From: Robert A Duff <bobduff@shell01.TheWorld.com>
Subject: Re: {Pre,Post}conditions and side effects
Date: Tue, 23 Dec 2014 20:03:37 -0500
Date: 2014-12-23T20:03:37-05:00 [thread overview]
Message-ID: <wccegrpq2om.fsf@shell01.TheWorld.com> (raw)
In-Reply-To: alpine.CYG.2.11.1412231747100.4196@WIL414CHAPIN.vtc.vsc.edu
Peter Chapin <PChapin@vtc.vsc.edu> writes:
> I think, actually, it is fairly common for assertions to require
> "gazillions" of CPU cycles. In fact, it is easy to come up with examples
> where the pre- and postconditions take far longer to execute than the
> subprogram to which they are attached.
>
> A simple example that I love because it is so simple yet so telling is
> binary search of an array. A reasonable precondition is that the array
> it is given is sorted. It might look like
>
> Pre => (for all I in A'First .. A'Last - 1 => (A(I) <= A(I + 1)))
>
> This takes O(n) time to evaluate. Yet binary search is an O(log(n))
> algorithm. For large arrays the precondition might take many thousands
> or even millions of times longer to execute than the subprogram itself.
>
> I don't think there is anything particularly unusual about this
> example. Here's another case: suppose you had a procedure that added an
> item to some kind of balanced binary tree. The postcondition might
> assert that the tree is still balanced in whatever sense is
> appropriate... probably an O(n) operation. Yet the insertion procedure
> is probably O(log(n)). It's easy to imagine many examples.
Yes, all that's true. But those would be better as
predicates/invariants instead of preconditions. For ex., if you know
Sort produces a sorted array, and Binary_Search takes a sorted array,
you don't have to check for sorted-ness on entry to Binary_Search if
you've got a Sorted_Array.
But note that your "Pre" above is a good example of what I was saying in
a somewhat-unrelated post in this thread: It is shorter and simpler
than doing a sort using some efficient sorting algorithm.
> For this reason I assume that in most cases programs must be deployed
> with assertions disabled or else there is little chance the program will
> be able to meet its performance goals.
Yes, or at least SOME assertions disabled.
I like to say, "If you don't need to disable assertions, then you don't
have enough assertions". It's a silly sound bite that is not always
true, but there's a grain of truth in it.
>... Thus putting anything resembling
> essential program logic in an assertion is, of course, just
> wrong.
Yes.
>... Forbidding assertions with side effects might be nice, but the
> programmer still has to be careful with them anyway.
Yes.
-- Bob
next prev parent reply other threads:[~2014-12-24 1:03 UTC|newest]
Thread overview: 107+ messages / expand[flat|nested] mbox.gz Atom feed top
2014-12-22 16:22 {Pre,Post}conditions and side effects Jean François Martinez
2014-12-22 17:18 ` Brad Moore
2014-12-23 8:22 ` Jean François Martinez
2014-12-23 17:05 ` Robert A Duff
2014-12-23 21:09 ` Jean François Martinez
2014-12-23 21:35 ` Robert A Duff
2014-12-23 23:02 ` Peter Chapin
2014-12-24 1:03 ` Robert A Duff [this message]
2015-04-24 8:59 ` Jacob Sparre Andersen
2015-04-24 9:18 ` J-P. Rosen
2015-04-24 23:39 ` Randy Brukardt
2015-04-24 12:10 ` G.B.
2015-04-24 14:40 ` Jacob Sparre Andersen
2015-04-24 16:29 ` G.B.
