Quoting Ian Lynagh <[EMAIL PROTECTED]>:

> I'm not sure what you mean here. In my mind, the new hunk
>
>     newhunk 8 -3 +6
>     old
>     foo
>     new
>     wibble
>
> (or equivalent) means "Go to the 8th byte. The next 3 bytes will be
> "foo". Remove them. Insert the 6 bytes "wibble".
>
> There are various details you can fiddle with, like whether we note how
> many lines are added/removed, what the starting line number is etc
> (which comes down to "is the extra complexity worth it for how important
> it is for users") but I'd really like a format that is trivial to apply
> and coalesce.

I'm probably the last person that should chime in since I've not followed the
complete discussion about the new hunk format (and if I'm preaching to the
choir then please just ignore this)... I hope whatever you decide on allows for
fseek() (or similar) to jump to the relevant parts of the patch files.  It
sounds like it may be possible given the above format (although the offset
computation might be a bit hairy, and easily broken if the formatting is
tweaked).  I just want to stress that I think this is important for allowing
large patches to be handled efficiently.  I'm just envisioning something where
darcs can skip from one patch to the next without reading anything inbetween
from the file.  I guess the next step would be to make a tool (as part of darcs
or external) that allows upgrading from the old patch format?

Thanks,
Jason

_______________________________________________
darcs-devel mailing list
[email protected]
http://www.abridgegame.org/cgi-bin/mailman/listinfo/darcs-devel

Reply via email to