9

I have a file that has declarations in it with capital words, something like:

public final Foo BAR;

these declarations are not all uniform. Is there a way for me to grep out just the capital lettered words from the line? I know how to find lines that have capital letters in them, but thats not what I want. I want only the matching capital lettered word. Thanks!

1
  • 3
    Have you considered running ctags on your source code and parsing that instead? You will avoid a lot of false-positive matches this way. For example, the solution you ask for will also show capitalized words in comments. Nov 7, 2012 at 20:30

3 Answers 3

15

Actually with the -o switch, GNU grep returns only what it has matched. How about:

grep -oP "\w*[A-Z]+\w*" yourfile.txt

Note that this regex will match words with capital letters anywhere in them, not necessarily the beginning. You should tune it to meet your needs if they differ.

As pointed out below, this might not be the most portable of solutions. A portable alternative in Perl is

perl -nE 'say $1  while /(\w*[A-Z]+\w*)/g' yourfile.txt
7
  • That will match words that have a capital anywhere in the word, not necessarily at the beginning.
    – bahamat
    Nov 7, 2012 at 20:55
  • 1
    I am aware of that. The OP's example didn't specify capital letters in the beginning; in any case, by his use of grep, I was hoping he's comfortable enough with regex to use one more fine-tuned to his needs.
    – Joseph R.
    Nov 7, 2012 at 20:59
  • 3
    I wasn't saying that to imply that you were wrong, but as a clarification for people who don't parse PCRE with wetware.
    – bahamat
    Nov 7, 2012 at 22:13
  • 4
    Note that this option is GNU specific and will not work on several implementations of grep. Using sed is the only portable way I know how to do this.
    – rahmu
    Nov 7, 2012 at 22:59
  • 1
    Or, even more elegantly, perl -e 'print "$_\n" for map {m/(\w*[A-Z]+\w*)/g} <>' yourfile.txt
    – Joseph R.
    Nov 7, 2012 at 23:20
6

you could use sed to only return the string and not the whole line

sed 's/.*\([A-Z]*\).*/\1/g' <file>

or

sed 's/[a-z]*\| //g' <file>
1
  • You need to add a -n switch to suppress the input to be echoed to the output. Apr 28, 2015 at 19:40
5

The POSIX Basic Regular expression standard (used by default in grep, vim, less, sed, etc.) uses \< and \> to signify word boundaries. This allows a letter to follow white-space, as well as non-alphanumeric characters like quotes, dashes, equal-signs, etc. Use the -o option to print each match on a new line, et voila:

grep -o '\<[A-Z][a-z]*\>' yourfile.txt

Again, you might need to have change the regexp to suit your needs. Maybe by allowing numbers or a second capital letter..? This does both..

grep -o '\<[A-Z][a-z0-9]*[A-Z][a-z0-9]*\>' yourfile.txt
2
  • Good one, thanks for the explanation on \< and \>!
    – delh
    Nov 9, 2012 at 0:45
  • No, \< is not in the POSIX spec of BREs. The link you provide is not to that spec but to some man page of some version of GNU grep. Feb 7, 2014 at 16:26

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .