bug#55636: 27.2; etags performance fix when working with very big TAGS files

[VAN VLIERBERGHE Stef]

Hi Eli,

For us the 10 sec is reduced to below 1 sec, loading the tags file is no longer 
noticed after this change.

I assume the reason is a huge amount of files all accessed over NFS, and 
expand-file-name does a lot of system calls that translate into network packets.

An alternative approach is to add some switch that allows a customization that 
simply never calls the expand-file-name, we generate tags files that already 
contain absolute paths so don't need any of this logic and disabling it would 
also be ok for us.

All the best,
Stef

-----Original Message-----
From: Eli Zaretskii <eliz@gnu.org>
Sent: 25 May 2022 19:16
To: DE BACKER Jurgen (EXT) <jurgen.de-backer.ext@eurocontrol.int>
Cc: 55636@debbugs.gnu.org; VAN VLIERBERGHE Stef 
<stef.van-vlierberghe@eurocontrol.int>; WAROQUIERS Philippe 
<philippe.waroquiers@eurocontrol.int>
Subject: Re: bug#55636: 27.2; etags performance fix when working with very big 
TAGS files

> Cc: jurgen.de-backer.ext@eurocontrol.int,
> stef.van-vlierberghe@eurocontrol.int,
>  philippe.waroquiers@eurocontrol.int
> Date: Wed, 25 May 2022 16:04:04 +0000
> From:  Jurgen De Backer via "Bug reports for GNU Emacs,  the Swiss
> army knife of text editors" <bug-gnu-emacs@gnu.org>
>
> We implemented a fix for making etags searching faster in our custom
> emacs build.
> In our project, sometimes a search could take more than 10 seconds.
> The fix is to expand filenames only if they are relative:
>
> (defun expand-if-relative (a-file)
> (if (file-name-absolute-p a-file)
>     a-file
>   (expand-file-name a-file)
>   )
> )
>
> and then use this function in the mapcar statement inside (defun
> tags-table-including ...), replacing expand-file-name:
>   ....
>   (if (member this-file (mapcar #'expand-if-relative
>                                           (tags-table-files)))
>                 ;; Found it.
>                 (setq found tables)
>                 )
>   ....

Thanks.  How much speedup does this bring in your use cases?  From 10 sec down 
to how long?

> Looking at expand-file-name doc, it does 2 things:
>   * make file name absolute
>   * and canonicalize it (removes the xxxx/.. dir components,
>        the . dir components, the double slashes).

It actually does more, like resolve the ~/ etc.

> We are wondering if the standard C function in emacs could do a fast
> "do nothing" when the file name is absolute and has nothing to
> cannonicalize.

I think testing for "nothing to canonicalize" is as complex as canonicalizing 
the file name.
____

This message and any files transmitted with it are legally privileged and 
intended for the sole use of the individual(s) or entity to whom they are 
addressed. If you are not the intended recipient, please notify the sender by 
reply and delete the message and any attachments from your system. Any 
unauthorised use or disclosure of the content of this message is strictly 
prohibited and may be unlawful.

Nothing in this e-mail message amounts to a contractual or legal commitment on 
the part of EUROCONTROL, unless it is confirmed by appropriately signed hard 
copy.

Any views expressed in this message are those of the sender.