On Thu, Feb 9, 2012 at 12:33 AM, Joachim Durchholz <j...@durchholz.org> wrote: > Am 09.02.2012 02:48, schrieb Sean Vig: > >> As a second note, when I asked Aaron about this or IRC, he mentioned it >> may >> not be possible at all to have something with an attribute named None, as >> would be the case for something like S.None going to BooleanNone. If this >> is the case, could this be hacked in somehow (Aaron suggested possibly >> working with __getattr__), or would it be better to name it something >> different so this isn't an issue? If so, are there any suggestions for >> what >> it should be called? > > > A quick scan of the English and German Wikipedia pages on three-valued logic > turned up a whole lot of different logic systems, with different truth > tables. The meaning of the third value could be things like: > - indeterminate/unknown > - irrelevant > - both > - neither > - nonsensical/paradox > > Is "none"/"no value" the right terminology for the kind of logic that's > being discussed here? > What's the truth table - does it match an existing three-valued logic? If > yes, I'd take the terminology from there.
Actually, "no value" is pretty close. None means that we don't know the answer, ether because it could be either True or False, or because we don't have sufficient algorithms to compute it (we don't differentiate between the two). So for example, if x is assumed to be positive, x > 2 would have the logical value of None, because we don't have enough information to determine if x is in the interval (0, 2] or (2, oo). On the other hand, something like (pi + E).is_irrational would be None, because this fact is not known for sure by anyone. As far as the truth table, the logical operands short-circuit as you would expect, but otherwise, an operand with None gives None again. So for example, True & None would be True, but False & None would be None. Not None would also be None. Note that Python's two valued logic assigns to None the boolean value of False. This is useful, because you can check for truth using the two valued logic (e.g., "if x.is_positive" will be correct for True, False, or None), but you have to be careful when checking for falsity. In that case, you have to check something like "if x.is_positive is False" instead of "if not x.is_positive", or you will get the wrong answer if x.is_positive is None. Better would be to have something like x.is_nonpositive, which would do the three-valued logic for you. But that's off topic. The point is that I think None is actually an OK name. As far as putting it in S, it turns out I was wrong. You can't do "S.None = something", because that's a syntax error (I tested it wrong). You also can't do it with True or False, as that's a syntax error in Python 3. Furthermore, S.True, S.False, and S.None are all SyntaxErrors in Python 3, so even __setattr__ wouldn't work (this could work in Python 2, where that is not a syntax error, but we have to support both). The Singleton registry is built to assign the exact name of the class to the attributes automatically (you just have to add "__metaclass__ = Singleton" to the top of the class definition). But you can define any additional attribute, and it seems to work. So you can just do S.Whatever = BooleanNone and so on, and it will work (or at least it worked in my tests). So we need to come up with good shortcut names. Here are some options: - S.T, S.F, and S.N? Those are easy to type, though they could be a little confusing. - S.true, S.false, S.none - Instead of None, use a name more indicative of what it means, like "maybe" or "unknown" (this option is independent of the others). - Don't have any shortcut names. If you want it, you can do S.BooleanTrue, etc., or just S(True). I'm slightly leaning toward the last option, though not too strongly. Aaron Meurer > > > -- > You received this message because you are subscribed to the Google Groups > "sympy" group. > To post to this group, send email to sympy@googlegroups.com. > To unsubscribe from this group, send email to > sympy+unsubscr...@googlegroups.com. > For more options, visit this group at > http://groups.google.com/group/sympy?hl=en. > -- You received this message because you are subscribed to the Google Groups "sympy" group. To post to this group, send email to sympy@googlegroups.com. To unsubscribe from this group, send email to sympy+unsubscr...@googlegroups.com. For more options, visit this group at http://groups.google.com/group/sympy?hl=en.
