Package duecredit

[yarikoptic]

Is used (optionally) but a growing number of packages, including recently merged feedstock PyMVPA.

I am one of the upstream developers and only slowly picking up speed with conda recipes, so if someone could provide it before I do, would be very much appreciated. In general it should't be too difficult, it does install fine with pip usually (citeproc and lxml are the heaviest dependencies iirc)

[synapticarbors]

Needs packages for citeproc-py first, and that in turn needs rnc2rng as a setup requirement.

[mscheltienne]

Hello @yarikoptic! I can give a hand on packaging those on conda-forge, but I have some concerns.

rnc2rng is available on conda-forge: https://anaconda.org/conda-forge/rnc2rng No issues here.
However, citeproc-py is not and is not maintained a lot. The tests are run up to 3.9, and I am seeing issues with compatibility with Python 3.7+. The last commit is a call for maintainers.
As it is, adding an unmaintained package to conda-forge is not a great idea. Several options:

• get the maintenance of citeproc-py up and going again.
• don't depend on citeproc-py, and use a less general purpose code/package for duecredit.

@smoia For now I am removing duecredit from the requirements of nigsp in conda-forge #21409.

[smoia]

Hey there!
If it's not a problem to have a different requirement in the library setup vs conda-forge, I'm happy with that solution.

@yarikoptic, how problematic is it to drop citeproc-py from duecredit?

[yarikoptic]

IIRC citeproc-py provides crucial functionality for duecredit -- formatting of the entries.
I see tests failing for 3.9 in brechtm/citeproc-py#129 (I just merged it), but seems to be ok in master , 3.6 errored, removing it since EOLed anyways (brechtm/citeproc-py#130), removed.
There is no test-related fixes since 0.6.0 release and we are green there... @mscheltienne what do you mean by

and I am seeing issues with compatibility with Python 3.7+.

? I see no issues. ideally it should just get feedstock in conda-forge. https://anaconda.org/Aramislab/citeproc-py hints there is one elsewhere already.

[mscheltienne]

I was just looking through the list of issues, e.g. brechtm/citeproc-py#103

---

If I read the code from duecredit correctly, citeproc-py is only used in one function io.py in:

• https://github.com/duecredit/duecredit/blob/03451059445fbf30958165b156a547ae610b6114/duecredit/io.py#L281-L351
  To format BibTex files?

Maybe it would be possible to drop this dependency, either in favor of an internal code doing the formatting you need or of another BibTex library, e.g. bibtexparser (https://github.com/sciunto-org/python-bibtexparser) which is maintained?

---

Also, side note, while I was looking at the code for duecredit, my IDE picked up many unused variables and imports. Maybe a flake8 run/workflow could help improve this point.

[maxb2]

ideally it should just get feedstock in conda-forge. https://anaconda.org/Aramislab/citeproc-py hints there is one elsewhere already.

FYI, I added a citeproc-py feedstock in conda-forge a while ago.

[mscheltienne]

@maxb2 Oh nice! Not sure how I missed it!
Then we can create the recipe for duecredit and keep the lack of maintenance of citeproc-py as an issue on duecredit's issue tracker.

@yarikoptic @smoia Do you want me to create the recipe and list both of you as maintainers?

[yarikoptic]

I don't mind, thank you

[yarikoptic]

Also, side note, while I was looking at the code for duecredit, my IDE picked up many unused variables and imports. Maybe a flake8 run/workflow could help improve this point.

Yeah, lots of things could be improved. Project grew from rapid prototype at a hackathon and didn't see much action recently. PRs are welcome!