2015-04-24 23:46 ` Randy Brukardt
2015-04-24 22:26 ` Peter Chapin
2015-04-25 0:13 ` Randy Brukardt
2015-04-25 1:01 ` Peter Chapin
2015-04-25 5:51 ` Dmitry A. Kazakov
2015-04-25 0:31 ` Bob Duff
2015-04-25 12:08 ` vincent.diemunsch
2015-04-25 16:37 ` Georg Bauhaus
2015-05-06 21:07 ` Randy Brukardt
2015-05-06 22:10 ` Paul Rubin
2015-05-07 9:01 ` Georg Bauhaus
2015-05-07 9:12 ` Dmitry A. Kazakov
2015-05-07 9:29 ` Georg Bauhaus
2015-05-07 9:31 ` Georg Bauhaus
2015-05-07 18:32 ` Randy Brukardt
2015-05-08 7:50 ` Dmitry A. Kazakov
2015-05-08 23:31 ` Randy Brukardt
2015-05-09 6:16 ` Dmitry A. Kazakov
2015-05-12 0:28 ` Randy Brukardt
2015-05-12 8:04 ` Dmitry A. Kazakov
2015-05-07 10:06 ` Stefan.Lucks
2015-05-07 12:16 ` Dmitry A. Kazakov
2015-05-07 18:00 ` Stefan.Lucks
2015-05-07 19:01 ` Randy Brukardt
2015-05-07 19:29 ` Niklas Holsti
2015-05-08 23:16 ` Randy Brukardt
2015-05-09 5:18 ` Niklas Holsti
2015-05-12 0:15 ` Randy Brukardt
2015-05-07 19:55 ` Dmitry A. Kazakov
2015-05-08 23:24 ` Randy Brukardt
2015-05-09 5:47 ` Dmitry A. Kazakov
2015-05-07 18:52 ` Randy Brukardt
2015-05-07 19:40 ` Stefan.Lucks
2015-05-08 7:28 ` Dmitry A. Kazakov
2015-05-08 22:58 ` Randy Brukardt
2015-05-08 22:52 ` Randy Brukardt
2015-05-09 0:14 ` Paul Rubin
2015-05-12 0:30 ` Randy Brukardt
2015-05-12 18:10 ` Paul Rubin
2015-05-12 22:01 ` Randy Brukardt
2015-05-13 9:35 ` Dmitry A. Kazakov
2015-05-13 11:53 ` G.B.
2015-05-13 12:47 ` Dmitry A. Kazakov
2015-05-13 14:06 ` G.B.
2015-05-13 14:21 ` Dmitry A. Kazakov
2015-05-13 16:33 ` G.B.
2015-05-13 19:15 ` Dmitry A. Kazakov
2015-05-14 1:36 ` Randy Brukardt
2015-05-14 7:10 ` Dmitry A. Kazakov
2015-05-14 1:32 ` Randy Brukardt
2015-05-14 7:19 ` Dmitry A. Kazakov
2015-05-12 0:36 ` Randy Brukardt
2015-05-11 10:35 ` Stefan.Lucks
2015-05-11 21:49 ` vincent.diemunsch
2015-05-11 22:49 ` Peter Chapin
2015-05-12 4:49 ` vincent.diemunsch
2015-05-12 23:25 ` Peter Chapin
2015-05-13 9:00 ` vincent.diemunsch
2015-05-12 4:42 ` vincent.diemunsch
2015-05-12 14:53 ` johnscpg
2015-05-13 9:14 ` vincent.diemunsch
2015-05-12 1:03 ` Randy Brukardt
2015-05-12 7:21 ` Georg Bauhaus
2015-05-12 22:08 ` Randy Brukardt
2015-05-12 8:02 ` Georg Bauhaus
2015-05-12 22:14 ` Randy Brukardt
2015-05-12 8:37 ` Stefan.Lucks
2015-05-12 11:25 ` Stefan.Lucks
2015-05-12 18:44 ` Paul Rubin
2015-05-12 20:52 ` Stefan.Lucks
2015-05-18 9:49 ` Jacob Sparre Andersen
2015-05-18 12:10 ` Stefan.Lucks
2015-05-19 7:46 ` Jacob Sparre Andersen
2015-06-09 7:55 ` Stefan.Lucks
2015-06-09 12:02 ` G.B.
2015-06-09 17:16 ` Stefan.Lucks
2015-05-12 18:39 ` Paul Rubin
2015-05-12 20:51 ` Stefan.Lucks
2015-05-12 14:21 ` Bob Duff
2015-05-12 22:37 ` Randy Brukardt
2015-05-13 6:58 ` Georg Bauhaus
2015-05-14 1:21 ` Randy Brukardt
2015-05-07 21:29 ` Georg Bauhaus
2015-05-08 23:11 ` Randy Brukardt
2015-05-08 23:19 ` tmoran
2014-12-23 21:53 ` Florian Weimer
2014-12-24 11:41 ` Brad Moore
2014-12-22 19:38 ` sbelmont700
2014-12-22 19:59 ` Brad Moore
2014-12-22 23:46 ` Randy Brukardt
2014-12-23 10:41 ` Georg Bauhaus
2014-12-22 23:32 ` Randy Brukardt
2015-04-24 17:59 ` Shark8
replies disabled
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox