Shell Programming and Scripting

Hi

My problem is that i need to measure given command statistics (time, mem, etc) with shell script
My script must basically do two things:
1. Measure overall command statistics (it is total time, maximum memory use during runtime, etc)
2. Measure command statistics in given periods of time during it's run (once a x seconds)
Then i have few questions:
1. what is the best method for checking command statistics during runtime (for example ps returns process time in seconds, but i need better resolution (are jeffes realible))
2. How to make gathering data during run and after that possible. my idea is that i test given command overall statistics with time command, and use some other command to test statistics during its runtime, but...
how to detect that command finished?
let's say that i use time for overall statistics, i have to do something like that
[time given_command...] &
part_of_script_testing_given_command_during_runtime
how to get pid of [time given_command...] in the main script?

and if you know some examples of such script, can you give some links?

if you know please help thx
Charles

You can get the pid of "time command" by grepping for "command" name, for example : "time cat" ; "pgrep cat" - this will return the PID of cat. Disadvantage : if you have multiple cat command, "pgrep" will return all of them.

Code:

for i in `seq 1 100`; do time ps -ef >> logfile.log; sleep 2; done

this will output the return of time command on STDOUT every two seconds, while storing the output of the actuall command in logfile.
Also, if your distro has it, take a look at the options of "usr/bin/time" - this is kinda different than the "time" command.
For the other things that you're asking, can you please provide some examples ?

hm... there is problem about /usr/bin/time. It doesn't return maximum resident set size, which i need too (it always return 0 instead). Is there a way to get maximum resident set size of a process that has finished?

also this:

Code:

for i in `seq 1 100`; do time ps -ef >> logfile.log; sleep 2; done

isn't exactly what i wast thinking about. what i aim is something like that:

Code:

time -o "logfile.txt" -f "parameters i want" my_given_command &
childpid=$!
for i in 'seq 1 100' do; gether_and_write_childpid_statistics; sleep 1; done;
write_statistics_from_time_command

But the problem with it is that:
1. i would like to know if childpid has finished and break the loop then (better if i could do it in sleep command, but i dont know if its possible). i don't want to gather data from finished process
2. i would like to have reliable command for monitoring childpid (ps only returns cpu time in seconds, i don't know if i can monitor it better)
3. and the problem with time that i write about at the beginning, how to obey it?

Thx

Also is there a command that wait's for child for specific amount of time and if child wont finish it return error after that period?