I need to find a regular expression, that ignores certain characters for usage in the bib2bib tool. For example: I need to find any occurence of the word "muller". But also a string like ''Hello, my name is Michael M\"uller, how are you?'' or ''There is M\"{u}ller''

should be found.

Edit: I need this to work not only for "muller" but dynamically for every word.

  • Can you give an example of how you want to use the bib2bib tool? Mar 16, 2015 at 21:46
  • If none of the answers presented so far meet your needs, it might help if you gave a more thorough explanation and more comprehensive examples of what you want. For example, would you want to match "muLLer"? (If you want case-insensitive matching across the board, say so.) How about "Mull\"er"? If the search string is "rene", would you want to match "Ren\'e"? Mar 16, 2015 at 23:51

3 Answers 3


If you want to remove things like \" and \"{ and }, you will have to preprocess your input file with a tool like sed before feeding it into bib2bib.


 sed -e 's/\\"\{\|\\"\|\}// input.bib > input.bib.preprocessed

Or to specifically convert things like \"{u} into u:

 sed -e 's/\\"{\(.\)}/\1/' -e 's/\\"//' input.bib > input.bib.preprocessed
  • But doesn't the preprocessing actually prohibit the match - as the characters which the asker wishes to match have actually been removed?
    – mikeserv
    Mar 16, 2015 at 23:21
  • @mikeserv I thought he wanted to ignore the \" and so on to be able to match e.g. muller with M\"uller. Mar 17, 2015 at 7:39
  • But how does it match it, though? I mean if you do : printf %s\\n 'M\"uller"' | sed 's/[\"]//g' | grep -i muller you don't match M\"uller" w/ grep but Muller - and that's what it prints. It doesn't print M\"uller".
    – mikeserv
    Mar 17, 2015 at 7:42

A fully portable solution could look like:

';printf %s\\n muller wright dummy >/tmp/patterns
tr '[:lower:][:upper:]' '[:upper:][:lower:]' </tmp/patterns |
paste '-d\n\n' - /tmp/patterns |
sed "N;s/./\\$n&/;:ul$n s/\(\n\)\(.\)\(.*\n\)\(.\)/\2\4\1\3/;tul"'

The output from that last sed looks like:


It would depend on the contents of patterns being only alphanumeric characters. If patterns contained, for example, either of [] it would require further testing to ensure that the square brackets were placed correctly each within their respective bracket expressions.

In any case, based on the example in question:


...is a regexp that will match a line containing any of muller or Muller or M"ulL\\\{"er.

With GNU sed you can handle the case conversions within sed itself, so:

sed -E 's/([[:upper:]]?)([[:lower:]]?)/\1\L\1\2\U\2/g' patterns



...fully fleshed out, you can get the same behavior as the previous tr|paste|sed combination (except that, this way, the aforementioned square-bracket problem is handled correctly) with just GNU sed like:

sed -E '
' </tmp/patterns

You haven't mentioned in what way you have your data available. To remove lines containing the posted patterns you can use grep:

grep -v -E '(muller|M\\"uller|M\\"{u}ller)'

(Note that the \ needs another escape.) To match the inverse, lines with the given patterns, omit the -v.

To define the regular expressions in a file use grep's option -f, as in:

grep -v -E -f file-with-regexps

It expects one regexp per line in that file.

  • Thanks for your reply. But this only works for muller. I need it to work dynamically. The data is in a file which is parsed with another tool (bib2bib), which accepts regular expressions.
    – User133713
    Mar 16, 2015 at 21:38
  • If you can't generalise from the suggestion please provide an example data file
    – roaima
    Mar 16, 2015 at 21:44
  • @user133713 No, it works for the three words that you asked for.
    – Janis
    Mar 16, 2015 at 21:52
  • @user133713 If that "bib2bib" tool has an own regexp parser you should better consult the documentation (or man page) of that tool. The solution provided uses standard Unix extended regexps.
    – Janis
    Mar 16, 2015 at 21:55
  • @User133713 Could you clarify what the process is you are working on? Please edit your question... Mar 16, 2015 at 21:58

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.