Filter bibliography file based on citations used in tex document

filter_bibliography.py

# coding: utf-8
"""
Created a filtered .bib-file from citations used in a specific .tex source file
"""

import pybtex
import pybtex.database
import pybtex.errors

import re

RE_CITATIONS = re.compile(r"\\cite[p|t]?\{(?P<ref>[\w,\s]+)\}")

def create_filtered_bibtex_file(fn_bibtex_in, fn_bibtex_out, needed_refs):
    # just issue warnings with duplicate entries
    pybtex.errors.strict = False

    bib_all = pybtex.database.parse_file(fn_bibtex_in)

    bib_new = pybtex.database.BibliographyData(
        {k: v for (k, v) in bib_all.entries.items() if k in needed_refs}
    )

    missing_refs = [r for r in needed_refs if not r in bib_new.entries.keys()]

    if len(missing_refs) > 0:
        print("Missing references:\n{}".format("\n".join(missing_refs)))

    bib_new.to_file(fn_bibtex_out)

def find_citations_in_latex_file(fn):
    with open(fn) as fh:
        file_content = fh.read()
        parsed_refs = RE_CITATIONS.findall(file_content)

    # regex doesn't split grouped refs, so we do that here
    all_refs = []
    for ref in parsed_refs:
        if "," in ref:
            all_refs += [s.strip() for s in ref.split(",")]
        else:
            all_refs.append(ref)

    return sorted(set(all_refs))

def main():
    import argparse
    argparser = argparse.ArgumentParser()
    argparser.add_argument('tex_document')
    argparser.add_argument(
        '--src-bibtex-path',
        default='$HOME/Documents/library.bib'
    )
    argparser.add_argument(
        '--bibtex-out', default='library.bib'
    )

    args = argparser.parse_args()

    needed_refs = find_citations_in_latex_file(args.tex_document)
    print("Found {} unique citations in `{}` ({})".format(
        len(needed_refs), args.tex_document, ", ".join(needed_refs)
        )
    )

    create_filtered_bibtex_file(
        fn_bibtex_in=args.src_bibtex_path,
        fn_bibtex_out=args.bibtex_out,
        needed_refs=needed_refs
    )
    print("Wrote new bibtex file to {}".format(args.bibtex_out))


if __name__ == "__main__":
    main()