I have been thinking about similar things, but then you get into the
question of combining multiple algebras with overlapping types. Or you
could have second-order constructs, such as creating your
MatrixVectorAlgebra on top of an algebra for the scalars. If you use
constructors as in your example you could pass things in.
I think it is doable, but certainly a challenge and sufficiently
different for your average OO developer to hate it :-)
Still haven't gotten around to give Fortress a closer look.
Peter
Reinier Zwitserloot wrote:
> You know, your post made me think of this construct:
>
> using ( new MatrixVectorAlgebra() ) {
> Matrix result = matrix * vector;
> }
>
> using ( new ExactMath() ) {
> BigInteger i = 10 * foobar;
> }
>
>
> the algebra controls both how literals get translated, and the
> meanings of all operators. Just throwing it out there. I'll need to
> think on it _a lot_ more, and it has the disadvantage that any given
> snippet of code is pretty much meaningless unless you know the algebra
> context that goes with it, but then that's kind of true for any
> sufficiently complex snippet (the need to know context), and its
> certainly true when using operator overloading - you'd need to know
> the type of the LHS before you know what the operator would do, in for
> example python. So that seems like an acceptable price to pay.
>
> Could be overkill, though. Feels very fortressy.
>
> On Aug 17, 2:03 am, Peter Becker <[email protected]> wrote:
>
>> To twist the thread a bit more: IMO the whole idea of having an operator
>> with mixed types on a class is bad. So is having the
>> Matrix.multiply(Vector) method. The Matrix-Vector product (left or
>> right) is not intrinsic to either type, it is part of a larger algebra.
>>
>> Unfortunately OO won the market to an extent where sanity was lost :-)
>>
>> Operator definitions make much more sense when they are seen as part of
>> an algebra and not hacked onto OO structures. If you want to do them on
>> polymorphic structures with mixed types you probably need to have
>> multiple dispatch in your runtime environment.
>>
>> Peter
>>
>>
>>
>> Jess Holle wrote:
>>
>>> I've certainly used operator overloading myself for graphics matrix math.
>>>
>>> That said, I don't see how it "speeds up execution". It just reduces
>>>
>>> vector2 = matrix.multiply( vector1 );
>>>
>>> to
>>>
>>> vector2 = matrix * vector1;
>>>
>>> Nice example of clear and concise (and appropriate) use of operator
>>> overloading -- but not of execution speed improvement.
>>>
>>> --
>>> Jess Holle
>>>
>>> Casper Bang wrote:
>>>
>>>> With all due respect, operator overloading in Fan is an entirely
>>>> different, much less scary beast than the infamous C++ implementation
>>>> which everyone seems to have in mind. I'm not much for guns, but to
>>>> continue gun metaphor: Python and C# both demonstrates that this is
>>>> not necessarily an exploding Uzi. But if your premise is that
>>>> programmers will do bad things when given the slightest chance, then
>>>> remember that you don't actually need a gun to kill somebody - it just
>>>> gets the job done easier and faster.
>>>>
>>>> Btw. just yesterday I witnessed an example of how operator overloading
>>>> simplifies code and speeds up execution:
>>>> http://blog.joa-ebert.com/2009/08/10/flirting-with-silverlight/
>>>>
>>>> /Casper
>>>>
>>>> On 13 Aug., 13:20, Ben Schulz <[email protected]> wrote:
>>>>
>>>>>> So operator overloading is bad because people have HR problems?
>>>>>>
>>>>> Why stop at operator overloading? To hell with all safe guards.. No
>>>>> more speed limits or safty belts; let's everyone just drive
>>>>> reasonably. And what's with architects and statics? It'll hold.
>>>>>
>>>>>> Don't blame the gun, blame the shooter.
>>>>>>
>>>>> Yes, because *that's* how you bring people back from the dead.
>>>>>
>>>>> With kind regards
>>>>> Ben
>>>>>
> >
>
--~--~---------~--~----~------------~-------~--~----~
You received this message because you are subscribed to the Google Groups "The
Java Posse" group.
To post to this group, send email to [email protected]
To unsubscribe from this group, send email to
[email protected]
For more options, visit this group at
http://groups.google.com/group/javaposse?hl=en
-~----------~----~----~----~------~----~------~--~---
