Comments on: Splitting Strings Natively with the Shell: Why

By: Luke Shumaker

Luke Shumaker — Wed, 02 Jun 2010 17:09:00 +0000

Sorry to burst your bubble, but:

sed -n ‘s@^$.*$:$.*$:$…*$:$.*$:$.*$:$.*$:/bin/nologin$@\1@p’ /etc/passwd

—Sed—

real 0m0.015s
user 0m0.000s
sys 0m0.004s

—Cut—

real 0m4.066s
user 0m1.336s
sys 0m2.004s

—Awk—

real 0m2.523s
user 0m0.576s
sys 0m1.716s

—Native—

real 0m0.066s
user 0m0.060s
sys 0m0.000s

Although, it wouldn’t have been possible if you’d specified a uid of, say, 15. I was able to rely on the fact that >10 = at least 2 digits for mine.

By: Катя

Катя — Wed, 27 Jan 2010 19:39:52 +0000

... Хм ...

By: Chris F.A. Johnson

Chris F.A. Johnson — Sun, 20 Dec 2009 05:25:29 +0000
That should be:
tr -d '\04'

By: Chris F.A. Johnson

Chris F.A. Johnson — Sun, 20 Dec 2009 05:24:18 +0000

tr -d ’4′

By: Brock Noland

Brock Noland — Thu, 10 Dec 2009 06:43:49 +0000

I do not believe there is a way to “filter” out the EOF.

By: Ben

Ben — Thu, 10 Dec 2009 06:26:53 +0000

Doh. I should have thought of using fifos…

Dumb suggestion: if you were totally desperate to use cut or awk in this way, could you convince tr to filter out EOF (that is, ^D = 0×04) to something less destructive? I wonder if tr is coded to allow this behavior. Just send a signal later to cut to get it to die.

By: Splitting Strings Natively with the Shell: Native vs Native

Splitting Strings Natively with the Shell: Native vs Native — Thu, 10 Dec 2009 04:57:42 +0000

[...] my previous post on why to split strings with bash itself, I used set to split the [...]

By: Brock Noland

Brock Noland — Thu, 10 Dec 2009 04:36:57 +0000

Yes, you are correct!

However I do not see how that would work. One thought would be to set them up to listen to a fifo, ensure they are line buffered, and have them ignore EOF, you could send them a single line and get a single response.

Like this:

# make fifo
mkfifo awk-input-username
mkfifo awk-output-username
awk -F: ‘{print $1} < awk-input-username > awk-output-username
….
# get username
echo $line > awk-input-username &
read username < awk-output-username

Of course they do NOT ignore EOF so the awk process will exit after reading a single line.

By: Ben

Ben — Thu, 10 Dec 2009 04:18:00 +0000

The overhead is most likely due to the system call overhead of spawning and starting (fork()ing and exec()ing) the awk and cut processes. It would be interesting to see how “already started” cut and awk compare with bash/IFS speedwise.