On 24/09/2013, at 9:02 AM, srean wrote:
> On Mon, Sep 23, 2013 at 5:19 PM, john skaller
> <skal...@users.sourceforge.net> wrote:
>>
>> On 24/09/2013, at 7:10 AM, srean wrote:
>>
>>> I call bollocks.
>>
The bottom line is this: if you actually want to learn Felix you have
to follow my instructions. Stop complaining about rants in the
mailing list, stop reading a heap of crap.
WRITE SOME CODE.
Obey the instructions in the very first slide of the slideshow
and first tutorial: you cannot learn a new programming language
by reading documentation. You have to get your hands dirty.
Read the docs AFTER you understand to firm up what you already know.
Pick a problem such as "print the first 10 fibonacci numbers",
implement it, and post your solution.
Try another problem like "calculate the gcd of two numbers",
which is Euclids algorithm.
OK, so did you know, Euclid's algorithm doesn't just work
for integers? It will work for a whole CLASS of structures
called integral domains.
SO now, you have a higher level problem: you already have
Euclid's algorithm and you can change "int" to T to make it
polymorphic but now it will not work.
Try it! Why doesn't it work? Because + is not defined for T.
So actually, you have to look up Wikipedia for Integral Domain
and find out what one is, and define a type class for them.
Which means adding the functions that are specified for
an integral domain.
If you want a model for the way this is done with classes
you can look in the library, which defines groups and
rings in
lib/std/algebraic.flx
Once you have an actual problem you can pattern match to guess
how to use classes to solve the problem and keep trying until
it works (or you have to ask for help).
And since we don't have a class IntegralDomain[T] in
the library .. we can then add it to the library :)
Shayne managed to do this for something MUCH harder: he actually
managed to get Monads to work in Felix. That requires a higher
order operator which Felix just barely does support,
When you have conceptual problems you may write some gibberish
that fails and you just post it and I'll try to fix it.
Do this enough times, you'll get the patterns right.
The actual understanding (rather than just copy/paste) will
come later.
And even later .. the understanding of what the issues are with
that technology.
Some of the theoretical motivations for these things are quite deep.
For example classes are strictly limited in what they can do.
They're a kind of type level OO, and we KNOW that doesn't work.
The RIGHT way to do this is to build your models on category theory,
the way Ocaml does with modules and functors. That system is much
more general but has two problems: its a pain to use, and, until
Ocaml 3.12 at least you couldn't have module variables so functor
arguments had to be literals.
You probably don't understand that, but the constraint is that
type classes only allow type arguments. They SHOULD allow
class arguments, i.e. you should be able to parametrise a class
with another class. In Ocaml you can. In particular you can thereby
pass in function as well as type variables.
You may not understand that constraint from a theoretical description
but a practical example of it is easy to understand: with type classes
you can provide integers with a total order relation < so
1 < 2
for example. Fine. But, what about the reverse order
2 < 1
This is a perfectly good total order., You could define that instead
but you cannot have both together and there is no way to disambiguate.
There is no problem with modules, because all uses have to provide
the disambiguating indicator, which is the main problem with them:
you have to provide that indicator even if there's only one case.
Felix has classes instead of Ocaml like modules because it is
intended as a scripting language and with that in mind the weaker
but easier to use classes were chosen (after all if it's good enough
for Haskell who am I to argue)
--
john skaller
skal...@users.sourceforge.net
http://felix-lang.org
------------------------------------------------------------------------------
October Webinars: Code for Performance
Free Intel webinars can help you accelerate application performance.
Explore tips for MPI, OpenMP, advanced profiling, and more. Get the most from
the latest Intel processors and coprocessors. See abstracts and register >
http://pubads.g.doubleclick.net/gampad/clk?id=60133471&iu=/4140/ostg.clktrk
_______________________________________________
Felix-language mailing list
Felix-language@lists.sourceforge.net
https://lists.sourceforge.net/lists/listinfo/felix-language
