Re: time command with shell builtins

[Dominik Vogt <dominik.vogt@gmx.de>]

On Mon, Jan 23, 2023 at 07:31:12PM +0100, Mikael Magnusson wrote:
> On 1/23/23, Dominik Vogt <dominik.vogt@gmx.de> wrote:
> > On Mon, Jan 23, 2023 at 02:42:05PM +0100, Roman Perepelitsa wrote:
> >> On Mon, Jan 23, 2023 at 2:40 PM Dominik Vogt <dominik.vogt@gmx.de> wrote:
> >> >
> >> > Is it possible to get timing statistics of shell builtins too?
> >> > Timing "echo" isn't very interesting, but timing loop constructs
> >> > would be:
> >> >
> >> >   $ time while foo; do bar done
> >>
> >> This:
> >>
> >>     % time ( while foo; do bar; done )
> >
> > That wasn't really the question.  Of course I can time a loop by
> > writing a different command, or by putting it in a pipe or file.
> >
> >   $ time echo foo | true
> >
> > I just want to get timing statistics of loops either explicitly by
> > prepending "time" or implicitly with REPORTTIME.
>
> As Bart already mentioned, the answer to your question is "no", but
> you can avoid some downsides of the subshell (eg, if your loop has
> side effects that are relevant to the rest of the script etc), by
> using SECONDS:
> % () { typeset -F4 SECONDS=0; sleep 1; () { typeset -F3 SECONDS=0;
> sleep 0.43; echo $SECONDS }; sleep 1; echo $SECONDS }
> 0.431
> 2.4329
> (the downside here is obviously that it doesn't split out cpu/system
> time for you, only elapsed time).

Well, the worst downside for me is that REPORTTIME does not work.
The use case is "oh, that command ran a long time, I'd really
like to know how long it took".  I see no solution for that if
re-running the command is no optiuon because it takes too long.

At the moment I'm writing some automation scripts that run for
hours and print their progress.  I might want to kill them after a
few hours and see how many seconds they ran and compare it to the
progress output.

Ciao

Dominik ^_^  ^_^

--

Dominik Vogt