Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!think.com!snorkelwacker.mit.edu!bloom-beacon!eru!hagbard!sunic!mcsun!ukc!edcastle!aiai!jeff
From: jeff@aiai.ed.ac.uk (Jeff Dalton)
Newsgroups: comp.lang.scheme
Subject: Re: Order of evaluation (was Re: evaluating () should be an error
Message-ID: <4375@skye.ed.ac.uk>
Date: 28 Mar 91 18:47:32 GMT
References: <2977@kraftbus.cs.tu-berlin.de> <1991Mar26.155905.12906@daffy.cs.wisc.edu> <kiran.670017354@copper> <1991Mar26.224805.23381@cs.ubc.ca> <harlan.670038398@copper>
Reply-To: jeff@aiai.UUCP (Jeff Dalton)
Organization: AIAI, University of Edinburgh, Scotland
Lines: 35

What it comes down to is that Scheme is the sort of language that
doesn't specify such things.  

The greater range of optimizations available when the order of
argument evaluation is not specified is not, in my view, a sufficient
reason for not specifying the order.  Indeed, a compiler would often
be able to reorder anyway, becuase it could often tell that some
expressions involve no side effects.  There are already compilers that
perform that sort of analysis, at least to some extent.  So I do not
think the efficiency gains, though real, are sufficiently great.  As
additional, though not necessarily convincing, evidence, I will note
that Common Lisp does specify the order of evaluation, yet I have not
heard loud complaints that it is a massive blow to efficiency there.

Here are 2 arguments in favor of specifying the order of argument
evaluation:

1. Programs will be more portable.  Since bugs related to argument
   order can be difficult to find, this may represent a significant
   savings in programmers' time, which is more important than machine
   cycles.

2. The operation of the program will be more closely related to the
   structure of the source text.  (Assuming, of course, that the
   specified order is left-to-right.)  Moreover, LET* will be used
   to accomplish one thing (values of later variables depend on
   values of earlier ones) rather than two (as before plus sequential
   computation of the values).

These arguments are not, of course, conclusive; but then neither
are the counterarguments.  In the end, it becomes a question of
which factors are considered most important.  For Scheme, one of
them was the desire to move in the direction of the Algol/C family.

-- jd