Shell Programming and Scripting

rdcwayx: I realized it works better if you do fgrep instead of just grep... but thanks!

Yes, fgrep is faster and has no metacharacters.

Just a note, fgrep is deprecated and grep -F is now favored instead..

Did egrep get decapitated, too?

Yup, grep -E and so is rgrep...

Never got into rgrep, too GNU. Besides, I like to do that with parallelism:

Code:

find ... -type f -follow |xdemux 32 'nice -19 fxargs2 grep ...'

xdemux is a simple C program that opens $i popen()'s to write command $2 and writes lines in rotation. Did I rewrite something already there?

fxargs2 is 'xargs -n 1000' simplified, fortified and sped up, where each line is one arg and no command line is created = no quoting challenges, just direct fork and exec "$* files".

Since the grep is CPU intensive, I get to use all 16 CPUs and call the 32 sequences of grep commands with 1000 files or 256K each, whichever comes first. I usually set N to CPU cores x 2, since there is I/O and some grep commands get bigger files than others.

I suspect I am an unrepentant power user . . . .

Funny, that is my experience too that usually the sweet spot is at two threads or subprograms per core in a situation with IO. That my style too. No use letting all those expensive CPUs sit idly by. I am wondering about fxargs2. Since with xargs there are no quoting challenges either, so what could be improved?