add `read-buffer-completion-ignore-case' ?

add `read-buffer-completion-ignore-case' ?

[John Paul Wallington]

Is it okay to install the following change?

It adds a customizable variable, `read-buffer-completion-ignore-case',
that specifies the case-sensitivity of completion for `read-buffer'.
Its default value is nil, meaning that `read-buffer' is case-sensitive
when completing buffer names.

2008-06-08  John Paul Wallington  <jpw@pobox.com>

	* minibuf.c (read_buffer_completion_ignore_case): New variable.
	(syms_of_minibuf): Declare and initialise it.
	(Fread_buffer): Bind `completion-ignore-case' to respect it.
	Mention it and `read-buffer-function' in docstring.

--- minibuf.c.~1.353.~	2008-06-05 20:28:48.000000000 +0100
+++ minibuf.c	2008-06-08 22:10:01.000000000 +0100
@@ -109,6 +109,9 @@
 /* Function to call to read a buffer name.  */
 Lisp_Object Vread_buffer_function;
 
+/* Nonzero means completion ignores case when reading buffer name.  */
+int read_buffer_completion_ignore_case;
+
 /* Nonzero means completion ignores case.  */
 
 int completion_ignore_case;
@@ -1178,7 +1181,9 @@
  If DEF is a list of default values, return its first element.
 If optional third arg REQUIRE-MATCH is non-nil,
  only existing buffer names are allowed.
-The argument PROMPT should be a string ending with a colon and a space.  */)
+The argument PROMPT should be a string ending with a colon and a space.
+
+See also `read-buffer-completion-ignore-case' and `read-buffer-function'.  */)
      (prompt, def, require_match)
      Lisp_Object prompt, def, require_match;
 {
@@ -1189,47 +1194,54 @@
   if (BUFFERP (def))
     def = XBUFFER (def)->name;
 
-  if (NILP (Vread_buffer_function))
-    {
-      if (!NILP (def))
-	{
-	  /* A default value was provided: we must change PROMPT,
-	     editing the default value in before the colon.  To achieve
-	     this, we replace PROMPT with a substring that doesn't
-	     contain the terminal space and colon (if present).  They
-	     are then added back using Fformat.  */
+  {
+    int count = SPECPDL_INDEX ();
+    
+    specbind (Qcompletion_ignore_case,
+	      read_buffer_completion_ignore_case ? Qt : Qnil);
 
-	  if (STRINGP (prompt))
-	    {
-	      s = SDATA (prompt);
-	      len = strlen (s);
-	      if (len >= 2 && s[len - 2] == ':' && s[len - 1] == ' ')
-		len = len - 2;
-	      else if (len >= 1 && (s[len - 1] == ':' || s[len - 1] == ' '))
-		len--;
+    if (NILP (Vread_buffer_function))
+      {
+	if (!NILP (def))
+	  {
+	    /* A default value was provided: we must change PROMPT,
+	       editing the default value in before the colon.  To achieve
+	       this, we replace PROMPT with a substring that doesn't
+	       contain the terminal space and colon (if present).  They
+	       are then added back using Fformat.  */
 
-	      prompt = make_specified_string (s, -1, len,
-					      STRING_MULTIBYTE (prompt));
-	    }
+	    if (STRINGP (prompt))
+	      {
+		s = SDATA (prompt);
+		len = strlen (s);
+		if (len >= 2 && s[len - 2] == ':' && s[len - 1] == ' ')
+		  len = len - 2;
+		else if (len >= 1 && (s[len - 1] == ':' || s[len - 1] == ' '))
+		  len--;
 
-	  args[0] = build_string ("%s (default %s): ");
-	  args[1] = prompt;
-	  args[2] = CONSP (def) ? XCAR (def) : def;
-	  prompt = Fformat (3, args);
-	}
+		prompt = make_specified_string (s, -1, len,
+						STRING_MULTIBYTE (prompt));
+	      }
 
-      return Fcompleting_read (prompt, intern ("internal-complete-buffer"),
-			       Qnil, require_match, Qnil, Qbuffer_name_history,
-			       def, Qnil);
-    }
-  else
-    {
-      args[0] = Vread_buffer_function;
-      args[1] = prompt;
-      args[2] = def;
-      args[3] = require_match;
-      return Ffuncall(4, args);
-    }
+	    args[0] = build_string ("%s (default %s): ");
+	    args[1] = prompt;
+	    args[2] = CONSP (def) ? XCAR (def) : def;
+	    prompt = Fformat (3, args);
+	  }
+
+	return Fcompleting_read (prompt, intern ("internal-complete-buffer"),
+				 Qnil, require_match, Qnil, Qbuffer_name_history,
+				 def, Qnil);
+      }
+    else
+      {
+	args[0] = Vread_buffer_function;
+	args[1] = prompt;
+	args[2] = def;
+	args[3] = require_match;
+	return Ffuncall(4, args);
+      }
+  }
 }
 \f
 static Lisp_Object
@@ -2117,6 +2129,10 @@
 	       doc: /* If this is non-nil, `read-buffer' does its work by calling this function.  */);
   Vread_buffer_function = Qnil;
 
