On 20120502_102011, Johannes Schauer wrote: > Hi, > > I'm looking for a unix tool that does nothing else than increment and > print an integer with a fixed frequency. As a bonus it should be able to > execute a command with a fixed frequency. The special requirement: it > should precise in the interval. > > Thus, the following will not work: > > #!/bin/sh > while true; do > print_counter > sleep 1 > done > > because each loop iteration will take one second plus how long > print_counter takes to execute. > > Even the following will not work: > > #!/bin/sh > while true; do > print_counter & > sleep 1 > done > > Because even the time it takes to fork print_counter will accumulate > over time. > > Another inconvenience of the above is, that not all sleep(1) > implementations do accept floating point arguments to run the above > faster than every second. > > What I need is something that on each step recalculates how long to > sleep based on the current time so that the overall frequency remains > stable. > > I do not need each print_counter to be executed exactly at a very > precise moment but just that the overall frequency stays the same. After > having it run 1000 times with a frequency of 1Hz I want 1000 seconds to > have passed. > > Some libraries like glib provide functions like g_timout_add or > g_timeout_add_seconds but that does not 'catch up' as I require above. > Is there some library that does? > > There is the watch(1) tool which already partly does what I want using > the --precise switch. But it runs in fullscreen and even when using > --precise, it will not compensate for commands that take longer than the > given interval length as it doesnt fork them. Also, watch(1) will not > allow intervals smaller than 0.1 seconds. > > So basic requirement: print a counter (counting up or down) in a fixed > frequency. Either a tool that does that or a library that I can use to > code it. > > Bonus: execute a command with a fixed frequency and in contrast to > watch(1) even when executing the command takes longer than the interval. > > So is there a utility that just implements a simple, precise counter? > > Is there a better version of watch(1) that is not fullscreen, allows > faster than 0.1 second intervals and forks the application so that their > runtime can exceed the interval time? > > cheers, josch
Don't use sleep, or any of the loop structures that you discuss above. Use crontab (see man crontab). This is not the /etc/crontab that drives some system services, but the one that provides a crontab for each user Typing crontab -e <return> in an xterm, will get you started, but perhaps you should read the man page first. Your crontab file will be created automagically in /var where the system expects it to be. I think you should be careful to not set the schedule to execute you job too frequently. This system will launch a new job on schedule, whether or not the previously launched job has completed. For some jobs this can cause problems. If you want to do something that takes 1 or 2 sec., every 5 sec., crontab works without using locks on files, or other safety mechanisms. Or look at the package liblockfile1. -- Paul E Condon pecon...@mesanetworks.net -- To UNSUBSCRIBE, email to debian-user-requ...@lists.debian.org with a subject of "unsubscribe". Trouble? Contact listmas...@lists.debian.org Archive: http://lists.debian.org/20120502143339.gb13...@big.lan.gnu
