Christiaan- Thanks for the response. I've clarified a couple of my points below and added a few additional comments, but I don't mean to be demanding about this at all. If these capabilities aren't appropriate for inclusion in BibDesk, I'll do my best to handle what's needed for an author-date version of CiteInPages in the script.
On Dec 31, 2007, at 11:07 AM, Christiaan Hofman wrote: ... > A problem of sorting support in templates is that it is 'static': you > have to hard code the possible sortings, there is no way to pass an > arbitrary sort key in a template. Also, it would be useless and > counter intuitive in templates used for export, as it overrides the > natural order from the items or the table. Having said this, it would > be possible to make a few sorting mechanisms (such as author search) > available, as we do in Skim. In fact I just added one. I think that the need is mainly for sorting bibliographies by authors and then by date as a secondary sort, so the number of keys involved should be limited. For scripts, this could be partly handled using the table sort, if there was a way within a script to trigger the table to sort by authors/date prior to extracting bibliographies. However, I didn't see anything in the BibDesk Applescript dictionary for table sorting. Author-date bibliographies have a bunch of low-incidence conditions that can be challenging, though. For example, if there's no author you want the reference to sort in the bibliography by the first field, which could be a company name, the title, or something else. To know what will be first in a particular bibliography, you need the template--or it's output--which is one reason I thought of doing the sorting as part of the templating process >> There are a couple of other edge conditions in author-date citations >> that turn out to be challenging in an Applescript implementation: >> >> 1. When first authors with the same last name and publication year >> but >> different first names appear in a bibliography list, the in-text >> citations should show the initials of both those authors. Citations >> for the rest of the references should show just the last names as >> usual. >> >> 2. When there are several publications in a year with the same first >> several authors, the in-text citations and the bibliographies should >> distinguish those publications by, for example, an "a," "b," "c," and >> so forth after the year. >> > > That's why a cite key is useful. I'd call it more a problem of the > approach of not using cite keys. I'm not sure I follow. CiteInPages uses cite keys directly as in-text working citations. In final formatting (after the manuscript is complete), the cite keys are used to look up references and are replaced by templated representations of these references for in-text citations. The bibliography is generated using a second template with the list of the unique cite keys in the manuscript. There's no problem targeting the correct references in working citations, in BibDesk or in the script using cite keys. The problem is that the templated output doesn't have a way to allow the reader to visually link in-text citations with the correct references in the bibliography list when there are overlapping initial author names and years. For example, APA specifies that in-text citations of references with "colliding" last names and years should be represented in special ways (1 and 2 in my previous note above). The template mechanism doesn't appear to cover these cases at the moment, which is understandable. This is also an issue that would occur with BibTeX/LaTeX; it must be handled as part of the overall manuscript formatting process there, working directly from the bib file. >> An alternative might be to address this issue within BibDesk. The >> complete idea would be to allow a list of publications to be >> extracted >> from BibDesk via Applescript in a particular primary and secondary >> sort order based on existing fields (authors and year, for example). >> Each publication would also include a templated citation string and >> bibliography string, based on citation and bibliography templates >> specified at extraction time. There would be a switch (or template >> tag) that allowed citation and bibliography strings to include >> additional designators such as initials or characters after the year >> to deal with the name conflicts noted above. >> > > As noted above, this is not possible in the templating mechanism. In > your script, the publications come from the text, so it's not > something extracted from BibDesk. So I don't really understand what > you mean here. I don't see anything here that BibDesk could/should > ever handle. The concept is that a script would read through the text document and gather a list of the unique cite keys. This list would be passed to BibDesk en bloc along with template names, and at that point BibDesk would have everything it needs to deal with the correct representation of that list of references. BibDesk currently can return a list of publication objects corresponding to a cite key list. These publication objects have a styled text attribute that contains a text representation of its reference that seems to be based on the view defined for the preview (I may be wrong on that--if so, sorry). A publication object could have two new attributes similar to styled text, CitationStyledText and BibliographyStyledText, that contained the text representations for in-text citations and bibliography entries, respectively. These attributes would acquire values on extraction, if appropriate templates were specified, and the values would not need to be stored by BibDesk. So, extraction of a set of references from a BibDesk library using Applescript might look something like this: set refList to get publications for myCiteKeys using myCitationStyle for CitationTemplate and using myBibStyle for BibliographyTemplate sort by BibliographyStyledText The returned list publication objects (in refList) would have individual citation and bibliography text representations in their corresponding attributes as specified by the templates, including the use of author initials or extra characters appended to the years when name/year collisions occurred. For this to work correctly, of course, you'd need to do the whole reference extract at once after assembling the cite key list from the document. As you point out, you couldn't do it piecemeal. In-text citations could then be replaced directly using cite keys to match to the correct publications and the bibliography could be built easily by cycling through the list in order. Since the templating system--particularly with the new editor--puts a lot of power in users' hands I'd prefer not to have to bypass it to deal with these author-date citation issues. However, I'll certainly understand if this isn't on the near-term road map for BibDesk and my comments here are just meant to clarify the initial suggestion. Best wishes, Jim Harrison UVa ------------------------------------------------------------------------- This SF.net email is sponsored by: Microsoft Defy all challenges. Microsoft(R) Visual Studio 2005. http://clk.atdmt.com/MRT/go/vse0120000070mrt/direct/01/ _______________________________________________ Bibdesk-users mailing list [email protected] https://lists.sourceforge.net/lists/listinfo/bibdesk-users
