This script takes a .bib file as argument and makes several changes to that file. The changes are designed to make the file nice biblatex according to my notion of nice. The changes are made in place if a file is given as the --input argument; a backup is made first. Otherwise the script reads from STDIN and writes to STDOUT.

The script can be run on a file like this:

$ python3 ./convertbibliography.py --input <file>

The script depends on the bibtexparser (0.6.1) and titlecase (0.7.2) modules.

The file CB_customs.py contains several functions that are used by bibtexparser. These are not all used by my script, but I have left them in the file in case they are useful to somebody.

The script prints messages about what it is doing if it is called with the --verbose flag.

The script makes the following changes:

• Author and editor names are put in LastName, FirstName order
• LaTeX non-ASCII characters are converted to unicode
• If a record doesn't have a citekey then one is added of the form 'Foo' + a unique numeral
• Spaces are removed from citekeys
• Unicode encoding is set
• Titles are put in Title Case
• 'booktitle' fields are removed
• 'language' fields are removed where the language is English; if there is a 'language' field a 'langid' field is set too
• 'journal' fields are replaced by 'journaltitle'
• Some philosophy journals that are often missing definite articles have them added
• Books and collections have 'pages' fields removed
• Dashes are replaced with double hyphens ('--') in 'pages', 'volume', and 'number' fields
• 'pages' fields have 'pp.', 'p.' removed
• Abstracts are removed
• ISBNs and ISSNs are removed
• EPUB identifiers are removed
• 'copyright' fields are removed
• 'publisher' fields are removed from articles
• 'link' fields are removed
• '&', '%', '_' are escaped
• Ampersands ('&') in titles are replaced by 'and'
• Several fields used by JSTOR and CiteULike are removed
• 'edition' fields are converted from ordinals to numerals
• If a book or collection has a volume number its type is changed to mvbook or mvcollection
• if 'and' occurs in a 'publisher' field it is protected
• The URL for a resolver is removed from 'doi' fields
• Keywords are removed
• Entire fields that are protected have their protection removed
• Quotes are replaced with an appropriate smart single quote
• If there is a colon (':') in a 'title' or 'journaltitle' field it is deleted and the string after it is removed and placed in a 'subtitle' or 'journalsubtitle' field as appropriate
• 'series' fields are removed
• If an article lacks a doi field then an attempt is made to get one through the CrossRef API (the --no-doi option suppresses this)
• 'month' fields are removed
• 'numpages' fields are removed
• 'eprint' fields are removed
• 'year' fields are converted to 'date'
• 'issue' fields are converted to 'number'
• 'leading zeroes' are removed from fields with numbers