Re: [Maxima-discuss] declared constants cannot be formal parameters

[Pankaj S. <pan...@gm...>]

On Sun, Aug 10, 2014 at 5:24 AM, <
max...@li...> wrote:

> I think that all Maxima declarations and assumptions are global.
> That is, when you say declare(n,constant),  ALL stack bindings
>   are ignored.  Why?  because only name-value pairs are eligible for
> binding locally.
> There is a declaration  local()  which provides an opportunity to bind
> name-function pairs.
>
> We could consider whether this is optimal or should be changed ...
>
> (a) technical standpoint (supporting programmers)
> (b) easiest to explain / principle of least surprise for users
> (c) not breaking existing code
> (d) fixing apparently surprising "features" -- later revealed to be more
> like bugs.
> (e) other criteria?
>
> At the moment (d)  seems to push us in the direction of allowing
> constants to be bound.
>
> It  would be possible to change meval and friends to do lexical scoping,
> or a flag
> which makes the change.  Some years ago Maple did a revision (or a
> revolution!)
> and changed its scoping.   I think it is a credit to the generally
> shallow programming
> in these systems that for a huge majority of the existing code, it
> doesn't matter.
>
> A programmer deeply into Scheme probably would write programs where it
> does matter.
>
> RJF
>

​>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>
I would like to second RJF's opinion on global space assignment of all
declaration,assumptions and rules and functions.

a) The idea that maxima could offer to resolve is using "context". In
documentation there are references to subcontexts but nothing about how
they are used from the parent context.​
creating a "supcontext" doesn't add facts() from both into one.

b) Also there is no information if some variable could be used from another
context using some fully qualified name.

c)properties(function_name) shall also list the context it is relevant to.
Means even functions could be categorized.

d) This feature is also important in case someone wants to divide somewhat
similar rules into multiple containers(contexts). Otherwise a person has to
write lot of predicates to differentiate all similar lhs rules from others.

e) Another use case, one can add frequently used rules to a subcontext​ and
search first in it, as in Rete's algo in term rewriting systems. Its useful
only if list of rules is exceptionally long.

f) A temporary context could be used that will be killed automatically once
the task is achieved.

There can be many more uses of this interaction of contexts.

-- 
Regards,
Pankaj Sejwal
___________________________________________
"The more I read, the more I acquire, the more certain I am that I know
nothing.” - Voltaire