Calling a token 'any' calls for trouble

[p6rt]

Migrated from rt.perl.org#77350 (status was 'open')

Searchable as RT77350$

[p6rt]

From @moritz

22​:39 < moritz_> rakudo​: grammar A { token TOP { <any> }; token any {
'foo' | 'bar' } }; say A.parse('foo')
22​:39 <+p6eval> rakudo d1015f​: OUTPUT«Too many positional parameters
passed; got 2 but expected 1␤ in 'Any​::any' at line 1361​:CORE.setting␤
in 'A​::any' at line 22​:/tmp/PUiWn8O0Uy␤ in 'A​::TOP' at line
22​:/tmp/PUiWn8O0Uy␤ in 'Grammar​::parse' at line 5864​:CORE.setting␤ in
main program body at line
22​:39 <+p6eval> ..22​:/tmp/PUiWn8O0Uy␤»
22​:40 < moritz_> that's a bug, right?
22​:40 < pmichaud> well, the grammar is overriding the .any method
22​:41 < moritz_> pmichaud​: but it's not the overridden .any method that
is called, right?
22​:41 < pmichaud> moritz_​: I'm still trying to figure out what's
happening there.

[p6rt]

From @coke

No change in the past year.

--
Will "Coke" Coleda

[p6rt]

@coke - Status changed from 'new' to 'open'

[p6rt]

From @moritz

The problem is that the $*ACTIONS contextual is set to a type object
(used to be Any, now Mu), so $*ACTIONS.any existed, but caused an error.

commit 0973612366de17d6eab26e975663e63170e5d1d5 eased some of the pain
by switching it from Any to Mu, but the underlying problem is still there.

[p6rt]

From @Mouq

On Thu Oct 18 02​:13​:15 2012, moritz wrote​:

The problem is that the $*ACTIONS contextual is set to a type object
(used to be Any, now Mu), so $*ACTIONS.any existed, but caused an error.

commit 0973612366de17d6eab26e975663e63170e5d1d5 eased some of the pain
by switching it from Any to Mu, but the underlying problem is still there.

00​:11 <Mouq> r​: grammar A { token TOP { <any> }; token any { 'foo' | 'bar' } }; say A.parse('foo')
00​:11 <camelia> rakudo-parrot 79167a, rakudo-jvm 79167a, rakudo-moar 79167a​: OUTPUT«｢foo｣␤ any => ｢foo｣␤␤»

Closeable with tests

[p6rt]

From @moritz

On 04/14/2014 06​:11 AM, Alexander Moquin via RT wrote​:

On Thu Oct 18 02​:13​:15 2012, moritz wrote​:

The problem is that the $*ACTIONS contextual is set to a type object
(used to be Any, now Mu), so $*ACTIONS.any existed, but caused an error.

commit 0973612366de17d6eab26e975663e63170e5d1d5 eased some of the pain
by switching it from Any to Mu, but the underlying problem is still there.

00​:11 <Mouq> r​: grammar A { token TOP { <any> }; token any { 'foo' | 'bar' } }; say A.parse('foo')
00​:11 <camelia> rakudo-parrot 79167a, rakudo-jvm 79167a, rakudo-moar 79167a​: OUTPUT«｢foo｣␤ any => ｢foo｣␤␤»

Closeable with tests

The problem can still be reproduced with method names from class Mu, for
example 'so'​:

<moritz> m​: grammar A { token TOP { <so> }; token so { 'foo' | 'bar' }
  }; say A.parse('foo')
<camelia> rakudo-moar 07dcb2​: OUTPUT«Too many positional parameters
  passed; got 2 but expected 1␤ in method so at
  src/gen/m-CORE.setting​:834␤ in any !reduce at
  gen/moar/stage2/QRegex.nqp​:759␤ in any !cursor_pass at
  gen/moar/stage2/QRegex.nqp​:721␤ in regex so at
  /tmp/4Fzij7cVZH​:1…»

[p6rt]

From @jdv

I added a test in S05-grammar/inheritance.t but I'm unsure if
this is a valid bug. If the default actions dispatch off the Mu
type object then this is expected behavior. Maybe they can be
sidestepped as a special case for grammar actions? Or at least
a better error message could be spit out.

[p6rt]

From @moritz

The correct solution would be to track the presence or absence of an action class by an out-of-bounds mechanism, like a separate variable.

I just don't know how to do that without making parsing significantly slower.