Diff + patch - Sum instead of replace

Question

Say I have a file A, which contains this:

a = 5

And a file B, like this:

b = 5

Now, it's obvious that diff will produce something like this (the patch):

1c1

< a = 5

---

> b = 5

Patching file A will obviously replace its contents with those of file B, resulting in file A containing

b = 5

What I want to do, however, is different. I want the contents of files A and B to merge, so that after patching file A it will contain

a = 5

b = 5

My case if course way more complex than my example but I think I've figured it out except for the summative diff/patch at the end.

tl;dr: I want the diff/patch to sum rather than replace the differences in the files. How can I do that?

This is not what diff and patch are for. You will need to use an actual tool that supports merges. — Ignacio Vazquez-Abrams, Jun 19 '11 at 21:41

binfalse · Answer 1 · 2011-06-19T21:58:15.267

2

You don't want to patch some files, you are trying to merge them. There are lot's of tools out there that will help you, e.g. take a look at kdiff!

Edit: Some more tools:

meld
gPyFm
diff3
tkdiff
P4Merge

At least vim -d will do the job ;-)

edited Jun 19 '11 at 21:58

answered Jun 19 '11 at 21:45

binfalse

508
4
10

The problem with kdiff is that it requires the whole kde base, which is quite heavy considering that I do most of my work on CLI. I installed meld but when I tell it to merge, it does the same as diff+patch. – Chinoise Jun 19 '11 at 21:50
of course each tool has its dependencies.. since i don't know your env i can't recommend anything based on this. but i know that `kdiff3` does a really good job. nevertheless i added some more programs, maybe you'll find some more... – binfalse Jun 19 '11 at 22:01

score 0 · Answer 2 · answered Jun 19 '11 at 22:00

You tell very very little about what problem you are trying to solve (or you talk too much about how the approach you took did not work). What's more interesting is what you want done.

Perhaps you could be done with

sort -u A B

Because it will give the output you require with the inputs you provide.

Oben Sonne · Answer 3 · 2011-06-20T12:24:52.773

The diff command from the GNU diffutils packages provides an option to merge two different files by inserting #ifdef preprocessor instructions. For instance consider 2 files f1 and f2 (displayed side-by-side):

line 1     |   line 1
line 2     |   line 2
line 3a    |   line 3a
line 4     |   line 4
line 5a    |   line 5a

Calling diff -D MARKER f1 f2 will produce this merged output:

line 1
line 2
#ifndef MARKER
line 3a
#else /* MARKER */
line 3b
#endif /* MARKER */
line 4
#ifndef MARKER
line 5a
#else /* MARKER */
line 5c
#endif /* MARKER */

You can strip the preprocessor instructions to get a clean merged file, e.g. by using grep:

diff -D MARKER f1 f2 | grep -v "MARKER"

Note that the token MARKER must not exist in the original input files, so you should better use something more special, e.g. a random token like c11ced4751ef4b3aa60e512bdaa184210441b4ed0.

Diff + patch - Sum instead of replace

3 Answers3

Linked