+  DEFVAR_BOOL ("read-buffer-completion-ignore-case", &read_buffer_completion_ignore_case,
+	       doc: /* *Non-nil means when reading a buffer name completion ignores case.  */);
+  read_buffer_completion_ignore_case = 0;
+
   DEFVAR_LISP ("minibuffer-setup-hook", &Vminibuffer_setup_hook,
 	       doc: /* Normal hook run just after entry to minibuffer.  */);
   Vminibuffer_setup_hook = Qnil;



2008-06-08  John Paul Wallington  <jpw@pobox.com>

	* cus-start.el (read-buffer-completion-ignore-case): Add.

--- cus-start.el.~1.123.~	2008-06-05 20:28:43.000000000 +0100
+++ cus-start.el	2008-06-08 22:17:50.000000000 +0100
@@ -275,6 +275,7 @@
 				   (choice (const nil)
 					   (function-item iswitchb-read-buffer)
 					   function))
+	     (read-buffer-completion-ignore-case minibuffer boolean "23.1")
 	     ;; msdos.c
 	     (dos-unsupported-char-glyph display integer)
 	     ;; process.c

RE: add `read-buffer-completion-ignore-case' ?

[Drew Adams]

> Is it okay to install the following change?

I didn't try it, but if it does what you say then it's OK by me.

> It adds a customizable variable, `read-buffer-completion-ignore-case',
> that specifies the case-sensitivity of completion for `read-buffer'.
> Its default value is nil, meaning that `read-buffer' is case-sensitive
> when completing buffer names.

t might be a better default value, as some others explained. t would also be my
preference, but I'm OK with either t or nil.

Thx.

doc strings for read-*-completion-ignore-case [was: add `read-buffer-completion-ignore-case' ?]

[Drew Adams]

> It adds a customizable variable, `read-buffer-completion-ignore-case',
> that specifies the case-sensitivity of completion for `read-buffer'.
> Its default value is nil, meaning that `read-buffer' is case-sensitive
> when completing buffer names.

Actually, the doc strings for both variables (file and buffer) seem incomplete
(and even not quite correct) to me: "Non-nil means when reading a buffer name
completion ignores case."

I would like to see a little more, saying explicitly that this means only when
using `read-buffer' or an `interactive' spec with spec "B..." or "b...".
Similarly for file names and `read-file-name'.

The reason:

1. It's not obvious that if you read a file name using, say, `completing-read'
with file-name candidates as the TABLE arg, this variable has no effect. You
must use `read-file-name'. Similarly for reading a buffer name using
`completing-read' with a list of buffer names as TABLE arg - the variable has no
effect.

2. It's not obvious that (interactive "B...") uses `read-buffer', and so on.

At least the latter was not obvious to me. I had to study the C code just now to
see how it is that a change to `read-buffer' also affects (interactive "B...").

Both #1 and #2 are important to know for those programming in Emacs Lisp. #1 is
presumably the reason for the name `read-buffer-...' instead of
`read-buffer-name-...'.

And #1 is presumably the reason for both names instead of, say, just
`file-name-completion-ignore-case' and `buffer-name-completion-ignore-case'.
These variables do not, _in general_, affect completion when reading a buffer
name or a file name - they are limited to `read-buffer' and `read-file-name'
(and `interactive'). 

The doc strings should help dispel the idea that this is about completion of
buffer/file names - it is about `read-buffer' and `read-file-name',
specifically.

---

Also, the doc strings are not great grammatically. It would be clearer to say:
"Non-nil means that file-name completion ignores case." 

Or, if you really want to keep the current order (which is less clear, IMO),
then: "Non-nil means that, when reading a file name, completion ignores case."
The "that" here helps understanding, especially for those whose mother tongue is
not English.

Re: add `read-buffer-completion-ignore-case' ?

[Stefan Monnier]

> Is it okay to install the following change?

I don't think we need yet-another configuration variable for that.
We should just reuse read-file-name-completion-ignore-case.


        Stefan

Re: add `read-buffer-completion-ignore-case' ?

[Miles Bader]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
> I don't think we need yet-another configuration variable for that.
> We should just reuse read-file-name-completion-ignore-case.

