My understanding is these lines should sort inline citations like [8, 9, 10, 11] but I have a manuscript where the citations appear [10, 11, 8, 9]

I wonder if this behavior is related to http://docs.citationstyles.org/en/stable/specification.html#sorting

cs:citation and cs:bibliography may include a cs:sort child element before the cs:layout element to specify the sorting order of respectively cites within citations, and bibliographic entries within the bibliography. In the absence of cs:sort, cites and bibliographic entries appear in the order in which they are cited.

If we do find a way to get citations to sort within a series of citations, would we want this as the default behavior? Or is it useful that citations remain in the order of the raw markdown?

@slochower I am not sure why the order in your manuscript is not [9, 10, 8, 11], because 8 is the only work that has been cited prior in the text.

I wonder if this behavior is related to http://docs.citationstyles.org/en/stable/specification.html#sorting

But we do have <sort> before <layout>.

@slochower I am not sure why the order in your manuscript is not [9, 10, 8, 11], because 8 is the only work that has been cited prior in the text.

Why wouldn't you expected 8 to be the first (followed by 9, 10, and then 11)? Maybe I'm misunderstanding the sorting.

The input is:

A complete description of the APR method has been characterized in the literature [@doi:10.1021/acs.jctc.5b00405; @doi:10.1002/jcc.23398; @doi:10.1007/s10822-014-9735-1; @doi:10.1021/ct5004109].

The output is:

A complete description of the APR method has been characterized in the literature [10,11,8,9].

However, this output corresponds to:

@doi:10.1007/s10822-014-9735-1; @10.1021/ct5004109; @doi:10.1021/acs.jctc.5b00405; @doi:10.1002/jcc.23398

So the input order is not preserved as I had assumed. Instead what I think is happening is that the rendered citations are being sorted as strings rather than numbers!

@adam3smith sorry to bug you, but any idea how we get our citation-numbers to sort as numbers rather than strings?

Instead what I think is happening is that the rendered citations are being sorted as strings rather than numbers!

Ah, good insight!

Wonder if it could be related to something with pandoc-citeproc not recognizing the strings as numeric: jgm/pandoc-citeproc#208

Yes, this appears to be the result of poorly written specifications and pandoc-citeproc reading them overly literally: citation-number is, for reasons I'm uncertain about, not listed as a number variable and should, as per the specs, therefore be parsed as a string during sort. That's obviously nonsense, though, and citeproc-js (and I assume most/all other processors) treat it as an integer when sorting. I'd appeal to John to change pandoc-citeproc's behavior, even though very technically it is correct.

cc @rmzelle who may know more on why this wasn't included as a number.

(to be clear -- I verified this issue with a local copy of pandoc)

Looping in @jgm. Is it intentional that pandoc-citeproc treats citation-number as a string and not an integer, and therefore sorts citation-number alphabetically?

citation-number is, for reasons I'm uncertain about, not listed as a number variable and should, as per the specs, therefore be parsed as a string during sort.

@adam3smith are you referencing these lines in the csl-data.json schema?

I was looking at the corresponding section in the specifications but that's just a different way of looking at the same thing, yes.