On Sun, 17 Feb 2008 20:12:26 -0500 Behdad Esfahbod <behdad@xxxxxxxxxx> wrote: > On Sun, 2008-02-17 at 16:40 -0800, Andrew Farris wrote: > > Behdad Esfahbod wrote: > > >> On Sun, 17 Feb 2008 08:15:20 -0800, Arjan van de Ven wrote > > >> > > >>>> This should really be fixed in bash to short-circuit. > > >>> well......... not sure it can be done in the bash language. > > >>> (it may well guarantee that both get executed) > > > > > > There's nothing to execute there. All the expansions happen > > > first, so if you have things like [ -n "$something" -a "`id`" = > > > 0 ], the `id` call is made before passing control to "test". The > > > short-circuit would just happen in the evaluation in test. > > > There's no user-visible difference there. > > > > If thats the case shouldn't a statement like that always be nested > > so that `id` is not called if it is not necessary, possibly saving > > many disk accesses? It would seem that using the more complex > > constructs could be much slower if all the cases are evaluated > > before testing any of them. > > I believe that's correct. The initial case that Arjan brought up > though was [ -f somefile -a -f anotherfile ]. That doesn't have this > problem. > > /me is surprised that this kind of micro-optimizing has such drastic > effects that Arjan has measured. well looking to see if a file (that doesn't exist) exist is an expensive operation. Sure, the kernel caches files that don't exist *after the first time you ask for them*. Without that the kernel has to walk the entire directory on disk. so an extra -f is.. not cheap. > -- If you want to reach me at my work email, use arjan@xxxxxxxxxxxxxxx For development, discussion and tips for power savings, visit http://www.lesswatts.org -- fedora-devel-list mailing list fedora-devel-list@xxxxxxxxxx https://www.redhat.com/mailman/listinfo/fedora-devel-list
