Improvement to message-x.el : Header name completion.

Improvement to message-x.el : Header name completion.

[Matthieu Moy]

Hi,

I've got an interesting small piece of code to submit :

Both message-x.el  and message-tab  provide completion in  the headers
content, but none of them provide completion on the header's *name*. 

For example, when I want to  add a followup-to field, I never remember
the shortcut,  (and also  never remember wether  there are one  or two
`l' ;-) so I have to search it in the menu. 

With this, just type "Fo TAB" and you get the field automagically. 

I've done it in message-x.el, which I'm still using because TAB brings
me to the  body when I wish, but the  code should be cut-and-past-able
to message.el.

I've signed the FSF Papers, so you can use this code as you wish. 

-- 
Matthieu


--- message-x.el.orig	Wed Sep 25 17:23:08 2002
+++ emacs-lisp/message-x.el	Wed Sep 25 17:16:56 2002
@@ -3,7 +3,10 @@
 
 ;; $Id: message-x.el,v 1.23 2001/05/30 21:04:47 grossjoh Exp $
 
-;; Author: Kai Grossjohann <grossjohann@ls6.informatik.uni-dortmund.de>
+;; Author: Kai Grossjohann
+;; <grossjohann@ls6.informatik.uni-dortmund.de>
+;; Modified by Matthieu MOY on September 25th 2002 to add completion
+;; in header names.
 ;; Keywords: news, mail, compose, completion
 
 ;; This file is not part of GNU Emacs.
@@ -217,6 +220,25 @@
                   t)
                  (t nil))))))
 
+(defvar message-x-all-headers '("To" "Cc" "Bcc" "Gcc" "Newsgroups"
+				"From" "Subject" "Mail-Followup-To"
+				"Reply-To" "Followup-To"
+				"Keywords" "Summary" 
+				"Distribution")
+  "List of header names availlable for completion."
+  )
+
+(defvar message-x-all-headers-completion-alist
+  (mapcar 
+   (lambda (x)
+     `(,(concat x ": ")))
+   message-x-all-headers)
+  "Simple transformation of `message-x-all-headers' used for completion.")
+
+(defvar message-x-all-headers-regexp
+  (regexp-opt message-x-all-headers)
+  "Regexp matching the name of a header.")
+
 (defun message-x-tab (&optional skip-completion)
   "Smart completion or indentation in message buffers.
 
@@ -234,30 +256,57 @@
 `message-x-call-completion-function' for details on how to check
 whether the completion function has done something.
 
+Completion also works on headers names. For example, to insert a
+Followup-To field, type \"Fo TAB\" which will be expanded into
+\"Followup-To: \".
+
 A non-nil optional arg SKIP-COMPLETION (prefix arg if invoked
 interactively) means to not attempt completion.  Instead,
 `message-x-unknown-header-function' function is called in all headers,
 known or unknown."
   (interactive "P")
-  (let* ((alist message-x-completion-alist)
-         (which-header (message-x-which-header))
-         header)
-    (run-hook-with-args 'message-x-before-completion-functions which-header)
-    (while (and (not skip-completion)
-                alist
-                (let ((mail-abbrev-mode-regexp (eval (caar alist))))
-                  (not (mail-abbrev-in-expansion-header-p))))
-      (setq alist (cdr alist)))
-    (cond ((and alist (not skip-completion))
-           (let ((p (point))
-                 (func (cdar alist)))
-             (unless (message-x-call-completion-function func)
-               (funcall message-x-unknown-header-function))))
-          ((message-x-in-header-p)
-           (funcall message-x-unknown-header-function))
-          (t
-           (funcall message-x-body-function)))
-    (run-hook-with-args 'message-x-after-completion-functions which-header)))
+  (let* ((begin (point))
+	 (end (save-excursion (beginning-of-line) (point)))
+	 (to-be-completed (buffer-substring-no-properties begin end)))
+    (cond ((string-match "[a-zA-Z]*\: .*" to-be-completed)
+	  (let* ((alist message-x-completion-alist)
+		 (which-header (message-x-which-header))
+		 header)
+	    (run-hook-with-args 'message-x-before-completion-functions which-header)
+	    (while (and (not skip-completion)
+			alist
+			(let ((mail-abbrev-mode-regexp (eval (caar alist))))
+			  (not (mail-abbrev-in-expansion-header-p))))
+	      (setq alist (cdr alist)))
+	    (cond ((and alist (not skip-completion))
+		   (let ((p (point))
+			 (func (cdar alist)))
+		     (unless (message-x-call-completion-function func)
+		       (funcall message-x-unknown-header-function))))
+		  ((message-x-in-header-p)
+		   (funcall message-x-unknown-header-function))
+		  (t
+		   (funcall message-x-body-function)))
+	    (run-hook-with-args 'message-x-after-completion-functions
+				which-header))
+	  )
+	  (t (let ((completion (try-completion to-be-completed message-x-all-headers-completion-alist)))
+	       (cond 
+		((save-excursion (beginning-of-line)
+				 (looking-at (concat "\\("
+						     message-x-all-headers-regexp
+						     "\\): ")))
+		 ;; We are inside a valid header name. Goto it's
+		 ;; value (after the `: ')
+		 (goto-char (match-end 0)))
+		((stringp completion) ; Completion is possible. Do it
+		 (progn
+		   (delete-region begin end)
+		   (insert completion)))
+		(completion (message "Complete field name"))
+		(t (error "no completion")))
+	       ))
+	  )))
 
