Re: Refining ediff's idea of a difference?

[Eli Zaretskii <eliz@gnu.org>]

> From: Doug Lewan <dougl@shubertticketing.com>
> Date: Fri, 30 Jan 2015 23:11:41 +0000
> 
> Using ediff I find that very often one difference one difference seems inappropriately large and might obscure a small difference.
> 
> In the example at hand I have the two following diffs:
> 
>     <<<< A
>      -clrobj server_info_blk -fid server_info_begin_time  -val 2015-01-29 15:28:47
>     >>>> B
>      -clrobj server_info_blk -fid server_info_begin_time  -val 2015-01-29 15:27:28
>     -fid server_info_reply_que  -val 
>     -fid server_info_tran_msg_type  -val 
>     -fid server_info_tran_id  -val 
>     -fid server_info_seq_msg_type  -val 
>     -fid server_info_seq_id  -val 
>     ...
> 
> The lines following diff A are these
>     <<<< A
>     -fid server_info_reply_que  -val 
>     -fid server_info_tran_msg_type  -val 
>     -fid server_info_tran_id  -val 
>     -fid server_info_seq_msg_type  -val 
>     -fid server_info_seq_id  -val 
> which look, to my eye, identical to those flagged as different in diff B.
> 
> I'd rather those lines are recognized as identical, so the diff is minimal. Is there a conventional way to do that?

Ediff just obeys here what the Diff utility outputs.  So try playing
with Diff switches, like -d, -b, -B, -w, --horizon-lines.

> I haven't found anything obvious in ediff customization. Browsing the code didn't reveal much more than the fact that ediff seems to use diff with no options. It clearly does more with that information.

The relevant defcustom is ediff-diff-options.