From @masak

r30183​:
$ ./perl6 -e 'grammar A { token TOP { <foo> }; token foo { d } }; "d"
~~ A​::TOP' # works
$ ./perl6 -e 'grammar A { token TOP { <text> }; token text { d } };
"d" ~~ A​::TOP' # fails!
Null PMC access in type()
[...]

Same goes for 'rule text' or 'regex test'. My uneducated guess is that
something inside PGE is called 'text' and clashes.

From @pmichaud

On Tue Aug 12 11​:58​:42 2008, masak wrote​:

r30183​:
$ ./perl6 -e 'grammar A { token TOP { <foo> }; token foo { d } }; "d"
~~ A​::TOP' # works
$ ./perl6 -e 'grammar A { token TOP { <text> }; token text { d } };
"d" ~~ A​::TOP' # fails!
Null PMC access in type()
[...]

Same goes for 'rule text' or 'regex test'. My uneducated guess is that
something inside PGE is called 'text' and clashes.

You're correct -- .text is a method on Match objects (see S05), and at
the moment that tends to conflict with any rules named 'text' in a
grammar. This may be fixed when we convert PGE to generate multimethods
for regexes instead of normal methods, but ultimately there's a deep
underlying conflict with Match methods and grammar methods that we'll
need to resolve somehow.

Thanks,

Pm

The RT System itself - Status changed from 'new' to 'open'

From @chrisdolan

On Sun Sep 14 07​:51​:56 2008, pmichaud wrote​:

You're correct -- .text is a method on Match objects (see S05), and at
the moment that tends to conflict with any rules named 'text' in a
grammar. This may be fixed when we convert PGE to generate multimethods
for regexes instead of normal methods, but ultimately there's a deep
underlying conflict with Match methods and grammar methods that we'll
need to resolve somehow.

It seems to be Regex.pir that is taking all of the invocations. I have
discovered that, beyond <text>, I cannot create tokens/rules named
"null", "ws", etc. The latter is particularly problematic in writing
parsers in Perl6. The following simple example shows the failure​:

grammar WSOverride {
  token TOP { <tok_foo> <.ws> <tok_bar> };
  token tok_foo { foo };
  token tok_bar { bar };
  token ws { [ \h | \v | '%' ]+ };
}

ok('foo bar' ~~ WSOverride​::TOP); # succeeds
ok("foo\nbar" ~~ WSOverride​::TOP); # succeeds
ok("foo%%%\nbar" ~~ WSOverride​::TOP); # fails

From @pmichaud

On Fri, Oct 31, 2008 at 08​:47​:09PM -0700, Chris Dolan via RT wrote​:

It seems to be Regex.pir that is taking all of the invocations. I have
discovered that, beyond <text>, I cannot create tokens/rules named
"null", "ws", etc. The latter is particularly problematic in writing
parsers in Perl6. The following simple example shows the failure​:

grammar WSOverride {
token TOP { <tok_foo> <.ws> <tok_bar> };
token tok_foo { foo };
token tok_bar { bar };
token ws { [ \h | \v | '%' ]+ };
}

ok('foo bar' ~~ WSOverride​::TOP); # succeeds
ok("foo\nbar" ~~ WSOverride​::TOP); # succeeds
ok("foo%%%\nbar" ~~ WSOverride​::TOP); # fails

First, note that C< 'foo bar' ~~ WSOverride​::TOP > isn't
really the correct syntax for invoking a rule -- but the
synopses don't really specify what the correct syntax is yet.
At the moment we do know that it's not likely to be
C< 'foo' ~~ Grammar​::rule >.

Beyond that, the problem appears to be that Rakudo is not
presently creating its grammars properly -- it appears to
be deriving them directly from Any instead of from a PGE
grammar class of some sort. As a result, inheritance and
method dispatch don't seem to be working quite properly,
and we'll need to get that fixed.

Pm

From @masak

Resolving this ticket, since the original problem seems to have been solved.

Beyond that, the problem appears to be that Rakudo is not
presently creating its grammars properly -- it appears to
be deriving them directly from Any instead of from a PGE
grammar class of some sort. As a result, inheritance and
method dispatch don't seem to be working quite properly,
and we'll need to get that fixed.

Ticket [perl #​65474] covers this issue, along with an example of what fails.

@masak - Status changed from 'open' to 'resolved'