Re: [PATCH] Permit user to select keys when using sdiff merge

[Chuck Swiger]

On Sep 22, 2004, at 2:11 AM, Paul Eggert wrote:
> Thanks for explaining it.

Sure, sorry for the delay in responding.

> How about if we just hardcode the "1" and "2" keys as alternatives to 
> the existing "l" and "r" keys?  That would be simpler than adding a 
> new option.

Hardcoding "1" and "2" would suit my personal preferences just fine.  
It would involve removing the getenv() call from the proposed patch, 
which is simple to do.

However, the person who made the initial feature request wanted 
different keys.  I think a solution that lets the user select their 
preferred keys is better than hardcoding any individual preferences, 
but I suppose that's up to Tom Lord or whoever is maintaining diffutils 
and is willing to commit such a change.

        -----

Perhaps I should backtrack a little, and add some thoughts.  Good 
patches are short, fix problems or add functionality cleanly, and 
include documentation.  Changing which keystrokes are used for sdiff at 
compile time is trivial: it only takes about 10 lines.  Letting sdiff 
choose which keys to use at runtime is harder since it requires 
changing a switch statement to a set of if/thens, since one cannot test 
against a variable via switch.

Once I actually got into the code, I discovered that the code for 
handling user commands during a merge is a little flakey and didn't 
seem to be documented anywhere-- it would complain rather than accept 
words like "quit" or "help".  I happen to think that accepting 'q' but 
not "quit" is silly, so the change I proposed handles such cases more 
reasonably than the original code.  [IMHO]

It also makes the proposed patch longer than I would like.

--
-Chuck