Regardless of what some newbies think, buffer names and filenames are
not the same.  It is quite reasonable for someone to want to have one be
case-insensitive but not the other.

-Miles

-- 
Kilt, n. A costume sometimes worn by Scotchmen [sic] in America and Americans
in Scotland.

RE: add `read-buffer-completion-ignore-case' ?

[Drew Adams]

> Stefan Monnier <monnier@iro.umontreal.ca> writes:
> > I don't think we need yet-another configuration variable for that.
> > We should just reuse read-file-name-completion-ignore-case.
> 
> Regardless of what some newbies think, buffer names and filenames are
> not the same.  It is quite reasonable for someone to want to 
> have one be case-insensitive but not the other.

1+

Re: add `read-buffer-completion-ignore-case' ?

[Eli Zaretskii]

> From: Miles Bader <miles.bader@necel.com>
> Date: Mon, 09 Jun 2008 11:59:21 +0900
> Cc: emacs-devel@gnu.org, John Paul Wallington <jpw@pobox.com>
> 
> Stefan Monnier <monnier@iro.umontreal.ca> writes:
> > I don't think we need yet-another configuration variable for that.
> > We should just reuse read-file-name-completion-ignore-case.
> 
> Regardless of what some newbies think, buffer names and filenames are
> not the same.  It is quite reasonable for someone to want to have one be
> case-insensitive but not the other.

I don't necessarily disagree, but we do know which buffers are
associated with a file and which aren't.  So we could DWIM in each
case as appropriate.  If we want that, of course.

Re: add `read-buffer-completion-ignore-case' ?

[David Kastrup]

Eli Zaretskii <eliz@gnu.org> writes:

>> From: Miles Bader <miles.bader@necel.com>
>> Date: Mon, 09 Jun 2008 11:59:21 +0900
>> Cc: emacs-devel@gnu.org, John Paul Wallington <jpw@pobox.com>
>> 
>> Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> > I don't think we need yet-another configuration variable for that.
>> > We should just reuse read-file-name-completion-ignore-case.
>> 
>> Regardless of what some newbies think, buffer names and filenames are
>> not the same.  It is quite reasonable for someone to want to have one be
>> case-insensitive but not the other.
>
> I don't necessarily disagree, but we do know which buffers are
> associated with a file and which aren't.  So we could DWIM in each
> case as appropriate.  If we want that, of course.

I don't think we do.  For non-file buffers, I don't really see that
having different completion realms for "*Messages*" and "*mail*" is
really an essential part of people's workflow.  So it is not that
important to differentiate here as well.

What about buffers associated with a file, but with a file name not
matching the buffer name?

There is so much potential for messiness that I would not want to go
there.

-- 
David Kastrup

Re: add `read-buffer-completion-ignore-case' ?

[Juanma Barranquero]

On Mon, Jun 9, 2008 at 10:45, Eli Zaretskii <eliz@gnu.org> wrote:

> I don't necessarily disagree, but we do know which buffers are
> associated with a file and which aren't.  So we could DWIM in each
> case as appropriate.  If we want that, of course.

I like switching buffers in a case-insensitive way, even when the
buffer is not visiting a file. It's not that often that I have
*Messages* and *messages* and *Scratch* and *scratch*, etc... Like,
never, really.

So it's better IMO not to tie buffer-selection casefolding with
filename casefolding (I'd set both to case-insensitivity at once, but
it's clear some people would prefer otherwise).

   Juanma

Re: add `read-buffer-completion-ignore-case' ?

[John Paul Wallington]

I wrote:

> Is it okay to install the following change?
>
> It adds a customizable variable, `read-buffer-completion-ignore-case',
> that specifies the case-sensitivity of completion for `read-buffer'.
> Its default value is nil, meaning that `read-buffer' is case-sensitive
> when completing buffer names.
>
> 2008-06-08  John Paul Wallington  <jpw@pobox.com>
>
> 	* minibuf.c (read_buffer_completion_ignore_case): New variable.
> 	(syms_of_minibuf): Declare and initialise it.
> 	(Fread_buffer): Bind `completion-ignore-case' to respect it.
> 	Mention it and `read-buffer-function' in docstring.
>
> --- minibuf.c.~1.353.~	2008-06-05 20:28:48.000000000 +0100
> +++ minibuf.c	2008-06-08 22:10:01.000000000 +0100
[...]

