On Wed, 22 Jul 2009, Robert Dodier wrote:
< On Wed, Jul 22, 2009 at 1:13 PM, Richard Fateman<fateman at cs.berkeley.edu> wrote:
<
< > The way I understand the declarations ?(like float, ) is that they do
< > not inhibit assignment of values that disagree with the declarations.
< > They are advice to the compiler that should enable it to compile code
< > that is more specific, and will work faster. ?If object of the wrong
< > type are used, the computation may be incorrect.
This is the theory, but see below.
<
< The point of declarations is to allow the application of identities which
< are true for any value in the declared class.
< (This subsumes compiler advice and lots more.)
< If later on you substitute an element from outside the declared class,
< you'll very likely get an incorrect result.
I think that maxima is inconsistent in these regards. Here are some examples:
(%i2) define_variable(xt,[],list);
(%o2) []
(%i3) xt : 1;
Error: xt was declared mode list, has value: 1
-- an error. To debug this try debugmode(true);
(%i4) define_variable(yt,0,integer);
(%o4) 0
(%i5) yt : %e;
Error: yt was declared mode fixnum, has value: %e
-- an error. To debug this try debugmode(true);
(%i6) define_variable(zt,0.0,float);
(%o6) 0.0
(%i7) zt : [];
(%o7) []
You can see that the the assignment operator ':' does some type-checking,
but it is inconsistent. It was the first 2 examples that prompted my
original question. Also, when I looked at 'mset' in src/mlisp.lisp, I
could see that there is some type-checking (notably, for 'sysconst').
I have since realised that the online documentation for define_variable
explains how to force type-checking, so that one can hack a 'const'
class, for example.
Leo
--
The University of Edinburgh is a charitable body, registered in
Scotland, with registration number SC005336.