Re: [RFC] Diff (blame) optimization: how to go forward

Johan Corveleyn wrote on Fri, 8 Oct 2010 at 01:44 -0000:
> With suffix-lines-to-keep=50, you'd need to insert a
> block of text that has its last 50 lines identical to the 50 lines
> preceding the insertion point, to mess up the diff result.
>
> - If really necessary, we could say default=50, but it can be
> overridden by the user with another -x option.
>
> So the main question is: is such a change in behavior (unlikely to
> have an impact in most realistic cases) acceptable for this kind of
> performance improvement?

Just to clarify: are you asking whether it's acceptable to have a "not
nice" (but still semantically correct) diff output in case the user adds
a block whose last 50 lines match the 50 lines prior to the first
difference?

And 'not nice' just means "the preceding, rather than trailing, instance
of the repeated block would be grabbed by the /^+/ lines in the diff",
right?

In this case, I think I'm going to answer your long mail with just...

    Yes.

or actually...

    Yes, and let's move on to deeper problems (if any).

(I'm not /particularly/ worried about a "different-than-expected, but
still semantically correct, diff" --- though, yes, it's nice to output
the "expected" diff if that's possible without too much effort and
without breaking other use cases.)

:-),

Daniel
(p.s. I don't have as much time to look into this as I'd like to...)
Received on 2010-10-08 15:09:41 CEST