added ipython.__bibtex__ attribute #6755
Conversation
|
I think we should think a bit more if we want that, or an tuple of citation, especially for big projects with many article. Also I'm not sure if the I also think that the data of the We should probably discuss of a format before jumping into this. |
|
I like the idea. So far I just found this info: http://ipython.org/citing.html |
|
Yes, but if ever we add things we need (at least) a list. And we should absolutely be coherent across packages. I saw that you opened similar PR on other packages. If we do that across scipy, we shoudl absolutely have the same structure everywhere, which require a bit of thinking about : 1 - is it a attribute or a function ? (function give os more flexibility, not sure if you can do properties module wide), and does it takes params ? Bibtex is nice, but hard to parse, should we have a |
|
Note, that I have seen the subject on Julia IIRC, where they wonder wether you like this even on a per-function basis, to know that a specific function implement an algorithm for a particular paper. I can see lots of project needing that too. |
|
This is a great idea! But does require some discussion.
This would be a great topic for discussion for a board of advisors for Scientific Computing... not to get bogged down in a "standards process" but for a group of people to quickly, efficiently make recommendations for broad use and adoption. I do think that Bibtex is the right format to choose, but maybe there should be a way for these to register themselves. Use case:
|
|
Lets not jump to far forward, and think at module level. Also here are the 3 other issues linked to that : scikit-learn/scikit-learn#3789 IMHO if should be stored as something more friendly and robust (eg, json), and |
|
i did cross post the following in the 3 other issues to try to regroup the discussion:
|
|
@jakevdp already did that on astroML : astroML/astroML@57fba48 |
|
I don't believe Bibtex would be the ideal format. It is from a pre-Internet age, so specifying links or other online resources is always kind of akward. Also, there is more than scientific literature. Technical reports often are referenced very differently (i.e. the author might be irrelevant). So I would pledge for a more flexible format, and as @Carreau suggested, an easier to parse one. |
|
Perhaps a citation could be in any one of a number of styles. That would suggest using JSON doesn't help all that much for Bibtex. For example, the title value could be "This is the title", or "{T}his is the {T}itle"... it will still need to be parsed. Perhaps Bibtex-string or Bibtex-JSON are possible formats. |
|
I am deeply skeptical of any format for citations which is not bibtex. It is a stable, well understood, non-proprietary, pure-text format. It has entry types for tech reports as will as articles, books, etc. It should also be injestable by any citation manager worth it's salt (yes, no true Scottsman but come on). At least the APS styles do a decent job of formatting links/eprints. |
|
I don't think the suggestion is to reinvent the bib format. It is no doubt a good format for scientific citations. I use it regularly. That said, it is not such a good format for internet and non-scientific resources citations. If some of the projects with the highest visibility in the scientific/engineering Python community implement a citation standard that is based on bibtex only, others will follow. I believe we should allow different formats. Say e.g. |
|
There is a json format for citations which at least zotero can export, and |
|
@Carreau I would also think that a language kernel would have citation information too. Another way to go about all of this might be to have a centralized repository (such as a github repo) of associated project names to citation entries. Combined with some server-side code, one could then get citation data for a named project (library, language kernel, etc). |
|
I'm going to close as every project is basically waiting for other project to decide what to do. Feel free to start a discussion on what format how should be done in a central place and reopen later if there is a concensus. Thanks ! |
|
Just putting a reference to code DOI here: |
|
mpl has (finally) merged our version. |
Added a bibtex attribute for quick command-line access to the recommended citation info. Idea by @jakevdp: https://twitter.com/jakevdp/status/524329573580697600