Question about sort_OmittedBibRefMixedNumericStyle

[jgm]

Something I don't understand here.
The layout is

      <choose>
        <if match="none" type="personal_communication">
          <group delimiter=". ">
            <text variable="citation-number"/>
            <group delimiter=", ">
              <text macro="author" />
              <text variable="title"/>
            </group>
          </group>
        </if>
      </choose>

I would assume that this means that items with type personal_communication produce no output. But oddly, the expected RESULT includes the citation number:

<div class="csl-bib-body">
  <div class="csl-entry">1. Anderson, Book One</div>
  <div class="csl-entry">2. [CSL STYLE ERROR: reference with no printed form.]</div>
  <div class="csl-entry">3. Crane, Book Two</div>
</div>

See the 2. in the second entry. Why?
(Edit: the second entry has type personal_communication.)

[bdarcus]

I agree that looks wrong, and that the 2 div shouldn't be there (right?).

Only thing I can think of that would explain is if we're missing some spec language that would allow unprinted references?

Can @adam3smith or @fbennett confirm?

[adam3smith]

I think John's question isn't why the <div> element is there, but why it contains the citation number, which is inside the if condition.

I agree that logically this seems wrong. That said, I don't think the behavior for citations without a printed form is currently defined in the specs (right?) and I'm guessing Frank added the citation number for easier identification of the problematic item. I think to proceed we'd first want to specify the no-printed-form behavior (which exists to avoid terrible messes, such as items with invisible citations, bibliographies skipping citations numbers, etc.) and then decide if there should or shouldn't be a special rule for including citation numbers either way or not.

[bdarcus]

So we need language in the spec, ideally for 1.1, that addresses two cases:

1. nocite (but reference printed in bib)
2. noref (this one; cite, but no bib entry)

... and then to update this and any related test(s) accordingly?

Right?

[bwiernik]

Related issue: citation-style-language/documentation#74

I think we would need to handle three cases:

1. nocite: uncited items that are included in the bibliography (this is pandoc's nocite or if a user used Zotero's Edit Bibliography function to add an uncited item to the bibliography)
2. "noref": cited items that are excluded from the bibliography using a conditional (this case). This should be allowed as many styles require it.
3. "citation with no visible form": A cited item for which the citation (not bibliography) is empty. This should return an error. An item appearing in a citation must have some visible output or it can lead to problematic behavior for users, especially for word processor-based applications. This is different from nocite.

[jgm]

I guess the idea must be that in a numeric style, it would be confusing to omit an item altogether, since then the numbers would not be consecutive (EDIT: and there might be citations that refer to this number). So this is handled by a special case, even though it goes against the strict letter of the spec.

[jgm]

But in my opinion a better solution would just be to write the style differently:

        <group delimiter=". ">
          <text variable="citation-number"/>
           <choose>
             <if match="none" type="personal_communication">
               <group delimiter=", ">
                 <text macro="author" />
                 <text variable="title"/>
              </group>
            </if>
          </choose>
        </group>

[bwiernik]

@jgm See my comments here: citation-style-language/documentation#92

[bwiernik]

But in my opinion a better solution would just be to write the style differently:

I agree. Issues about processing noref citations aside, this test seems to be written wrong.