 (define-key message-mode-map "\t" 'message-x-tab)
 

Re: Improvement to message-x.el : Header name completion.

[Lars Magne Ingebrigtsen]

Matthieu Moy <Matthieu.Moy@imag.fr> writes:

> Both message-x.el  and message-tab  provide completion in  the headers
> content, but none of them provide completion on the header's *name*. 

I forget now -- why isn't message-x.el in the Gnus distribution?

> For example, when I want to  add a followup-to field, I never remember
> the shortcut,  (and also  never remember wether  there are one  or two
> `l' ;-) so I have to search it in the menu. 
>
> With this, just type "Fo TAB" and you get the field automagically. 

This seems quite convenient, so it would be nice to include it...

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Improvement to message-x.el : Header name completion.

[Kai Großjohann]

[-- Attachment #1: Type: text/plain, Size: 1037 bytes --]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> Matthieu Moy <Matthieu.Moy@imag.fr> writes:
>
>> Both message-x.el  and message-tab  provide completion in  the headers
>> content, but none of them provide completion on the header's *name*. 
>
> I forget now -- why isn't message-x.el in the Gnus distribution?

message-x.el offers similar functionality to message-tab, except it
has more hair.

I think that some people like the hair, but I decided that down this
path lies madness.

Specifically, message-x can advance point to the next header if there
is nothing to complete in the current header.  This allows people to
type an email address using tab completion, then hit tab again to go
to the Subject header (say), then use tab again to go to the body.

In principle, it is a desirable feature.  But it is really really
difficult to implement correctly, and I'd rather not perpetuate that
hairy stuff.

But how about folks have a look at message-x to get their own
impression.  It's not that long...
-- 
Ambibibentists unite!


[-- Attachment #2: message-x.el --]
[-- Type: application/emacs-lisp, Size: 10795 bytes --]

Re: Improvement to message-x.el : Header name completion.

[Lars Magne Ingebrigtsen]

kai.grossjohann@uni-duisburg.de (Kai Großjohann) writes:

> I think that some people like the hair, but I decided that down this
> path lies madness.

Well, has that ever stopped us before?

> Specifically, message-x can advance point to the next header if there
> is nothing to complete in the current header.  This allows people to
> type an email address using tab completion, then hit tab again to go
> to the Subject header (say), then use tab again to go to the body.

I think that's neat.  How about adding the patch that was posted up
there in the root of this thread, adding a variable to message.el
that allows selection between traditional message-tab and message-x,
and including it in the Gnus distribution?

The name of the file/functions aren't all that, er, intuitive.
message-x sounds like it does unknown and mysterious things, which
might be true enough, but how about messagetab.el or something like
that?  Difficult to find a name thats 13 characters or less...

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen

Re: Improvement to message-x.el : Header name completion.

[Henrik Enberg]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> The name of the file/functions aren't all that, er, intuitive.
> message-x sounds like it does unknown and mysterious things, which
> might be true enough, but how about messagetab.el or something like
> that?  Difficult to find a name thats 13 characters or less...

OTOH, it follows Emacs conventions, i.e. dired-x.el.

Re: Improvement to message-x.el : Header name completion.

[Kai Großjohann]

Lars Magne Ingebrigtsen <larsi@gnus.org> writes:

> I think that's neat.  How about adding the patch that was posted up
> there in the root of this thread,

Good idea, modulo paperwork.

> adding a variable to message.el that allows selection between
> traditional message-tab and message-x, and including it in the Gnus
> distribution?

Why not add the code (it's not that much) to message.el.  Yeah, I'll
put that on my todo list.
-- 
Ambibibentists unite!

Re: Improvement to message-x.el : Header name completion.

[Lars Magne Ingebrigtsen]

kai.grossjohann@uni-duisburg.de (Kai Großjohann) writes:

>> I think that's neat.  How about adding the patch that was posted up
>> there in the root of this thread,
>
> Good idea, modulo paperwork.

He said that he had signed the FSF paperwork, I think...

> Why not add the code (it's not that much) to message.el.  Yeah, I'll
> put that on my todo list.

Cool.

-- 
(domestic pets only, the antidote for overdose, milk.)
   larsi@gnus.org * Lars Magne Ingebrigtsen