bug#26026: Defining a method named zero? breaks primitive zero?

bug#26026: Defining a method named zero? breaks primitive zero?

[Alejandro Sanchez]

If I define a ‘zero?’ predicate method for a custom class the primitive ‘zero?’ is lost. Here is a simple vector module:

	;;; File vector2.scm
	(define-module (vector2)
	  #:use-module (oop goops)
	  #:export (<vector2> get-x get-y zero?))
	
	(define-class <vector2> ()
	  (x #:init-value 0 #:getter get-x #:init-keyword #:x)
	  (y #:init-value 0 #:getter get-y #:init-keyword #:y) )

	(define-generic zero?)
	(define-method (zero? (v <vector2>))
	  (and (zero? (get-x v))
	       (zero? (get-y v))))

In the Guile REPL try executing the following code:

	scheme@(guile-user)> (use-modules (oop goops) (vector2))
	scheme@(guile-user)> (zero? (make <vector2>))

This will display 

	WARNING: (guile-user): `zero?' imported from both (ice-9 r5rs) and (vector2)
	ERROR: In procedure scm-error:
	ERROR: No applicable method for #<<generic> zero? (1)> in call (zero? 0)
	
	Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
	scheme@(guile-user) [1]> ,bt
	In vector2.scm:
	     11:7  2 (_ #<<vector2> 105e87e00>)
	In oop/goops.scm:
	   1438:4  1 (cache-miss 0)
	In unknown file:
	           0 (scm-error goops-error #f "No applicable method for ~S in call ~S" (#<<gen…> …) …)

Apparently the problem is that ‘zero?’ is defined in two modules and the vector2 definition overrides it. This isn’t the case with other primitives like ‘+’ or ‘*’, so this seems like a bug? I had built Guile from HEAD a few days ago, my package manager shows 6fff84d as the version number, so I guess that must be the hash of the commit HEAD was pointing to at that time.

bug#26026: Defining a method named zero? breaks primitive zero?

[Andy Wingo]

On Wed 08 Mar 2017 12:07, Alejandro Sanchez <hiphish@openmailbox.org> writes:

> If I define a ‘zero?’ predicate method for a custom class the primitive ‘zero?’ is lost. Here is a simple vector module:
>
> 	;;; File vector2.scm
> 	(define-module (vector2)
> 	  #:use-module (oop goops)
> 	  #:export (<vector2> get-x get-y zero?))
> 	
> 	(define-class <vector2> ()
> 	  (x #:init-value 0 #:getter get-x #:init-keyword #:x)
> 	  (y #:init-value 0 #:getter get-y #:init-keyword #:y) )
>
> 	(define-generic zero?)
> 	(define-method (zero? (v <vector2>))
> 	  (and (zero? (get-x v))
> 	       (zero? (get-y v))))
>
> In the Guile REPL try executing the following code:
>
> 	scheme@(guile-user)> (use-modules (oop goops) (vector2))
> 	scheme@(guile-user)> (zero? (make <vector2>))
>
> This will display 
>
> 	WARNING: (guile-user): `zero?' imported from both (ice-9 r5rs) and (vector2)
> 	ERROR: In procedure scm-error:
> 	ERROR: No applicable method for #<<generic> zero? (1)> in call (zero? 0)
> 	
> 	Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
> 	scheme@(guile-user) [1]> ,bt
> 	In vector2.scm:
> 	     11:7  2 (_ #<<vector2> 105e87e00>)
> 	In oop/goops.scm:
> 	   1438:4  1 (cache-miss 0)
> 	In unknown file:
> 	           0 (scm-error goops-error #f "No applicable method for ~S in call ~S" (#<<gen…> …) …)
>
> Apparently the problem is that ‘zero?’ is defined in two modules and
> the vector2 definition overrides it. This isn’t the case with other
> primitives like ‘+’ or ‘*’, so this seems like a bug? I had built
> Guile from HEAD a few days ago, my package manager shows 6fff84d as
> the version number, so I guess that must be the hash of the commit
> HEAD was pointing to at that time.

Actually the (vector2) module makes a fresh definition for zero?.  You
can tell because zero? is in its export list.  So instead of extending
the primitive-generic that is zero?, you are making a new definition.
See:

  scheme@(guile-user)> (define-module (foo) #:export (zero?))
  $1 = #<directory (foo) 1203c80>
  scheme@(foo)> (zero? 0)
  <unnamed port>:4:0: <unnamed port>:4:0: Unbound variable: zero?

  Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.

If you want to extend a primitive-generic, then do that by not exporting
zero?.  In a way it's like mutating the primitive in place, giving it
additional powers.

Andy

bug#26026: closed (Re: bug#26026: Defining a method named zero? breaks primitive zero?)

[Alejandro Sanchez]

[-- Attachment #1: Type: text/plain, Size: 5776 bytes --]

This does not work. If I remove the export of ‘zero?’ I get another error:

	scheme@(guile-user)>  (zero? (make <vector2>))
	<unnamed port>:3:1: <unnamed port>:3:1: In procedure =: Wrong type: #<<vector2> 106606e20>
	
	Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
	scheme@(guile-user) [1]> ,bt
	In current input:
	      3:1  0 (_)
	scheme@(guile-user) [1]>

That is why I was exporting ‘zero?’ to begin with. I do not have to export ‘+’ or ‘-‘ for example.

