybe that was the discussion that gave me ideas imageJ might be distributed this way.

No need to apologize. It also might have been this discussion:

https://github.com/conda-forge/staged-recipes/pull/7067#pullrequestreview-177561145

It was there that we decided to call the conda package pyimagej instead of simply imagej, in case we start shipping ImageJ itself (as an end-user application) via conda in the future.

[cheesema]

Just an update on this. @joeljonsson is working on a re-design of the wrappers and they can be installed using pip. Will make the code public once its in a shareable form.

We are focusing on using Python as a environment to play round with the LibAPR, with some custom viewers, and interactive generation.

[cheesema]

@joeljonsson

[cheesema]

There are a non-released version of PyLibAPR wrappers installable from: https://test.pypi.org/project/PyLibAPR-test/#files