On Wed, Sep 16, 2009 at 03:28:47PM -0400, Minimiscience wrote:
> I have a Perl 6 function currently written in the form:
>
> sub func($x) {
> for @data -> $datum {
> if condition($datum, $x) ff * {
> # Do various things
> }
> }
> # Do various other things
> }
>
> I want the "various things" code to be run for the first element in
> C<@data> for which C<condition($datum, $x)> returns true and for every
> element in C<@data> after that (regardless of whether C<condition()> is
> true, and ignoring the possibility of C<last>s) and for the state of
> whether C<condition()> has returned true to be reset for each invocation
> of C<func()>. I originally thought that the C<ff> operator with Whatever
> as its right-hand operand (so that it never unflipped/flopped) would do
> what I meant, but the documentation for Perl 5's C<..> in scalar context
> (which S03 currently defers to) seems to imply that C<ff>'s state is
> maintained across function invocations. Rakudo doesn't seem to have
> implemented C<ff> yet, so I can't test it, but if the state is indeed
> maintained across calls in Perl 6, how should I write C<func()> instead?
> The best way I can think of is the very C-like practice of creating a
> variable to store whether C<condition($datum, $x)> has returned true yet
> and updating it each iteration until true is returned, but it seems like
> Perl 6 should have a better way.
The theory here is that the C<ff> state is based on an implicit
C<state> variable declared in the same scope, and therefore would
have the same lifetime as that state variable. A state variable is
not like a C static, insofar as you get a new state variable when the
closure is cloned. In this case. the call to the C<for> loop should be
treating the block as an argument to the implementation of the loop,
and should therefore clone a new closure containing the C<ff>, so it
ought to reset the implicit state variable on every reinitialization
of the whole loop (but not on each loop iteration).
So I think in theory it should do what you want. (However, I can well
imagine an implementation botching the cloning of the loop body by
assuming it's static and not cloning it. That would not be Perl 6,
since in Perl 6 every block is assumed to be a closure, and C<for>
is assumed to work just like C<map>, which will certainly clone
its closure.)
Larry
