[darcs-users] diff algorithm: matching similar lines

[Ben Franksen]

Am 21.08.2016 um 22:23 schrieb Henning Thielemann:
> I did not want to propose new patch types, just improved diff (for now).
> 
> Say I have
> ------
> a
> 
> b
> 
> c
> ------
> 
> and alter it to
> 
> ------
>  b
> 
>  c
> ------
> 
> I guess that the record command would replace 'a' by ' b' and 'b' ' c'
> and drop 'c' (the current patience diff might replace all lines at
> once). But it would be more natural to drop 'a' and replace 'b' by ' b'
> and 'c' by ' c'.

A while ago I played with a completely new patch formalism (a mixture of
the approaches of Pijul and Focal) and as a related side project with a
new diff algorithm that is not line based and can detect moves of
sections of a file to a different place. It could certainly be used to
detect things like a change of indentation of an otherwise unchanged
line of text.

It did not occur to me until now that this might perhaps be useful to
Darcs. The algorithm is not as efficient as the usual candidates but it
should be just usable when there aren't too many differences. Also, for
Darcs the result needs to be converted to something line-based.

Anyway, if I find the time, I can perhaps make an experimental patch. Oh
and I need a name for the algorithm, so I know how to name the command
line option...

Cheers
Ben
-- 
"Make it so they have to reboot after every typo." ― Scott Adams