MaplePrimes - answers and comments on Question, subs command bug?!

rtable_eval

acer — Fri, 10 Feb 2012 02:36:10 Z

Curiously, something very similar was asked just a day or so ago.

> restart:

> b:=Vector([lambda-2, lambda+23]):

> assume(lambda, real);

> subs(lambda=1, b);

                                [lambda~ - 2 ]
                                [            ]
                                [lambda~ + 23]

> eval(b, lambda=1);   
                                [lambda~ - 2 ]
                                [            ]
                                [lambda~ + 23]

> subs(lambda=1, rtable_eval(b));

                                     [-1]
                                     [  ]
                                     [24]

> eval(rtable_eval(b), lambda=1);
                                     [-1]
                                     [  ]
                                     [24]

> rtable_eval(b,inplace): # only need this once

> subs(lambda=1, b);

                                     [-1]
                                     [  ]
                                     [24]

> eval(b, lambda=1);             
                                     [-1]
                                     [  ]
                                     [24]

acer

Full/partial evaluation

Joe Riel — Fri, 10 Feb 2012 06:19:33 Z

To better understand this behavior, consider the following

proc()
local a,v;
global g;
    a := v;
    g := v;
    assume(v,positive);
    subs(v=1,[a,g,v]);
end proc();
                                   [v, 1, 1]

The substitution "works" for the global variable, g, but not the local variable, a. The reason for this is that global variables are fully evaluated, while local variables are not, they are evaluated only one level (cf. ?procedure). Because a is not fully evaluated, in the subs operation it evaluates to v, which is a different variable than v~. The v's in the call to subs evaluate to v~.

The same thing is happening with the substitution into a Vector (more generally, an ?rtable) even though you are doing so at top-level (i.e. not inside a proceure). For efficiency, the content of an rtable is not fully evaluated unless you force it to be.

thanks

itsme — Fri, 10 Feb 2012 03:06:19 Z

Thanks for your answer - I missed the previous question.

Do you consider this a desirable behavior (the need to set an "inplace" flag if one of the variables has assumptions set)? I would personally prefer the substitution behavior to be the same regardlsess of the assumptions.

thanks

itsme — Fri, 10 Feb 2012 07:00:22 Z

i see! Thanks for taking the time to explain.

It might be useful to at some stage include this behavior as a "gotcha" in the 'subs' (or 'assume') help page - just a trivial example would probably be helpful.

rtable_eval

pagan — Fri, 10 Feb 2012 09:20:49 Z

@itsme

Here is an Answer to the similar question of a few days ago. It illustrates that the issue is about evaluation rules for rtables, and not about assumptions.

So, if there were to be some extra Help examples, they would be better placed on pages like ?subs or ?eval or ?Vector than they would on ?assume