Re: ediff feature request: diffing line by line

[Alex Schroeder]

address@hidden (Michael Kifer) writes:

>     Karl> This basically means a diff is required that compares line
>     Karl> by line and that allows to say "next word" from version A or
>     Karl> "rest of line" from version B is wanted.
>
> Ediff is designed to parse the output of diff and then present it in
> different ways. What you are proposing is a kind of "incremental diff", if I
> understand it correctly. I think it would require a major generalization of
> the existing code to do that.

I'm not sure, as I understand it, all Karl is saying that we would
like to see more commands to act upon the differences -- copy the
replace the first word of chunk A with the first word of chunk B and
show me a new diff output.  The diff output -- the underlying calls to
diff and the processing of the output, and the ediff display need not
be changed.  I assume that is biggest part of the code... (without
looking at it).

Some of what Carlo Traverso wishes for could be implemented by other
people by writing some sort of "filters" -- functions that
get run when the files are loaded and not yet saved as temporary
files.  The diffing, parsing, mangling, displaying, etc. remains as it
is.  But authors of a specialized ocr-ediff could then write a
collection of functions (plus relevant customization variables) which
transform (normalize, canonicalize, whatever) the text of files,
before ediff gets started.  This might be nice to have in ediff itself
(some sort of hook) but is not really required.  Perhaps Carlo can
tell us wether that would be good enough?

Alex.
-- 
http://www.emacswiki.org/