On 2013-07-30, Stavros Macrakis <macrakis at alum.mit.edu> wrote:
> 1) What does it do for the naive user, the one who simply doesn't know that
> %i, %pi, etc. have special meanings?
Actually it seems best for them; they expect their %i, %pi, etc to act
like an ordinary variable, and indeed it does.
> 2) What does that do to the extremely common and very useful pattern of
> (val: <expression involving x>, ..., operation(val, x) )? In this
> expression, <expression involving x> is not within the scope of operation's
> x.
Yeah -- an implicit lexical environment for integrate, plot2d, etc
doesn't extend to these common, useful operations.
> 3) If a %pi leaks, how do we distinguish it from other %pi's?
Well, it's the one that doesn't have any interesting properties.
> 4) Binding constant symbols seems to be a small detail in your grand
> scheme. Let's figure out how to handle x and y first, then we can worry
> about %i and %pi.
Incidentally, it's not "my" scheme -- I'm surely not the first or last
to bring up lexical scope, and not even the one who raised it in the
current iteration. Associating ideas with people is problematic -- it
makes is more difficult to be critical of the idea since the criticism
seems to carry over to the person as well. And one might change his mind
as the discussion progresses, so that which was once "his" idea isn't
anymore.
Anyway I'm beginning to think any implicit lexical environment is likely
to be too hard to understand. An explicit lexical environment would
solve the name collision problem, and make it clear which symbols are
affected (so if a leaked symbol has different properties, at least it
isn't a surprise), and could also enclose all of the stuff that's relevant
(addressing item (2) above).
To return to plot2d, where we started a day or two ago, at this point
I'm inclined to say trying to use a constant should trigger a complaint.
(Actually COERCE-FLOAT-FUN should probably just use MSETQ to assign
a value -- that will trigger any complaints which assignment triggers.
There may be other kinds of forbidden assignments detected by MSETQ.)
Let the user enclose plot2d in a lexical environment if they want to
use a symbol which already has special properties. We'll want to create
a lexical environment for general use. As a proof of concept, consider
blex. [http://www.math.utexas.edu/pipermail/maxima/2012/029643.html]
best
Robert Dodier