If you use the rewrite option, it remains possible to write a precedenced
statement that is inherently ambiguous.  Rewriting is to prevent the
*implementation* of precedenced rules from *introducing* new ambiguities
not actually in the precedenced rule itself.


On Fri, Sep 2, 2016 at 1:30 PM, <[email protected]> wrote:

> The rewrite approach seems to work great for recursive rules, but it
> doesn't do anything for non-recursive rules.  Running the following gives
> me an ambiguous parse.  Since opt1a is a higher priority, I expect to see
> two productions of opt1a with no ambiguity.  Am I expecting too much here?
> Thanks for your help!
>
>
>     use strict;
>     use Marpa::R2;
>     use Data::Dumper;
>
>     my $grammar = "
>     options ::= option*
>     option  ::= opt1a || opt2a
>     opt1a   ::= 'a'
>     opt2a   ::= 'a' 'a'
>
>     :discard ~ whitespace
>     whitespace ~ [\\s]+
>     ";
>
>     my $grammar = Marpa::R2::Scanless::G->new( { source => \$grammar } );
>     my $input = 'a a';
>     my $value_ref = $grammar->parse( \$input );
>
>     my $value = ${$value_ref};
>     print "Output:\n".Dumper($value);
>
>
> On Wednesday, August 31, 2016 at 8:55:21 PM UTC-6, Jeffrey Kegler wrote:
>>
>> I've remembered a major reason for preferring the rewrite approach.  If
>> you parse then rank, you're using ambiguous parsing -- Marpa is cool with
>> that, but it is not necessarily as fast as an unambiguous parse.  If the
>> expression is of size X the parse could be as bad as O(X^3) and that can
>> make a difference if you have a lot of long expressions.
>>
>> <snip>
>>
>> --
> You received this message because you are subscribed to the Google Groups
> "marpa parser" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to [email protected].
> For more options, visit https://groups.google.com/d/optout.
>

-- 
You received this message because you are subscribed to the Google Groups 
"marpa parser" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to [email protected].
For more options, visit https://groups.google.com/d/optout.

Reply via email to