Shell Programming and Scripting

I still suspect breaking it up is the right answer.

You could remove the spool, use sed on the standard output to asterisk the passwords and write the same file name.

You can probably spool to a named pipe and have sed process that, too, but I am not a fan of named pipes. Even the <(...) and >(...) named pipes of bash may be named pipes in /var that accumulate, if you system does not have /dev/fd/3 and the like, or equivalent:

Code:

$ bash -c 'echo <( date )' 
/var/tmp//sh-np-1362160702
$ ls -l /var/tmp/sh-np-*
prw-------   1 my_id    develop          0 Mar  1 10:18 /var/tmp/sh-np-1362143650
prw-------   1 my_id    develop          0 Mar  1 10:18 /var/tmp/sh-np-1362160702
$ bash -c 'echo <( date )'                
/var/tmp//sh-np-1362140070
$ ls -l /var/tmp/sh-np-*  
prw-------   1 my_id    develop          0 Mar  1 10:18 /var/tmp/sh-np-1362140070
prw-------   1 my_id    develop          0 Mar  1 10:18 /var/tmp/sh-np-1362143650
prw-------   1 my_id    develop          0 Mar  1 10:18 /var/tmp/sh-np-1362160702
$

I use 'mknod p' named pipes when that is the only fitting tool that is left. Too often, some process is left hung trying to open one. You can check your named pipes for hung processes with fuser.

On those nice systems, I love <(...) and >(...) with a passion! The pipe() call creates 2 fd's, which exist in process's pseudo-file space as path names, so the shell can use these names like named pipes, but the process exit() cleans up these named pipes. If a child inherits the same fd #, it has the same path in that process's pseudo-file space, and if the parent closes his fd, it still exists in the child's /dev/fd/. You can create beautiful trees of pipes and processes and get pipeline parallelism, straight parallelism and no temp files, file name collision or temp space worries. Here is a nice parallelism, n parallel sorts into a quick merge:

Code:

sort -m <(sort a*) <(sort b*) ... | ....