Re: Functional-style pattern matching
Hi,
Little Walker wrote:
> I've been looking around to see if there's been any discussion of
> introducing functional programming-style pattern matching for method/
> function dispatch. Could someone point me to any such discussions?
It's done multi dispatch in Perl 6, and you can find an introduction in
a book chapter which we're writing these days:
http://github.com/perl6/book/blob/master/src/multi-dispatch.pod
> I've seen that perl6 is taking on a lot of ideas from the functional
> world (lazy evaluation for instance). With multi-method dispatch on
> the agenda pattern matching seems inevitable but I haven't been able
> to find information about this or any related discussions.
>
> This is a contrived example of what I'm referring to:
>
> sub traverse([Leaf $a]) {
> # do something
> }
>
> sub traverse([Tree $left, Tree $right]) {
> traverse($left);
> traverse($right);
> }
>
> my $t = Tree(...);
> traverse($t);
You got it almost right. The simples approach is to use "normal" matching:
multi traverse(Leaf $a) {
# do something
}
multi traverse(Tree $t) {
traverse($t.left);
traverse($t.right);
}
The latter uses a bit of additional code for unpacking the $t parameter;
you can do that in a signature too (which is not yet covered in the book
chapter; you can find some documentation here:
)
multi traverse (Tree $t ( $left, $right) ) {
traverse($left);
traverse($right);
}
Since you don't actually access $t anywhere, there's no reason to give
it a name, so you can write this actually as
multi traverse (Tree $ ( $left, $right) ) {
traverse($left);
traverse($right);
}
Hope that helps,
Moritz
--
Moritz Lenz
http://perlgeek.de/ | http://perl-6.de/ | http://sudokugarden.de/
Re: Functional-style pattern matching
> That's almost exactly the example from: > > http://perlcabal.org/syn/S06.html#Unpacking_tree_node_parameters 1. I feel incredibly embarrassed to have missed this 2. This is awesome!
Re: Functional-style pattern matching
> Which is pretty powerful, really. Absolutely - I think you're referring to the 'type subset' stuff which is great. > This is where Perl 6 is not the same as functional > languages, since it's got an imperative OO element as well. True, there can be friction between the functional style and OO, but look at how Scala manages it with case classes. When you look at the implementation, really it boils down to syntactic sugar but then so do many of the cool new features in Perl 6! I bring this up because when thinking of what will be possible with lazy evaluation, junctions, named parameter shorthand, closures, etc., etc., somehow pattern matching screams out at me. It can be concise, expressive and unambiguous, and very complementary to Perl 6's existing feature set. I know a lot of work went into bringing functional and OO together to make Scala happen, so certainly there may be impracticalities in just 'adding pattern matching'. Scala case classes: http://www.scala-lang.org/node/107 or http://programming-scala.labs.oreilly.com/ch06.html#CaseClasses
Re: Functional-style pattern matching
Does the unpacking participate in dispatch? If a Hash comes in as $t with no 'left' key, will it fail to match? On Tuesday, March 9, 2010, Little Walker wrote: >> Which is pretty powerful, really. > > Absolutely - I think you're referring to the 'type subset' stuff which > is great. > >> This is where Perl 6 is not the same as functional >> languages, since it's got an imperative OO element as well. > > True, there can be friction between the functional style and OO, but > look at how Scala manages it with case classes. When you look at the > implementation, really it boils down to syntactic sugar but then so do > many of the cool new features in Perl 6! > > I bring this up because when thinking of what will be possible with > lazy evaluation, junctions, named parameter shorthand, closures, etc., > etc., somehow pattern matching screams out at me. It can be concise, > expressive and unambiguous, and very complementary to Perl 6's > existing feature set. I know a lot of work went into bringing > functional and OO together to make Scala happen, so certainly there > may be impracticalities in just 'adding pattern matching'. > > Scala case classes: http://www.scala-lang.org/node/107 or > http://programming-scala.labs.oreilly.com/ch06.html#CaseClasses > > -- Mark J. Reed
Re: Functional-style pattern matching
Mark (>):
> Does the unpacking participate in dispatch? If a Hash comes in as $t
> with no 'left' key, will it fail to match?
Yes.
$ perl6 -e 'sub foo(%h($left)) { say $left }; foo({ left => "OH HAI" })'
OH HAI
$ perl6 -e 'sub foo(%h($left)) {}; foo({ no => "left key" })'
Not enough positional parameters passed; got 0 but expected 1
[...]
The error message is perhaps slightly LTA, but at least Rakudo
(correctly) fails to match.
// Carl
Re: Functional-style pattern matching
Oh, wow. I was just asking about the spec; didn't know this stuff
already worked. Rakudos to the team! :)
On Wed, Mar 10, 2010 at 9:18 AM, Carl Mäsak wrote:
> Mark (>):
>> Does the unpacking participate in dispatch? If a Hash comes in as $t
>> with no 'left' key, will it fail to match?
>
> Yes.
>
> $ perl6 -e 'sub foo(%h($left)) { say $left }; foo({ left => "OH HAI" })'
> OH HAI
>
> $ perl6 -e 'sub foo(%h($left)) {}; foo({ no => "left key" })'
> Not enough positional parameters passed; got 0 but expected 1
> [...]
>
> The error message is perhaps slightly LTA, but at least Rakudo
> (correctly) fails to match.
>
> // Carl
>
--
Mark J. Reed
