Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Stefan Monnier]

> +(defun gensym (&optional prefix)
> +  "Return a new uninterned symbol.
> +The name is made by appending `gensym-counter' to PREFIX.
> +PREFIX can be a string, and defaults to \"G\".
> +If PREFIX is a number, it replaces the value of `gensym-counter'."

I understand that Common-Lisp has that "PREFIX is a number"
functionality, but I could never find a use-case for it, so I wonder why
we carry over that baggage to Elisp?


        Stefan

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Mark Oteiza]

On 12/09/17 at 12:48pm, Stefan Monnier wrote:
> > +(defun gensym (&optional prefix)
> > +  "Return a new uninterned symbol.
> > +The name is made by appending `gensym-counter' to PREFIX.
> > +PREFIX can be a string, and defaults to \"G\".
> > +If PREFIX is a number, it replaces the value of `gensym-counter'."
> 
> I understand that Common-Lisp has that "PREFIX is a number"
> functionality, but I could never find a use-case for it, so I wonder why
> we carry over that baggage to Elisp?

It was just simpler to not duplicate the code for such a small
historical tidbit.  I have no problem with the idea of dropping it.

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Mark Oteiza]

On 12/09/17 at 01:21pm, Mark Oteiza wrote:
> On 12/09/17 at 12:48pm, Stefan Monnier wrote:
> > > +(defun gensym (&optional prefix)
> > > +  "Return a new uninterned symbol.
> > > +The name is made by appending `gensym-counter' to PREFIX.
> > > +PREFIX can be a string, and defaults to \"G\".
> > > +If PREFIX is a number, it replaces the value of `gensym-counter'."
> > 
> > I understand that Common-Lisp has that "PREFIX is a number"
> > functionality, but I could never find a use-case for it, so I wonder why
> > we carry over that baggage to Elisp?
> 
> It was just simpler to not duplicate the code for such a small
> historical tidbit.  I have no problem with the idea of dropping it.

Is this what you mean? (I lowered the prefix for good measure)

diff --git a/doc/lispref/symbols.texi b/doc/lispref/symbols.texi
index 2d9ec6fda3..cda5f1c40f 100644
--- a/doc/lispref/symbols.texi
+++ b/doc/lispref/symbols.texi
@@ -276,8 +276,7 @@ Creating Symbols
 @defun gensym &optional prefix
 This function returns a symbol using @code{make-symbol}, whose name is
 made by appending @code{gensym-counter} to @var{prefix}.  The prefix
-defaults to @code{"G"}.  If @var{prefix} is a number, it replaces the
-value of the counter.
+defaults to @code{"g"}.
 @end defun
 
 @defun intern name &optional obarray
diff --git a/lisp/emacs-lisp/cl-macs.el b/lisp/emacs-lisp/cl-macs.el
index eee5953882..3405c92e8d 100644
--- a/lisp/emacs-lisp/cl-macs.el
+++ b/lisp/emacs-lisp/cl-macs.el
@@ -161,9 +161,16 @@ cl--expr-depends-p
 
 ;;; Symbols.
 
-(defvaralias 'cl--gensym-counter 'gensym-counter)
+(defvar cl--gensym-counter 0)
 ;;;###autoload
-(cl--defalias 'cl-gensym 'gensym)
+(defun cl-gensym (&optional prefix)
+  "Generate a new uninterned symbol.
+The name is made by appending a number to PREFIX, default \"G\"."
+  (let ((pfix (if (stringp prefix) prefix "G"))
+	(num (if (integerp prefix) prefix
+	       (prog1 cl--gensym-counter
+		 (setq cl--gensym-counter (1+ cl--gensym-counter))))))
+    (make-symbol (format "%s%d" pfix num))))
 
 (defvar cl--gentemp-counter 0)
 ;;;###autoload
diff --git a/lisp/subr.el b/lisp/subr.el
index ebb8b53b50..52d4e190e7 100644
--- a/lisp/subr.el
+++ b/lisp/subr.el
@@ -286,13 +286,10 @@ gensym-counter
 (defun gensym (&optional prefix)
   "Return a new uninterned symbol.
 The name is made by appending `gensym-counter' to PREFIX.
-PREFIX can be a string, and defaults to \"G\".
-If PREFIX is a number, it replaces the value of `gensym-counter'."
-  (let ((pfix (if (stringp prefix) prefix "G"))
-        (num (if (integerp prefix) prefix
-               (prog1 gensym-counter
-                 (setq gensym-counter (1+ gensym-counter))))))
-    (make-symbol (format "%s%d" pfix num))))
+PREFIX is a string, and defaults to \"g\"."
+  (let ((num (prog1 gensym-counter
+               (setq gensym-counter (1+ gensym-counter)))))
+    (make-symbol (format "%s%d" prefix num))))
 
 (defun ignore (&rest _ignore)
   "Do nothing and return nil.

