arxiv_to_bibtex is a simple commandline tool that takes one or more urls to the arXiv and returns bibtex for them.

This is strongly inspired by Sven Porst's arxiv2bibtex, which I used for several years. But I was looking for a programmatic way to handle this conversion recently and I saw that Porst's code is intermixed in a cgi script that runs the site.

I wanted a version that is leaner and simpler, and which separates the code for interacting with the arXiv API and utility. In practice, this is nothing more than a small wrapper around the arXiv API.

This repository contains two distinct implementations. The first is a python package, arxiv_to_bibtex, contained in ./src. The second is a standalone HTML+CSS+javascript version that contains the same logic, but in javascript form.

As this is not on the Python Package Index at the moment, it is first necessary to obtain a copy of the code. The easiest way to do this is by cloning the repository (or just the /src/arxiv_to_bibtex directory). The directory /src/arxiv_to_bibtex is a callable module. For example, with this directory in your current directory, you can call it with

python -m arxiv_to_bibtex -h

to see a commandline usage message. Alternately, you can clone the repository and then perform a local pip installation (pip install . from the cloned repository). If you do this, you will now have a command arxiv_to_bibtex.

In either case, the commandline utility has exactly one interface command. This is to do

arxiv_to_bibtex --url [ARXIVURL ..]

with either one or more urls of arxiv papers. For example,

arxiv_to_bibtex --url https://arxiv.org/abs/2007.14324 https://arxiv.org/abs/2204.01651

will give the output

@misc{hulse2020arithmetic,
      title={Arithmetic Progressions of Squares and Multiple Dirichlet Series},
      author={Thomas A. Hulse and Chan Ieong Kuan and David Lowry-Duda and Alexander Walker},
      year={2020},
      howpublished="\url{http://arxiv.org/abs/2007.14324v4}",
      note={arXiv:math.NT:2007.14324v4},
}
@misc{anderson2022improved,
      title={Improved bounds on number fields of small degree},
      author={Theresa C. Anderson and Ayla Gafni and Kevin Hughes and
              Robert J. Lemke Oliver and David Lowry-Duda and
              Frank Thorne and Jiuya Wang and Ruixiang Zhang},
      year={2022},
      howpublished="\url{http://arxiv.org/abs/2204.01651v3}",
      note={arXiv:math.NT:2204.01651v3},
}

(I added a couple of linebreaks for readability). These are ready for direct insertion into a bibtex database.

As we can observe, the rule for the citation key is to take the last name of the first author, the year, and then the first word of the paper. There is a small amount of normalization: some punctuation is stripped from last names (so my last name would become lowryduda), and "unimportant" words are omitted from titles. There is a short list of "unimportant" words in the source, and the program simply looks for the first word in the title that isn't unimportant.

The basic utilities are exposed in the code as a very simple library. There is one tool of interest (that fits into my own use case). Internally, there are commands to

• find an arxiv identifier from a url (e.g. extract 2204.01651 from https://arxiv.org/abs/2204.01651),
• query the arxiv API for a url (or really, to batch a list of urls into a single API call),
• to parse the arxiv API response into a custom dataclass called BibtexRecord, and
• format a BibtexRecord into a bibtex string.

If you wanted to alter the output of this code to produce a different bibtex string for the same record, it is only necessary to write a formatter function that takes a BibtexRecord and outputs the desired string; then pass this formatter as the second optional parameter to arxiv.arxiv_to_bibtex.

The default formatter is called bibtex_record.standard_bibtex_formatter, and can be used as a model.

(I do not expect anyone to ever this except for me. If you do, let me know!)

I should preface by saying that I'm great at javascript. Nonetheless, the task of querying the arXiv API and producing bibtex is very straightforward and there is no reason why users of arxiv_to_bibtex need their own server.

In /html_version/, there is a single html page (with mediocre CSS included) that contains a javascript version of arxiv_to_bibtex. Opening this page in your preferred browser (or serving the page with, e.g., python -m http.server and opening that page) opens a web version.

The only interaction in this web version is to ping the arxiv's API and parse the response, all client-side with javascript.

At some point in the future, I'll probably have a version of this page running.