Re: QSH find with fgrep spawns many QZSHSH jobs -- MIDRANGE-L

OK. Looks like this is a winner:

QSH CMD('find -L ''/java'' -name ''*.java'' | xargs fgrep -il "SOMPFLM000"')

Thanks everybody.

PLA

Patrick L Archibald wrote:

Hey Scott

Thanks for all the info. I get "*qsh: 001-0085 Too many arguments specified on command.*" when using the fgrep command on a large directory. It works OK on smaller directories. With the '*find -L /java -name *.java -print | xargs fgrep -li "sompflm000*"' it runs OK interactively but I'm having some trouble submitting it. I'm not sure what is going on with that. I'll keep at it.

Thanks again, PLA
Hi Patrick,

I use the Qshell commands "find" and "fgrep" to search java source files for a string. It spawns a bunch of QP0ZSPWT jobs. Is there any environment setting or some other setting to stop it from spawning all these QZSHSH jobs?

Those jobs are where your "find" and "fgrep" commands are running. If you didn't spawn them, your commands wouldn't work.

[SNIP]

QSH CMD('find -L ''/java'' -name ''*.java'' -exec fgrep -il ''sompflm000'' {} \;')

Okay, here you've got a lot of jobs.
1) The first job is the one in which the QSH command is running.  It's
potentially receiving keyboard input and doing display output.  It has
spawned a "worker" job in the background to run "find".
2) The worker job with "find" is running as fast as it can searching your directory tree. As it runs, it sends output to the job that spawned it. Each time it finds a file that matches the pattern *.java it spawns another new job in which to run fgrep.
3) fgrep runs individually for each file found in it's own job.  It's
output is sent back up to the find command, who will then in turn send
it's output back to the QShell job.
This is done this way so that it acts like Unix -- which is what QShell is designed to do -- on a Unix system, a new process is created for each command. Though, Unix processes are somewhat more lightweight than OS/400 jobs are -- so the performance would be better on a real unix system.

I think there is a new QZSHSH job spawned for every file in the directories I'm searching. Is there anyway to stop this?
As I said above, you can't eliminate them completely.  You can, however,
minimize the number of jobs required.  fgrep is searching many different
files all in one run.  For example:
     cd /java/src
     fgrep -Ril "sompflm000" *
This will search every single file in the /java/src directory (-R makes it also search files subdirectories) for "sompflm000". Since there's only one invocation of "fgrep", it will only need 2 jobs (one for your terminal, and one more for fgrep) That's a far cry from the potentially thousands of jobs that you had when you were using "find."
The only problem with this statement is that it searches EVERYTHING (not
merely *.java).  That means that if there's a file that fgrep doesn't
understand, it might prematurely stop searching.
Another way to do it is to list the files individually. You can list many different filenames on the fgrep command:

fgrep -il "sompflm000" file1.java file2.java file3.java
And that works nicely enough, but obviously it's not always practical to
list every single file individually.
That's where the "xargs" Qshell command comes in. What "xargs" does is construct a command for you. It works by reading the command string that you give it as a parameter, and then appending all of the filenames that it reads from it's standard input, and the finally running the command.

For example:

find -L /java -name *.java -print | xargs fgrep -li "sompflm000"

the "find" command will, as before, find all of the filenames that match the pattern "*.java." It will print those filenames out to it's standard output.

The pipe will cause the stadnard output from "find" to be input to "xargs". So args will generate a command string that starts with "fgrep -li "sompflm000" and then it will append all of the filenames that "find" has printed.

This allows a total of 3-4 jobs to handle your entire request... one job for the terminal, one job for "find", one job for "fgrep" and I'm not sure if xargs requires it's own job, or if it shares one with fgrep. Either way, it's a far cry from the thousands of jobs that -exec would've generated, and will therefore run much quicker.
The only problem is that sometimes the command-line that xargs generates
will be too long for QShell to run. If that happens, you may have to go
back to the -exec method.
Hopefully you'll find this to be useful ;)

_______________________________________________ This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe, unsubscribe, or change list options, visit: http://lists.midrange.com/mailman/listinfo/midrange-l or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a moment to review the archives at http://archive.midrange.com/midrange-l.
_______________________________________________ This is the Midrange Systems Technical Discussion (MIDRANGE-L) mailing list To post a message email: MIDRANGE-L@xxxxxxxxxxxx To subscribe, unsubscribe, or change list options, visit: http://lists.midrange.com/mailman/listinfo/midrange-l or email: MIDRANGE-L-request@xxxxxxxxxxxx Before posting, please take a moment to review the archives at http://archive.midrange.com/midrange-l.