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Stefan Monnier]

> It was just simpler to not duplicate the code for such a small
> historical tidbit.

Damn!  So still no explanation for that weird functionality.


        Stefan

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Mark Oteiza]

On 12/09/17 at 02:28pm, Stefan Monnier wrote:
> > It was just simpler to not duplicate the code for such a small
> > historical tidbit.
> 
> Damn!  So still no explanation for that weird functionality.

Nope, the best answer I have is what the CL spec says :(

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Alex]

Stefan Monnier <monnier@IRO.UMontreal.CA> writes:

>> It was just simpler to not duplicate the code for such a small
>> historical tidbit.
>
> Damn!  So still no explanation for that weird functionality.
>
>
>         Stefan

According to [1] (page 53), MACLISP didn't have a visible
gensym-counter, and instead used an integer argument to gensym to allow
programs to set an internal counter. According to [2], it was only
decided in 1989 to have *gensym-counter* in CL, so before then having an
integer argument made sense if you wanted to manipulate the gensym
state.

I don't think an integer argument makes much sense when there's a
visible counter available. Checking it only in cl-gensym and not in
gensym is a nice idea, but I wonder if people who use an integer
argument here are the type of people to use `cl' instead of `cl-lib'.

Footnotes: 
[1]  http://www.softwarepreservation.org/projects/LISP/MIT/Moon-MACLISP_Reference_Manual-Apr_08_1974.pdf

[2]  https://www.cs.cmu.edu/Groups/AI/html/cltl/clm/node110.html

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Stefan Monnier]

> According to [1] (page 53), MACLISP didn't have a visible
> gensym-counter, and instead used an integer argument to gensym to allow
> programs to set an internal counter. According to [2], it was only
> decided in 1989 to have *gensym-counter* in CL, so before then having an
> integer argument made sense if you wanted to manipulate the gensym
> state.

An integer arg that lets you set the internal counter makes some
sense, yes.  But the integer arg implements in cl-lib.el (and in clisp,
FWIW) does not affect the internal counter.

> but I wonder if people who use an integer argument here are the type
> of people to use `cl' instead of `cl-lib'.

I think using the plural here is rather optimistic.  I'd already be
surprised if there's even just one user of that functionality.


        Stefan

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Mark Oteiza]

On 13/09/17 at 08:40am, Stefan Monnier wrote:
> > According to [1] (page 53), MACLISP didn't have a visible
> > gensym-counter, and instead used an integer argument to gensym to allow
> > programs to set an internal counter. According to [2], it was only
> > decided in 1989 to have *gensym-counter* in CL, so before then having an
> > integer argument made sense if you wanted to manipulate the gensym
> > state.

Thanks for digging that up.

> > but I wonder if people who use an integer argument here are the type
> > of people to use `cl' instead of `cl-lib'.
> 
> I think using the plural here is rather optimistic.  I'd already be
> surprised if there's even just one user of that functionality.

Only one way to find out! I suppose I'll apply the patch I sent earlier.

Re: [Emacs-diffs] master 35c893d 2/2: Move gensym to core Elisp

[Alex]

Mark Oteiza <mvoteiza@udel.edu> writes:

> On 13/09/17 at 08:40am, Stefan Monnier wrote:
>> > According to [1] (page 53), MACLISP didn't have a visible
>> > gensym-counter, and instead used an integer argument to gensym to allow
>> > programs to set an internal counter. According to [2], it was only
>> > decided in 1989 to have *gensym-counter* in CL, so before then having an
>> > integer argument made sense if you wanted to manipulate the gensym
>> > state.
>
> Thanks for digging that up.
>
>> > but I wonder if people who use an integer argument here are the type
>> > of people to use `cl' instead of `cl-lib'.
>> 
>> I think using the plural here is rather optimistic.  I'd already be
>> surprised if there's even just one user of that functionality.
>
> Only one way to find out! I suppose I'll apply the patch I sent earlier.

It doesn't really matter, but wouldn't it be a bit nicer if cl-gensym
and gensym shared the same counter and default prefix?