Your company here — click to reach over 10,000 unique daily visitors

ltx2crossrefxml - Man Page

create XML files for submitting to crossref.org


ltx2crossrefxml [-c config_file]  [-o output_file] [-input-is-xml]
               latex_file1 latex_file2 ...


-c config_file

Configuration file.  If this file is absent, defaults are used. See below for its format.

-o output_file

Output file.  If this option is not used, the XML is output to stdout.


Do not transform author and title input strings, assume they are valid XML.

The usual --help and --version options are also supported. Options can begin with either - or --, and ordered arbitrarily.


For each given latex_file, this script reads .rpi and (if they exist) .bbl files and outputs corresponding XML that can be uploaded to Crossref (<https://crossref.org>). Any extension of latex_file is ignored, and latex_file itself is not read (and need not even exist).

Each .rpi file specifies the metadata for a single article to be uploaded to Crossref (a journal_article element in their schema); an example is below. These files are output by the resphilosophica package (<https://ctan.org/pkg/resphilosophica>), but (as always) can also be created by hand or by whatever other method you implement.

Any .bbl files present are used for the citation information in the output XML. See the Citations section below.

Unless --rpi-is-xml is specified, for all text (authors, title, citations), standard TeX control sequences are replaced with plain text or UTF-8 or eliminated, as appropriate. The LaTeX::ToUnicode::convert routine is used for this (<https://ctan.org/pkg/bibtexperllibs>). Tricky TeX control sequences will almost surely not be handled correctly. If --rpi-is-xml is given, the author and title strings from the rpi files are output as-is, assuming they are valid XML; no checking is done. Citation text from .bbl files is always converted from LaTeX to plain text.

This script just writes an XML file. It's up to you to actually do the uploading to Crossref; for example, you can use their Java tool  crossref-upload-tool.jar (<https://www.crossref.org/education/member-setup/direct-deposit-xml/https-post>). For the definition of their schema, see <https://data.crossref.org/reports/help/schema_doc/4.4.2/index.html> (this is the schema version currently followed by this script).

Configuration File Format

The configuration file is read as Perl code. Thus, comment lines starting with # and blank lines are ignored. The other lines are typically assignments in the form (spaces are optional):

    $variable = value ;

Usually the value is a "string" enclosed in ASCII double-quote or single-quote characters, per Perl syntax. The idea is to specify the user-specific and journal-specific values needed for the Crossref upload. The variables which are used are these:

    $depositorName = "Depositor Name";
    $depositorEmail = 'depositor@example.org';
    $registrant = 'Registrant';  # organization name
    $fullTitle = "FULL TITLE";   # journal name
    $issn = "1234-5678";         # required
    $abbrevTitle = "ABBR. TTL."; # optional
    $coden = "CODEN";            # optional

For a given run, all .rpi data read is assumed to belong to the journal that is specified in the configuration file. More precisely, the configuration data is written as a journal_metadata element, with given full_title, issn, etc., and then each .rpi is written as journal_issue plus journal_article elements.

The configuration file can also define one Perl function: LaTeX_ToUnicode_convert_hook. If it is defined, it is called at the beginning of the procedure that converts LaTeX text to Unicode, which is done with the LaTeX::ToUnicode module, from the bibtexperllibs package (<https://ctan.org/pkg/bibtexperllibs>). The function must accept one string (the LaTeX text), and return one string (presumably the transformed string). The standard conversions are then applied to the returned string, so the configured function need only handle special cases, such as control sequences particular to the journal at hand.

Rpi File Format

Here's the (relevant part of the) .rpi file corresponding to the rpsample.tex example in the resphilosophica package (<https://ctan.org/pkg/resphilosophica>):

  %authors=Boris Veytsman\and A. U. Th{\o }r\and C. O. R\"espondent
  %title=A Sample Paper:\\ \emph  {A Template}

Other lines, some not beginning with %, are ignored (and not shown). For more details on processing, see the code.

