On Nov 22, 2007, at 4:07 PM, Waldemar Kornewald wrote:
The absence of precedence between operators is a small price to pay,
because you gain minimality, consistency, and semantics at the right
level: messages sent to objects, not some domain-specific
"mathematical" operation.
I do understand that you like this kind of elegance, but I think the
price to pay is pretty high because it increases the likelihood of
mistakes (in either math or Smalltalk "mode"). Modality is a
well-known cause of errors and should be avoided. Precedence could
simply apply to +-*/, so nothing complex to learn.
Ian, what do you think?
I think whichever makes the expression of intent the clearest is the
one to use in any given situation. That it might have to come with
copious comments, flashing red 'danger bars' in the editor, or
whatever, is a detail for the UI and tools to get right.
Clarity can be in the eye of the writer or the reader, so an ideal
solution might employ bijections between visual/editable
representation and stored semantic structures throughout the
programming tools. If X wants to write with conventional arithmetic
precedence, import (or write) the two or three lines of syntax
declaration that makes certain binary operators behave that way. If
Y wants to read the code without operator precedence, the underlying
structures are converted (in the absence of precedence modifications)
with the operators appearing in a different order, possibly
parenthesised. I can see how this could rapidly break down into
chaos and confusion, but before it's abandoned I think someone (or
several people) should make serious attempts to find a practical and
satisfying solution.
(If we take this to the extreme, even Forth programs might one day be
readable. ;-)
In the meantime: given that the object system was created to
eliminate C(++), and the message syntax was created to keep me sane,
and I can never remember the traditional (C or similar) precedence
rules without a huge poster on my wall and lots of redundant
parentheses to be 101% sure the compiler understands them too, I have
to side with the three-level 'message-passing' precedence rules.
But, with the promise that it will be trivial for anyone to change
the way any particular selectors are parsed in the very near future.
Cheers,
Ian
_______________________________________________
fonc mailing list
[email protected]
http://vpri.org/mailman/listinfo/fonc
