Prevent duplicate labels in :alpha style

[goerz]

This implements a stateful AlphaStyle that disambiguates the labels for the bibliography by appending a suffix "a", "b", etc. This emulates the behavior of the numeric citation style in LaTeX.

The existing :alpha style is automatically upgraded to the new AlphaStyle. It's a lot easier for users to pass style=:alpha than to instantiate the full AlphaStyle, in particular because AlphaStyle needs access to all the entries of the .bib file (so instantiating in manually would end up reading the .bib file twice). Moreover, there's no reason anyone would ever want to use the old "dumb" :alpha style.

In addition to the new AlphaStyle, this also fixes some inconsistencies relative to the alphabetic citation style in LaTeX:

• Up to four names are included in the label, not 3 (but 5 or more names results in 3 names and "+")
• The first letter of a "particle" of the name is included in the label, e.g. "vWB08" for a first author "von Winckel"

Closes #26

[goerz]

There's a preview at https://juliadocs.org/DocumenterCitations.jl/previews/PR31/gallery/#alphabetic_style

[goerz]

I'm going to mark this as "Draft" for now, since #29 convinced me that I'll probably want to tweak the arguments for the internal functions a little bit, and pass around the full CitationBibliography object (the instantiated plugin). The plugin object contains lots of state: all entries in the .bib file, all citations, which order they were cited in. The lack of access to this state in the various format_* routines already made AlphaStyle a little cumbersome, and implementing #29 or the most general version of #13 would likely be even more cumbersome.

So I'll probably make the changes to the internals and rebase this PR on top of that.

[goerz]

#29 convinced me that I'll probably want to tweak the arguments for the internal functions a little bit, and pass around the full CitationBibliography object (the instantiated plugin).

It turns out there's a more elegant solution that doesn't require breaking backwards compatibility at all (not that the internal functions are part of the stable API; but I'd still rather not change them unless absolutely necessary):

I've added a new internal function init_bibliography!(style, bib) that is called at the beginning of the ExpandBibliography pipeline step. For the default :numeric style, this function does nothing, but for the revised :alpha/AlphaStyle(), this is where the unique keys for all entries in the .bib file are calculated and cached in an internal attribute of AlphaStyle.

It's a much cleaner implementation for the "smart" :alpha style than the previous iteration, and it also addresses #29.

This should now be ready for a final review and merge, and then I'll follow up with #32

[kellertuer]

Nice, this looks good to me, though I can not for a 100% review the internal code – the functionality seems to work very nicely.

[goerz]

@fredrikekre Can I go ahead and merge, or did you want to further review this?

Also, since we didn't explicitly talk about this when QuantumCitations got merged back into DocumenterCitations and moved here: Should I still consider myself the primary maintainer of this project, as a direct continuation of QuantumCitations?

If yes, I would be a bit more direct in pushing the further development:

• I'll make general design decisions if they come up while dealing with PRs. I'll review and merge PRs.
• I'll organize tooling and CI mostly according to my needs and may push trivial and tooling-related commits directly to master
• I'll make any non-trivial changes via PRs and give people at least a week for review, but I'll ultimately merge my own PRs if there's no feedback. For any major changes or if I'm responding to feature requests, I'll actively solicit reviews/approval.
• I'll tag new releases and manage release notes, following semver.

None of this precludes any co-maintainers, i.e. you/anyone with push access to the repo from also doing any of those same things. In any case, especially with the transition to Documenter 1.0, I'll be coordinating with you and @mortenpi via comments here or on the #Documentation Slack channel.

[goerz]

If there are no further comments, I will merge this tomorrow