The %paperUrl value is what will be associated with the given %doi (output as the resource element). Crossref strongly recommends that the url be for a so-called landing page, and not directly for a pdf (<https://www.crossref.org/education/member-setup/creating-a-landing-page/>). Special case: if the url is not specified,  and the journal is Res Philosophica, a special-purpose search url using pdcnet.org is returned. Any other journal must always specify this.

The %authors field is split at \and (ignoring whitespace before and after), and output as the contributors element, using sequence="first" for the first listed, sequence="additional" for the remainder.

If the %publicationType is not specified, it defaults to full_text, since that has historically been the case; full_text can also be given explicitly. The other values allowed by the Crossref schema are abstract_only and bibliographic_record. Finally, if the value is omit, the publication_type attribute is omitted entirely from the given journal_article element.

Each .rpi must contain information for only one article, but multiple files can be read in a single run. It would not be difficult to support multiple articles in a single .rpi file, but it makes debugging and error correction easier when each uploaded XML contains a single article.

More About Author Names

The three formats for names recognized are (not coincidentally) the same as BibTeX:

   First von Last
   von Last, First
   von Last, Jr., First

The forms can be freely intermixed within a single %authors line, separated with \and (including the backslash). Commas as name separators are not supported, unlike BibTeX.

In short, you may almost always use the first form; you shouldn't if either there's a Jr part, or the Last part has multiple tokens but there's no von part. See the btxdoc (“BibTeXing” by Oren Patashnik) document for details.

In the %authors line of a .rpi file, some secondary directives are recognized, indicated by | characters. Easiest to explain with an example:

  %authors=|organization|\LaTeX\ Project Team \and Alex Brown|orcid=123

Thus: 1) if |organization| is specified, the author name will be output as an organization contributor, instead of the usual person_name, as the Crossref schema requires.

2) If |orcid=value| is specified, the value is output as an ORCID element for that person_name.

These two directives, |organization| and |orcid| are mutually exclusive, because that's how the Crossref schema defines them. The = sign after orcid is required, while all spaces after the orcid keyword are ignored. Other than that, the ORCID value is output literally. (E.g., the ORCID value of 123 above is clearly invalid, but it would be output anyway, with no warning.)

Extra | characters, at the beginning or end of the entire %authors string, or doubled in the middle, are accepted and ignored. Whitespace is ignored around all | characters.


Each .bbl file corresponding to an input .rpi file is read and used to output a citation_list element for that journal_article in the output XML. If no .bbl file exists for a given .rpi, no citation_list is output for that article.

The .bbl processing is rudimentary: only so-called unstructured_citation references are produced for Crossref, that is, the contents of the citation (each paragraph in the .bbl) is dumped as a single flat string without markup.

Bibliography text is unconditionally converted from TeX to XML, via the method described above. It is not unusual for the conversion to be incomplete or incorrect.  It is up to you to check for this; e.g., if any backslashes remain in the output, it is most likely an error.

Furthermore, it is assumed that the .bbl file contains a sequence of references, each starting with \bibitem{KEY} (which itself must be at the beginning of a line, preceded only by whitespace), and the whole bibliography ending with \end{thebibliography} (similarly at the beginning of a line). A bibliography not following this format will not produce useful results. Bibliographies can be created by hand, or with BibTeX, or any other method.

The key attribute for the citation element is taken as the KEY argument to the \bibitem command. The sequential number of the citation (1, 2, ...) is appended. The argument to \bibitem can be empty (\bibitem{}, and the sequence number will be used on its own. Although TeX will not handle empty \bibitem keys, it can be convenient when creating a .bbl purely for Crossref.

The .rpi file is also checked for the bibliography information, in this same format.

Feature request: if anyone is interested in figuring out how to generate structured citations (<https://data.crossref.org/reports/help/schema_doc/4.4.2/schema_4_4_2.html#citation>) instead of these flat text dumps, that would be great.


  ltx2crossrefxml.pl ../paper1/paper1.tex ../paper2/paper2.tex \
                      -o result.xml

  ltx2crossrefxml.pl -c myconfig.cfg paper.tex -o paper.xml


Boris Veytsman <https://github.com/borisveytsman/crossrefware>