splitting up a larger .bib file into one .bib file per entry

Question

Hi I have a large library of entries, and I'd like to split it up so there is a single .bib file for each entry. This seems to be the easiest and least error-prone way of importing them into R (as opposed to doing all of this inside R)

so for example:

@Article{     king2012cem,
  author    = {Iacus, Stefano M. and King, Gary and Porro, Giuseppe},
  title     = {Causal Inference without Balance Checking: {Coarsened}
          Exact Matching},
  number    = {1},
  pages     = {1-24},
  source    = {Crossref},
  timestamp = {2021-06-10},
  year      = {2012}
}

@Book{        koskinen2012ergm,
  author    = {Johan Koskinen, Dean Lusher, Garry Robins},
  title     = {Exponential Random Graph Models for Social Networks},
  subtitle  = {Theory, Methods, and Applications},
  source    = {Crossref},
  year      = {2012}
}

would each get their own .bib file, ideally named after their citekeys, so king2012cem.bib and koskinen2012ergm.bib.

I've looked at the manual for bibtool and biber to no avail. I could probably fix some regex to do it, but I'd rather use a tried-and-tested tool for it.

Thank you,

Answer 1

If you're working in R, you might want to parse the data in the .bib file as a native R object, bypassing the need to create a file for each bib item. I've created a gist showing the R code for that here: https://gist.github.com/jmclawson/def66ac8635db9c6131c3f3ae092f6e5

And here's a screenshot of the resulting data.frame:

Answer 2

Ok I got the answer myself, here it is:

using bibtool and bib2bib (from the bibtex2html), I split it up, and as a bonus is the R code for making a data.frame with all entries.

the -- 'print(xxx)' in bibtool is very useful formatting from bibtool (see the manual p. 27), which makes it possible to format a .bib in a completely standardised way, making sure everything is as it should be. unfortunately, the standard option is to break lines with \n at 77 chars and use tabs for alignment, which is not useful for parsing into R (long abstracts, titles etc).

# first make list of all entries for the loop
bib2bib -oc all-entries-list.bib  --no-comment library.bib

# then loop thru them
for xkey in $(cat all-entries-list.bib); do
  bibtool '--select{"^'$xkey'$"}' -i library.bib -o path/to/dir/"$xkey.bib" -- 'print.use.tab = {0}' -- 'print.line.length = {999999999}' 
done

and here is the code in R

# wd and libs
setwd('some/path/to/bibtex-entries')
library(fs)
library(data.table)
library(purrr)

# get all entries
fil1 <- dir_ls('input/singles-entries/work-bibliotek-entries/', regex='\\.bib')

clean_entries <- function(.x) {
        # .x <- fil1$path[[72]]; .x     
        # .x <- fil1[[72]]; .x      
        # .x <- fil1$path[[36]]; .x     
    # read lines and remove non-fields 
    x1 <- data.table(v1=readLines(.x)) %>% .[v1 %ilike% '@|=']
    # split into variabel name and variabel content
    # x1[,  (c('field', 'value')) := tstrsplit(v1, '^=+?')]
    x1[,  (c('field', 'value')) := tstrsplit(v1, '=(?= {)', perl=T)]
    x1[field %flike% '@', `:=` (
                field = 'entry-type',
                value = gsub('^@([A-Za-z]+).*$', '\\1',  v1))]

    # remove curlies and ending , from last line 
    x1[, value := gsub('(^[ ])\\{(.*)', '\\1\\2',  value)]
    x1[, value := gsub('(.+)\\},*$', '\\1',  value)]

    x2 <- data.table(x1$value) %>% data.table::transpose()  
    setnames(x2, x1$field)  
    x2
     # names(x2)
    # v(x2)  
}

# use safely() to run tru all of them and capture output or error without stopping 
bib1 <- map(fil1, safely(clean_entries))
map(bib1, 'error') %>% keep(~ !is.null(.x)) 

# rbind the list, with the name of the list element as column called 'citeky'
bib2 <- rbindlist( map(bib1, 'result'), idcol='citekey', fill=TRUE)
# remove path from citekey column
bib2[, citekey := gsub('^.*/(.*).bib$', '\\1', citekey)]