> On 19 Apr 2017, at 17:13, GNU bug Tracking System <help-debbugs@gnu.org> wrote:
> 
> Your bug report
> 
> #26026: Defining a method named zero? breaks primitive zero?
> 
> which was filed against the guile package, has been closed.
> 
> The explanation is attached below, along with your original report.
> If you require more details, please reply to 26026@debbugs.gnu.org.
> 
> -- 
> 26026: http://debbugs.gnu.org/cgi/bugreport.cgi?bug=26026
> GNU Bug Tracking System
> Contact help-debbugs@gnu.org with problems
> 
> From: Andy Wingo <wingo@igalia.com>
> Subject: Re: bug#26026: Defining a method named zero? breaks primitive zero?
> Date: 19 April 2017 at 17:12:12 GMT+2
> To: Alejandro Sanchez <hiphish@openmailbox.org>
> Cc: 26026-done@debbugs.gnu.org
> 
> 
> On Wed 08 Mar 2017 12:07, Alejandro Sanchez <hiphish@openmailbox.org> writes:
> 
>> If I define a ‘zero?’ predicate method for a custom class the primitive ‘zero?’ is lost. Here is a simple vector module:
>> 
>> 	;;; File vector2.scm
>> 	(define-module (vector2)
>> 	  #:use-module (oop goops)
>> 	  #:export (<vector2> get-x get-y zero?))
>> 	
>> 	(define-class <vector2> ()
>> 	  (x #:init-value 0 #:getter get-x #:init-keyword #:x)
>> 	  (y #:init-value 0 #:getter get-y #:init-keyword #:y) )
>> 
>> 	(define-generic zero?)
>> 	(define-method (zero? (v <vector2>))
>> 	  (and (zero? (get-x v))
>> 	       (zero? (get-y v))))
>> 
>> In the Guile REPL try executing the following code:
>> 
>> 	scheme@(guile-user)> (use-modules (oop goops) (vector2))
>> 	scheme@(guile-user)> (zero? (make <vector2>))
>> 
>> This will display 
>> 
>> 	WARNING: (guile-user): `zero?' imported from both (ice-9 r5rs) and (vector2)
>> 	ERROR: In procedure scm-error:
>> 	ERROR: No applicable method for #<<generic> zero? (1)> in call (zero? 0)
>> 	
>> 	Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
>> 	scheme@(guile-user) [1]> ,bt
>> 	In vector2.scm:
>> 	     11:7  2 (_ #<<vector2> 105e87e00>)
>> 	In oop/goops.scm:
>> 	   1438:4  1 (cache-miss 0)
>> 	In unknown file:
>> 	           0 (scm-error goops-error #f "No applicable method for ~S in call ~S" (#<<gen…> …) …)
>> 
>> Apparently the problem is that ‘zero?’ is defined in two modules and
>> the vector2 definition overrides it. This isn’t the case with other
>> primitives like ‘+’ or ‘*’, so this seems like a bug? I had built
>> Guile from HEAD a few days ago, my package manager shows 6fff84d as
>> the version number, so I guess that must be the hash of the commit
>> HEAD was pointing to at that time.
> 
> Actually the (vector2) module makes a fresh definition for zero?.  You
> can tell because zero? is in its export list.  So instead of extending
> the primitive-generic that is zero?, you are making a new definition.
> See:
> 
>  scheme@(guile-user)> (define-module (foo) #:export (zero?))
>  $1 = #<directory (foo) 1203c80>
>  scheme@(foo)> (zero? 0)
>  <unnamed port>:4:0: <unnamed port>:4:0: Unbound variable: zero?
> 
>  Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
> 
> If you want to extend a primitive-generic, then do that by not exporting
> zero?.  In a way it's like mutating the primitive in place, giving it
> additional powers.
> 
> Andy
> 
> 
> 
> 
> From: Alejandro Sanchez <hiphish@openmailbox.org>
> Subject: Defining a method named zero? breaks primitive zero?
> Date: 8 March 2017 at 12:07:56 GMT+1
> To: bug-guile@gnu.org
> 
> 
> If I define a ‘zero?’ predicate method for a custom class the primitive ‘zero?’ is lost. Here is a simple vector module:
> 
> 	;;; File vector2.scm
> 	(define-module (vector2)
> 	  #:use-module (oop goops)
> 	  #:export (<vector2> get-x get-y zero?))
> 	
> 	(define-class <vector2> ()
> 	  (x #:init-value 0 #:getter get-x #:init-keyword #:x)
> 	  (y #:init-value 0 #:getter get-y #:init-keyword #:y) )
> 
> 	(define-generic zero?)
> 	(define-method (zero? (v <vector2>))
> 	  (and (zero? (get-x v))
> 	       (zero? (get-y v))))
> 
> In the Guile REPL try executing the following code:
> 
> 	scheme@(guile-user)> (use-modules (oop goops) (vector2))
> 	scheme@(guile-user)> (zero? (make <vector2>))
> 
> This will display 
> 
> 	WARNING: (guile-user): `zero?' imported from both (ice-9 r5rs) and (vector2)
> 	ERROR: In procedure scm-error:
> 	ERROR: No applicable method for #<<generic> zero? (1)> in call (zero? 0)
> 	
> 	Entering a new prompt.  Type `,bt' for a backtrace or `,q' to continue.
> 	scheme@(guile-user) [1]> ,bt
> 	In vector2.scm:
> 	     11:7  2 (_ #<<vector2> 105e87e00>)
> 	In oop/goops.scm:
> 	   1438:4  1 (cache-miss 0)
> 	In unknown file:
> 	           0 (scm-error goops-error #f "No applicable method for ~S in call ~S" (#<<gen…> …) …)
> 
> Apparently the problem is that ‘zero?’ is defined in two modules and the vector2 definition overrides it. This isn’t the case with other primitives like ‘+’ or ‘*’, so this seems like a bug? I had built Guile from HEAD a few days ago, my package manager shows 6fff84d as the version number, so I guess that must be the hash of the commit HEAD was pointing to at that time.
> 
> 
> 


[-- Attachment #2: Type: text/html, Size: 15581 bytes --]