Re: GNU diff and git diff - difference on myers algorithm?

On Thu, Jul 16, 2015 at 9:22 PM, Jacob Keller [off-list ref] wrote:

On Thu, Jul 16, 2015 at 12:07 PM, Luis R. Rodriguez
[off-list ref] wrote:

quoted

On Fri, Jun 12, 2015 at 11:52 AM, Luis R. Rodriguez
[off-list ref] wrote:

quoted

OK wells I'm curious about more research / effort when trying to
evaluate a diff with two seprate but adjoining preprocessor directives
and if anyone has implemented an optimizaiton option to let the diff
generator join them.

For example, to let it infer that:

--- a/test.c
+++ b/test.c

@@ -10,8 +10,6 @@ int main(int argc, char *argv[])

 #ifdef FOO
        a = 4;
-#endif /* FOO */
-#ifdef FOO
        a = 5;
 #endif /* FOO */

is possible.

Anyone familiar if any tool exists today that would optimize this? Is
anyone working on it? Would git be a good place for such a thing? I'd
consider it as an option to optimize a diff. This for example is
extremely useful for us working with Coccinelle where we have a tool
writing code for us, while such an optimization might be useful to
Coccinelle it would seem like a rather generic feature, its just not
clear to me where to give such a tool a proper home.

 Luis

I do not understand exactly what would be optimized in this case?

In any regards, that's not a diff transformation, that is a code
transformation, and I would suggest starting with Coccinelle and
seeing if you can get that to do what you want.
http://coccinelle.lip6.fr/

Regards,
Jake

I misread your comment. Coccinelle is a tool that could probably be
coerced into doing this, but this is not a diff optimization unless I
am completely understanding it. This is a code transformation concept,
and doesn't have much to do with finding differences or code changes.
My above comment is correct, but I don't think this belongs in diff
parsing, but rather as part of something like Coccinelle

Regards,
Jake