Fetching BibTeX citations using rcrossref

Introduction

I was drafting a review recently where I wanted to cite 122 papers (the subjects of the review) in one reference. Ordinarily, I use the citation managers Zotero and Mendeley. I could have imported all 122 PDFs into Zotero and then output the citations in a BibTeX file. However, that felt like too much manual work so I searched for a more automated way.

I had a list of 122 DOIs and I wanted the BibTeX entry for each. My suspicion was that there would be an R package to automate this task. That is how I stumbled across the rcrossref package by Chamberlain et al. (2020).

Demonstration of rcrossref

rcrossref is on CRAN so you can install it in the usual way:

install.packages('rcrossref')

Let me immodestly demonstrate the package by fetching a BibTeX entry for one of my own papers (my first, in fact). You simply call the cr_cn function with the DOI you seek:

library(rcrossref)

doi <- '10.1186/s12874-017-0381-x'
refs <- cr_cn(dois = doi, format = "bibtex")

The citation can be returned in one of many formats. As stated, I want BibTeX.

Let’s see the result:

writeLines(refs)

## @article{Brock_2017,
##  doi = {10.1186/s12874-017-0381-x},
##  url = {https://doi.org/10.1186%2Fs12874-017-0381-x},
##  year = 2017,
##  month = {jul},
##  publisher = {Springer Science and Business Media {LLC}},
##  volume = {17},
##  number = {1},
##  author = {Kristian Brock and Lucinda Billingham and Mhairi Copland and Shamyla Siddique and Mirjana Sirovica and Christina Yap},
##  title = {Implementing the {EffTox} dose-finding design in the Matchpoint trial},
##  journal = {{BMC} Medical Research Methodology}
## }

Perfect.

The bulk solution

Having fetched one example successfully, I imagined a viable solution would be to write a for-loop. However, the solution I used was even simpler than that. You can send in a vector of DOIs via the dois parameter. Let me add another of my pivotal world-changing research papers to illustrate the point:

doi <- c(
  '10.1186/s12874-017-0381-x',
  '10.1016/S2213-2600(20)30033-3'
)

refs <- cr_cn(dois = doi, format = "bibtex")

The refs object is now a list of two BibTeX entries. Writing those references to a text stream is slightly more awkward than before:

writeLines(unlist(lapply(refs, paste, collapse=" ")))

## @article{Brock_2017,
##  doi = {10.1186/s12874-017-0381-x},
##  url = {https://doi.org/10.1186%2Fs12874-017-0381-x},
##  year = 2017,
##  month = {jul},
##  publisher = {Springer Science and Business Media {LLC}},
##  volume = {17},
##  number = {1},
##  author = {Kristian Brock and Lucinda Billingham and Mhairi Copland and Shamyla Siddique and Mirjana Sirovica and Christina Yap},
##  title = {Implementing the {EffTox} dose-finding design in the Matchpoint trial},
##  journal = {{BMC} Medical Research Methodology}
## }
## @article{Middleton_2020,
##  doi = {10.1016/s2213-2600(20)30033-3},
##  url = {https://doi.org/10.1016%2Fs2213-2600%2820%2930033-3},
##  year = 2020,
##  month = {mar},
##  publisher = {Elsevier {BV}},
##  author = {Gary Middleton and Kristian Brock and Joshua Savage and Rhys Mant and Yvonne Summers and John Connibear and Riyaz Shah and Christian Ottensmeier and Paul Shaw and Siow-Ming Lee and Sanjay Popat and Colin Barrie and Gloria Barone and Lucinda Billingham},
##  title = {Pembrolizumab in patients with non-small-cell lung cancer of performance status 2 ({PePS}2): a single arm, phase 2 trial},
##  journal = {The Lancet Respiratory Medicine}
## }

Having mastered that, however, writing the BibTeX entries to a file was a piece of cake:

myfile = file('~/bibliography.bib')
writeLines(unlist(lapply(refs, paste, collapse=" ")), con = myfile)

Job done? Not quite.

There was, in fact, one more challenge to overcome. To cite the papers in \(\LaTeX\), I would need the 122 citation keys, i.e. the identifying labels that come after the @article{. We can see that CrossRef, the service that converts the DOI into a BibTeX citation, labels papers using the pattern author_yyyy. I extracted the citation keys using a bit of RegEx and the stringr and purrr packages:

library(purrr)
library(stringr)

cite_keys <- refs %>% 
  map_chr(~ str_extract(.x, pattern = '\\@article\\{[A-Za-z_]*[0-9]+')) %>% 
  map_chr(~ str_extract(.x, pattern = '[A-Za-z_]*[0-9]+'))

cite_keys

## [1] "Brock_2017"     "Middleton_2020"

I am not great at RegEx so I imagine there is a better way of doing that, but my method worked. Importantly, the above will work with double-barreled names like name1_name2_yyyy. Finally, I pasted the contents of cite_keys into my tex file.

End product

This is the end result:

One hundred and twenty-two consecutive references compacted into one reference, as required.