Add preferred-citation field to the top-level.

[hainesr]

As discussed in #54.

[hainesr]

So. preferred-citation is now in ruby-cff.

Next question is how to surface it in the BibTeX/APA outputs. If there is a preferred-citation, should that be the one that is output, over the top-level data? It feels like this is the expectation when providing a preferred-citation but would appreciate thoughts from @sdruskat, @jspaaks and indeed others!

[sdruskat]

Great!

I think that what GitHub want (correct me if I'm wrong, @arfon) is indeed to use the preferred-citation whenever there is one, so as to respect the users' preferences (although this goes against the principles somewhat 😉).

[hainesr]

You know what? I'll make it configurable, then everyone has options! 🚀

[jspaaks]

You know what? I'll make it configurable, then everyone has options! rocket

But wait isnt the fact that the user added a preferred-citation already the configuration?

[sdruskat]

I thought that making it configurable would be the best option as well, but didn't want to put the burden on Rob. I do think that having this configurable is good, because then, other users of the Gem (or GitHub in the future!) could choose to either follow the user, or follow best practice (see the principles paper, "always cite the software") or just display both 🙂.

[hainesr]

I can make it configurable with a default.

What should the default be? Use preferred-citation (for the reasons @jspaaks mentions above)?

[sdruskat]

Best option for now, defo!

[arfon]

What should the default be? Use preferred-citation (for the reasons @jspaaks mentions above)?

This makes sense to me. I've been thinking of preferred-citation as an override for any other behavior so I think if it exists, this is what GitHub should use in the UI.

[hainesr]

I have pushed this to main now, with the following features:

• When outputting a citation string, honouring preferred-citation is the default, if present, but can be turned off.
• The BibTeX formatter respects the type of reference being generated and tries to use a sensible corresponding BibTeX type e.g. @misc for software and @book, @article, @inproceedings etc as appropriate.
• The APA formatter only adds '[Computer software]' if it is indeed software being referenced.
• The formatters expect validated CFF files as input. I felt that this was better than validating within the formatter so that for e.g. GitHub things aren't being validated multiple times when outputting multiple formats at once.

Happy to have thoughts on these changes and issues raised where I've missed something.

[arfon]

⚡ thanks @hainesr. I've taken these changes for a spin and have found a few issues with some of the outputs for articles. I've opened a PR to demonstrate the issues here: #71

[hainesr]

Thanks @arfon. I'll look at your PR in detail tomorrow. For reference I used the fixtures that @mfenner provided in #67 for the APA format.

[arfon]

Thanks @arfon. I'll look at your PR in detail tomorrow. For reference I used the fixtures that @mfenner provided in #67 for the APA format.

👍 thanks @hainesr 🙇. From what I can see in #67, the examples are almost (all?) software citations, not articles (i.e., preferred-citation entities).

Ultimately it will be great to switch to CSL for formatting these outputs :-)

[hainesr]

Hi @arfon, yes now I've had chance to look at your PR properly I see the difference!

Thanks, we do need more test coverage in this area and we're missing some of the useful article level data in the citations at the moment. I'll try and look at this today.