[Reality] grunk, problem stemming from query implementation

Christopher Armstrong reality@twistedmatrix.com
Fri, 14 Feb 2003 09:53:55 -0500


This was just going to be for glyph, but I figured since I'm using
email, and it's relevant, I'd send it to the list as well.

What we want:

  To be able to disambiguate one Thing at a time, rather than entire Actions.

Thus, how I naively figured it would work:

  parse methods would no longer return lists of Actions, but rather they would
  return a single Action, which itself would look around for the objects,
  and tell the parsing system if there are any ambiguities.

Problem:

  We have potentially more than one parse method to call for any input; each
  one would return a single Action. Not only do we need to disambiguate each
  Thing that's potentially involved, but also which Action. I don't see how
  this can work :-) But maybe someone else can.

Potential (but probably not) Solution:

  Have parse methods return the fine-grained Thing ambiguities, and
  not construct any Actions yet? This moves the `look around for'
  behavior back to the parser, though. IIRC, we were explicitly moving
  this behavior to the Action. Do we really need to? Even if we do, it
  seems like this isn't quite going to work. My brain keeps fogging over
  as I try to think of use cases. :P

  Hrm... Or, is there a way we could have the parse method return a list of
  possibilites or proto-Actions or somesuch, such that we can throw each
  possibility from each parse method into a single multiple choice
  question?



Maybe I should just figure out a way to leave the current
(non-implementation of a) disambiguation system alone and still be
able to cram in the query system... :P

Confused,

-- 
 Twisted | Christopher Armstrong: International Man of Twistery
  Radix  |          Release Manager,  Twisted Project
---------+     http://twistedmatrix.com/users/radix.twistd/