Create CITATION.md

[ivalaginja]

Addresses #24036 by adding a CITATION.md file to the repository.

I copied the citation instructions from here: https://pandas.io/about/citing.html
and added a section for the published software on Zenodo. Following this recommendation, I did this for the latest released version (v1.0.1) with a note for the user to go fetch the citation from Zenodo for the version they are actually using.

Tagging @TomAugspurger and @jreback since you were active in the linked issue.

I still recommend to update the citation request on the pandas website directly (https://pandas.io/about/citing.html), as well as on the Scipy website (https://www.scipy.org/citing.html#pandas), on https://pandas.pydata.org/ and maybe provide a <package>.__citation__ variable as suggested here.

[TomAugspurger]

Is CITATION.md a standard place for this information?

[TomAugspurger]

How was this list determined? I'd say perhaps keep it as "Wes McKinney and the pandas development team" or just "The pandas development team". Curious what others think though.

[jreback]

yes this should be just the pandas development team

[ivalaginja]

This is how you guys released it on Zenodo. I will update it in the PR and if you wish, I think you can even go into Zenodo and adjust that so that people fetch it correctly when they get their BibTeX entry from there.

[ivalaginja]

Is CITATION.md a standard place for this information?

Yes, the format doesn't matter, but authors will look for either citation instructions on a website or a CITATION file on the repository. The format doesn't matter (I just find markdown more neat than a plain text file). STScI has released a recommended style-guide for software releases based on what the community usually uses, and it recommends including a CITATION file.

[datapythonista]

Thanks for the work on this @ivalaginja

@pandas-dev/pandas-core a discussion we probably had to have earlier... What is our preferred way of citing pandas? In the old website we've got the two papers, and I just copied them to the new, but instead of making the users decide one randomly, we should probably decide ourselves on what's the proper way to cite pandas, and give them just one option.

Is this zenodo thing the preferred way? Should we leave just that and drop the papers?

[datapythonista]

We probably don't need the logo stuff in this file, just the citing part

[datapythonista]

There are not two papers anymore, I think this comment is outdated.

[datapythonista]

Probably worth having the link to Zenodo, so users can click and go there.

[datapythonista]

Not sure, but may be instead of 1.0.1 we could have something like REPLACE_BY_USED_VERSION? Or may be we can simply remove the version and make things easier?

[datapythonista]

I'd remove this section.

[datapythonista]

I don't see numpy, scipy or matplotlib implementing a CITATION.md file, and they are much more academic than us. I'd update the website and avoid duplication if this is not a standard in the scientific Python world.

[JoKeyser]

The point is to make it easy for anyone searching for information how to cite it properly. There doesn't seem to be a standard yet; adding this information to the Readme would also do the trick.

[ivalaginja]

I fully agree with @JoKeyser. This is not a standard yet, and yes adding it to the readme would do the same trick. The fact that numpy, Scipy et al. don't have something like this is annoying to problematic, rather than leading the way.

[datapythonista]

I don't think adding a CITATION.md file in pandas is going to create a standard. And I don't think it's great having to maintain two copies of the same information.

There is a link "Citing pandas" in the navigation bar of the website. If you want to add a comment in the README linking to that page, since people may go to the GitHub and not the website to look for it, that sounds good. And if you want to improve the citation page, that's also welcome.

We can surely discuss it again if other projects are happy to adopt this as a standard, but I think at this stage this adds little value, and gives us extra work of maintaining duplicate information, and adds even more noise to our root directory.

[ivalaginja]

The whole point of this is not actually to create a CITATION file but to update the citation instructions in general because there is no mention at all about the actual software citation. Adding that information into the README works just as well, but the website definitely needs to get updated.

And you're right, pandas adding a CITATION file wouldn't set a standard - it would make it follow it. And yes, it would not be necessary if the existing places were up to date.

[datapythonista]

Ok, that sounds good. Closing this, feel free to open a PR to update the existing citing page https://pandas.pydata.org/about/citing.html

Thanks!