Compacting `find` name patterns

Question

I am using

find . -name '*.[cCHh][cC]' -exec grep -nHr "$1" {} ';'
find . -name '*.[cCHh]' -exec grep -nHr "$1" {} ';'

to search for a string in all files ending with .c, .C, .h, .H, .cc and .CC listed in all subdirectories. But since this includes two commands this feels inefficient.

How do I write a regex to include .c,.C,.h,.H,.cc and .CC files using one single regex?

EDIT: I am running this on bash on a Linux machine.

Solution 1

As you (incorrectly – what you used is a shell pattern) mentioned it in the subject, you should use regular expressions:

find . -iregex '.*\.[ch]+'

The above is lazy approach, which will also find .ch, .hh and alike, if there exists. For exact matches you still have to enumerate what you want, but that is still easier with regular expressions:

find . -regex '.*\.\(c\|C\|cc\|CC\|h\|H\)'

Solution 2

Portably/standardly (POSIX, Unix (SUS) and Linux (LSB) standards) and efficiently, you'd write it:

find . \( -name '*.cc' -o -name '*.CC' -o -name '*.[cChH]' \) \
  -type f -exec grep -n -- "$1" /dev/null {} +

The most important point here is to use + instead of ;. Otherwise, you'll run one grep command per file.

The -H option is GNU specific, but adding /dev/null (which makes sure grep gets at least two files to look in) guarantees that grep displays the file name.

You'll need "--" unless you can make sure that $1 will never start with -.

Adding -type f here, to avoid looking into non-regular files (like directories), but as that means it also excludes symlinks, you may wish to leave it out.

Solution 3

Can be shortened to this single line:

find -type f -regextype posix-egrep -iregex '.*\.(cc|h|c)$' -exec grep -nHr "$1" {} \;

Comments

[rush]

By the way, you can use '+' at the end of find instead of ';'. It will accelerate the command due to shell will execute one grep per many files, not one grep per file as with ';'.

[manatwork]

Your regular expression is wrong. It says “any character 0 or more times, followed by one of the enumerated strings”. On my machine that finds a lot of .sh script files…

[daisy]

@manatwork right, updated the answer

[Stéphane Chazelas]

knitpicking here, but the above would match .cC or .Cc files which were not requested. Also note that the $ is not needed as GNU find's regexps are implicitely anchored.

[Richard Fortune]

-H versus /dev/null: very nice! FWIW, the -H option does seem to be widely available (I see it on FreeBSD 9, BusyBox grep, and Mac OS X grep).

[Stéphane Chazelas]

@dubiousjim AFAIK FreeBSD and MacOS/X greps are the GNU grep.

[Richard Fortune]

Oh yeah, you're right! That surprises me about FreeBSD. Most of their tools aren't Gnu. For instance, their sed and awk aren't.

[Arpith]

How is this different from using find . -name '.*\.\(c\|C\|cc\|CC\|h\|H\)' ?

[manatwork]

@Arpith, with -name you specify a shell pattern, with -regex you specify a regular expression. That '.*\.(c\|C\|cc\|CC\|h\|H)' string interpreted as shell pattern will rarely match anything, but certainly not what you intended in your question: pastebin.com/yhddCnbv