Doh.  That patch didn't work, 'cos I forgot to unbind_to at the end.
Here's a revised one:

--- minibuf.c.~1.353.~	2008-06-05 20:28:48.000000000 +0100
+++ minibuf.c	2008-06-09 12:57:35.000000000 +0100
@@ -109,6 +109,9 @@
 /* Function to call to read a buffer name.  */
 Lisp_Object Vread_buffer_function;
 
+/* Nonzero means completion ignores case when reading buffer name.  */
+int read_buffer_completion_ignore_case;
+
 /* Nonzero means completion ignores case.  */
 
 int completion_ignore_case;
@@ -1178,17 +1181,25 @@
  If DEF is a list of default values, return its first element.
 If optional third arg REQUIRE-MATCH is non-nil,
  only existing buffer names are allowed.
-The argument PROMPT should be a string ending with a colon and a space.  */)
+The argument PROMPT should be a string ending with a colon and a space.
+
+See also `read-buffer-completion-ignore-case' and `read-buffer-function'. */)
      (prompt, def, require_match)
      Lisp_Object prompt, def, require_match;
 {
   Lisp_Object args[4];
+  Lisp_Object tem;
   unsigned char *s;
   int len;
+  int count = SPECPDL_INDEX ();
+    
 
   if (BUFFERP (def))
     def = XBUFFER (def)->name;
-
+  
+  specbind (Qcompletion_ignore_case,
+	    read_buffer_completion_ignore_case ? Qt : Qnil);
+  
   if (NILP (Vread_buffer_function))
     {
       if (!NILP (def))
@@ -1218,9 +1229,9 @@
 	  prompt = Fformat (3, args);
 	}
 
-      return Fcompleting_read (prompt, intern ("internal-complete-buffer"),
-			       Qnil, require_match, Qnil, Qbuffer_name_history,
-			       def, Qnil);
+      tem = Fcompleting_read (prompt, intern ("internal-complete-buffer"),
+			      Qnil, require_match, Qnil, Qbuffer_name_history,
+			      def, Qnil);
     }
   else
     {
@@ -1228,8 +1239,9 @@
       args[1] = prompt;
       args[2] = def;
       args[3] = require_match;
-      return Ffuncall(4, args);
+      tem = Ffuncall(4, args);
     }
+  return unbind_to (count, tem);
 }
 \f
 static Lisp_Object
@@ -2117,6 +2129,10 @@
 	       doc: /* If this is non-nil, `read-buffer' does its work by calling this function.  */);
   Vread_buffer_function = Qnil;
 
+  DEFVAR_BOOL ("read-buffer-completion-ignore-case", &read_buffer_completion_ignore_case,
+	       doc: /* *Non-nil means when reading a buffer name completion ignores case.  */);
+  read_buffer_completion_ignore_case = 0;
+
   DEFVAR_LISP ("minibuffer-setup-hook", &Vminibuffer_setup_hook,
 	       doc: /* Normal hook run just after entry to minibuffer.  */);
   Vminibuffer_setup_hook = Qnil;

Re: add `read-buffer-completion-ignore-case' ?

[Stefan Monnier]

>> I don't think we need yet-another configuration variable for that.
>> We should just reuse read-file-name-completion-ignore-case.
> Regardless of what some newbies think, buffer names and filenames are
> not the same.  It is quite reasonable for someone to want to have one be
> case-insensitive but not the other.

Maybe it's reasonable, but I don't think it matters.
So this is irrelevant without a more concrete reason.


        Stefan

Re: add `read-buffer-completion-ignore-case' ?

[Miles Bader]

Stefan Monnier <monnier@iro.umontreal.ca> writes:
>> Regardless of what some newbies think, buffer names and filenames are
>> not the same.  It is quite reasonable for someone to want to have one be
>> case-insensitive but not the other.
>
> Maybe it's reasonable, but I don't think it matters.
> So this is irrelevant without a more concrete reason.

I've found case-insensitivity in buffer-name matching generated a lot of
annoying "conflicts" due to non-file buffers (e.g. *Group* and *grep*).

[I currently use case-insensitive at work, and case-sensitive at home :-]

Since there wasn't much choice in the past, I've not experimented with
other combinations, but I'd like to.

-Miles

-- 
A zen-buddhist walked into a pizza shop and
said, "Make me one with everything."

Re: add `read-buffer-completion-ignore-case' ?

[Stefan Monnier]

>>> Regardless of what some newbies think, buffer names and filenames are
>>> not the same.  It is quite reasonable for someone to want to have one be
>>> case-insensitive but not the other.
>> 
>> Maybe it's reasonable, but I don't think it matters.
>> So this is irrelevant without a more concrete reason.

> I've found case-insensitivity in buffer-name matching generated a lot of
> annoying "conflicts" due to non-file buffers (e.g. *Group* and *grep*).

> [I currently use case-insensitive at work, and case-sensitive at home :-]

> Since there wasn't much choice in the past, I've not experimented with
> other combinations, but I'd like to.

But that'll only affect people who've always used case-insensitive
matching everywhere until now.


        Stefan