On Mon, Dec 27, 2010 at 11:27 AM, Justin C. Walker <[email protected]> wrote: > > On Dec 27, 2010, at 10:58 , Robert Bradshaw wrote: > >> On Sat, Dec 25, 2010 at 5:13 PM, Justin C. Walker <[email protected]> wrote: >>> Hi, all, >>> >>> A couple of questions regarding mutabilty: >>> >>> First, I noticed a comment at the beginning of quadratic_forms/binary_qf.py >>> from Stein: "make immutable". Can anyone (William?) elaborate on what >>> that means? They don't currently (e.g., sage-4.6.1-a3) seem to be very >>> immutable: >>> sage: q1=BinaryQF((-2,0,1)) >>> sage: q1._a=3 >>> sage: q1 >>> 3*x^2 + y^2 >> >> sage: q1.discriminant() >> 8 > > This is "accidental" (because of the @cached_method decoration), right? > Anything that refers, uncached, to q._a will pick up the current value of > that variable.
Yep. > Could one remove all mention of _a,_b,_c in the implementation, and keep > those values in a tuple? Well, of course, one could. That would make the > instance relatively immutable (except for clobbering the whole tuple). I don't think clobbering the whole tuple would be that much harder, but it would add an extra level of indirection everywhere. >> It's generally unsafe to set underscore attributes directly, as it's >> unclear what may be cached or what invariants may need to be > > I understand that aspect of the implementation. I wondered (based on > William's comment at the head of the file) whether one really could make the > class (or rather, its instances) immutable. > >>> Second, on this note, is there anywhere written a discussion of what >>> immutability means in Python? A quick check of the doc (for Python 2.[67]) >>> doesn't seem to lead anywhere. >> >> Immutability is just a matter of convention. > > And that means 'no' (except for, say, tuples). > >>> For example, can one declare a "class" immutable and have it stick, or is >>> there extra work involved to ride herd on all "instance variables" that one >>> would want to be unchangeable directly? >> >> As mentioned, one can use __setattr__. It's easy from Cython to make a class >> immutable from Python as well. > > This because the implementation is hidden in C, correct? Such a class would > not be immune to fiddling from other C code. Exactly. The language for the most part doesn't prevent you from meddling with internals, just in case you need to, but by convention an underscore is used to denote something internal that you shouldn't change/call/access unless you really know what you're doing (e.g. as an implementer of the class itself). - Robert -- To post to this group, send an email to [email protected] To unsubscribe from this group, send an email to [email protected] For more options, visit this group at http://groups.google.com/group/sage-devel URL: http://www.sagemath.org
