mjhjmtu

We have several find functions defined in our bash environment to exclude folders (often large or auto-generated) before grepping. An example of one of those is this:

function grepsrc()
c

Would using multiple -o -iname be faster than the -iregex?

function grepsrc()

 find . -type d ( -name .repo -o -name .git -o ) -prune -o 
 -type f ( -iname '*.h' -o -iname '*.c' -o -iname '*.cc' -o 
 -iname '*.cpp' -o -iname '*.S' -o -iname '*.java' -o 
 -iname '*.xml' -o -iname '*.sh' -o -iname '*.mk' -o 
 -iname '*.aidl' -o -iname '*.vts' ) 
 -exec grep --color=auto -n "$@" +

On my own tests, the former has an average time of

real 0m3.175s
user 0m3.021s
sys 0m0.145s

while the latter has an average of

real 0m3.170s
user 0m3.024s
sys 0m0.137s

So no real significant difference on my dataset, but I may be missing something.

There is no significant difference.

find is I/O bound, not CPU bound. Any string operation such as globbing or regexp matching will be dwarfed by disk operations. So your result above is to be expected.

What can (and typically does) affect find performance is the order of tests. For example if you're looking for directories, moving -type d before, say, -name tests can speed things up by telling find it doesn't need to look at files. But changes that affect only name matching don't have any significant effect on speed.