bug#57996: Re[5]: bug#57996: 28.2; imenu doesn't differentiate overloaded c++ functions

["Chris Hecker" <checker@d6.com>]

[-- Attachment #1: Type: text/plain, Size: 6349 bytes --]


Er, wait, this won't handle nested alists, I'll fix it tomorrow.

Chris


------ Original Message ------
From: "Chris Hecker" <checker@d6.com>
To: "Alan Mackenzie" <acm@muc.de>; "Philip Kaludercic" 
<philipk@posteo.net>
Cc: 57996@debbugs.gnu.org
Sent: 2022-10-05 06:45:05
Subject: Re[4]: bug#57996: 28.2; imenu doesn't differentiate overloaded 
c++ functions

>
>Okay, I figured out a way to do this that works pretty easily, and the 
>same idea could be integrated into imenu fairly trivially:
>
>>(defun checker-imenu-make-unique-alist (index-alist)
>>   "De-duplicate the imenu alist by adding indices to duplicate names."
>>   (let ((h (make-hash-table :test #'equal)))
>>     (mapc (lambda (el)
>>             (let* ((k (car el))
>>                    (v (gethash k h)))
>>               (if v
>>                   (let ((n (car v)))
>>                     (if (= n 1)
>>                         (setcar (car (cdr v)) (format "%s(1)" k)))
>>                     (setq n (1+ n))
>>                     (puthash k '(n nil) h)
>>                     (setcar el (format "%s(%d)" k n)))
>>                 (puthash k (list 1 el) h))))
>>           index-alist)))
>>(advice-add 'imenu--truncate-items :filter-return 
>>#'checker-imenu-make-unique-alist)
>
>So this just hooks the imenu--truncate-items function because it was 
>the easiest function that was hookable on the pre-cached side of the 
>code (meaning it doesn't get called every imenu, just on *rescan*).  It 
>puts the function names in a hash, and if they are dupes, it puts (n) 
>after the name.  Works great.
>
>Chris
>
>
>
>------ Original Message ------
>From: "Chris Hecker" <checker@d6.com>
>To: "Alan Mackenzie" <acm@muc.de>; "Philip Kaludercic" 
><philipk@posteo.net>
>Cc: 57996@debbugs.gnu.org
>Sent: 2022-10-05 04:15:12
>Subject: Re[3]: bug#57996: 28.2; imenu doesn't differentiate overloaded 
>c++ functions
>
>>
>>Hmm...
>>
>>item is selected by the user.  This function is called with
>>arguments consisting of the item name, the buffer position, and
>>the ARGUMENTS.
>>
>>This looks like it might work...  I just opened my .emacs and found 
>>I'd already hacked my own version of the c++ matching function, so 
>>maybe I'll try this.
>>
>>A few minutes later...
>>
>>Okay, so this function does indeed get called, but both Function 
>>elements in the imenu list pass the marker for the first Function, so 
>>that's unfortunate...  I will look at that...
>>
>>Another few minutes...
>>
>>Well, it looks like imenu--generic-function generates the right alist 
>>with the two functions and the two different markers, so it's 
>>something about choosing in the buffer...
>>
>>Here's the alist return, looks good:
>>
>>(("Function" . #<marker at 2 in c.cpp>)
>>  ("Function" . #<marker at 41 in c.cpp>)
>>  ("Bar" . #<marker at 95 in c.cpp>))
>>
>>I should sleep, but maybe there's just a bug in the code that selects 
>>the function, or it searches by name instead of by index.   I wish 
>>there was some way for the regex match to return a mangled name...I'll 
>>look into imenu next.
>>
>>Chris
>>
>>------ Original Message ------
>>From: "Chris Hecker" <checker@d6.com>
>>To: "Alan Mackenzie" <acm@muc.de>; "Philip Kaludercic" 
>><philipk@posteo.net>
>>Cc: 57996@debbugs.gnu.org
>>Sent: 2022-10-05 03:47:06
>>Subject: Re[2]: bug#57996: 28.2; imenu doesn't differentiate 
>>overloaded c++ functions
>>
>>>
>>>Yeah, I should probably switch to something a little more modern, but 
>>>imenu has the advantage of just being there and working most of the 
>>>time across all machines and shells and stuff.  It definitely gets 
>>>confused occasionally (like it doesn't find inline functions in class 
>>>declarations) but this overload thing seemed like it might be a 
>>>simple fix.
>>>
>>>
>>>The scanning interface to imenu allows just function names to be
>>>collected.  It doesn't allow anything extra (such as a line number) to
>>>be included into the alist.
>>>
>>>I guess you could mangle the name to include the line number or match 
>>>number...kinda hacky but it'd work...maybe I'll take a look.
>>>
>>>Chris
>>>
>>>
>>>------ Original Message ------
>>>From: "Alan Mackenzie" <acm@muc.de>
>>>To: "Philip Kaludercic" <philipk@posteo.net>
>>>Cc: "Chris Hecker" <checker@d6.com>; 57996@debbugs.gnu.org
>>>Sent: 2022-10-05 03:31:11
>>>Subject: Re: bug#57996: 28.2; imenu doesn't differentiate overloaded 
>>>c++ functions
>>>
>>>>Hello, Chris and Philip.
>>>>
>>>>On Sun, Oct 02, 2022 at 13:13:16 +0000, Philip Kaludercic wrote:
>>>>>  "Chris Hecker" <checker@d6.com> writes:
>>>>
>>>>>  > With this dumb c++ file:
>>>>>  > ----
>>>>>  > int Function( int n ) {
>>>>>  >   return n;
>>>>>  > }
>>>>>  > int Function( float v ) {
>>>>>  >   return (int)(v + 0.5);
>>>>>  > }
>>>>>  > ----
>>>>
>>>>>  > Hitting imenu only gives a single Function entry.  It should probably
>>>>>  > give two, maybe with a line number after them like "Function(123)" or
>>>>>  > whatever.  Currently there's no way to get to the second Function from
>>>>>  > imenu.
>>>>
>>>>imenu is old and rather simplistic.  It parses a buffer, then stores the
>>>>results in an association list.  It then uses the function assoc on that
>>>>list to get "the" match.  What we could do with is a function which gets
>>>>_all_ the matches from an alist, and I've asked on emacs-devel about
>>>>this.
>>>>
>>>>>  Note that this is not the case when using Eglot and a LSP server like
>>>>>  clangd.
>>>>
>>>>Much more modern!
>>>>
>>>>>  I've CC'ed Alan to see if he knows how this could be done by c++-mode
>>>>>  itself.
>>>>
>>>>I'm pretty sure it couldn't be.  I think it would involve enhancing
>>>>imenu.  The scanning interface to imenu allows just function names to be
>>>>collected.  It doesn't allow anything extra (such as a line number) to
>>>>be included into the alist.
>>>>
>>>>I've looked at problems with imenu in C++ Mode before, but got bogged
>>>>down without coming up with a workable solution.  There the problem was
>>>>identically named methods in different classes, or something like that.
>>>>
>>>>So, maybe we can enhance imenu.  But not for Emacs 29.
>>>>
>>>>--
>>>>Alan Mackenzie (Nuremberg, Germany).

[-- Attachment #2: Type: text/html, Size: 11719 bytes --]