separate prefix algorithm from semantics

[benbrastmckie]

I thought it might be helpful to open an issue for this. The aim is to refactor the prefix function so as not to depend on which operators the language includes, remaining agnostic about the semantics and even arity for those operators. Here is a sketch what I was thinking:

• define find_op_and_args(string, parens_number=0, triggered=False) algorithm:
• if triggered and parens_number=0, then return
• if first character is a left parentheses:
  • cut and store the first character in left_arg
  • set parens_number =+ 1
  • run find_op_and_args(string, parens_number, true)
• if first character is a right parentheses:
  • cut and store the first character in left_arg
  • set parens_number =- 1
  • run find_op_and_args(string, parens_number, triggered)
• else:
  • cut and store first character in left_arg
  • run find_op_and_args(string, parens_number, triggered)
• define prefix algorithm:
• infix_sentence takes one of three forms:
  • (left_arg \op right_arg) in the case of a binary operator
  • \op single_arg in the case of a unary operator
  • \op in the case of a zero-place operator
• converting to prefix form is easy for the latter two forms
• if the first form, then the algorithm proceeds by:
  • delete first left parentheses
  • if the next character is a space, delete and check the next character
  • if the next character is a parentheses, then set left_arg, op, right_arg = find_op_and_args(infix_sentence)
  • else:
  • store everything up to first \ as left_arg
  • store the next word as op
  • store the remainder minus the final parentheses as right_arg
  • return [op, prefix(left_arg), prefix(right_arg)]

There may well be ways to improve on this but thought this might help to provide a way in.

[benbrastmckie]

I almost got pure_prefix to work. It runs but has errors for complicated cases. I can't quite see what the issue is. If you have a chance to take a look I'd be curious to know if you can spot the error. Seems like it is very close.

[benbrastmckie]

Never mind I got it!

[benbrastmckie]

I made all of the syntax functions methods of the ModelSetup class. If it makes more sense, we could move some of these to a syntax module, though right now no module is imported by hidden_things.py. I suspect elements are still missing from the syntax functions from before, in particular, I did not include the return [Const(token, AtomSort)] though commented out where I think it should go.