git diff numbers
diff --git a/numbers b/numbers
index 5f5fbe7..d184fef 100644
--- a/numbers
+++ b/numbers
@@ -1,3 +1,3 @@
-1
+4
+3
2
-3
Nu
The diffing tools are usually implemented in terms of the Myers' diff algorithm. There isn't much that you can do to control the behaviour of GNU/git/diff. (There are a couple of switches that you can pass to diff to affect the behaviour, but in your case they are irrelevant.)
You could simply post-process the output and remove the duplicate lines, for example you can pipe your diff through the following awk script that will remove (duplicate) -/+ reordering.
git diff | awk '{ seen[substr($0,2)]++; l[i++] = $0; } END { for (j = 0; j < i; ++j) if (seen[substr(l[j],2)] < 2) print l[j] }'
For your example, the output would be,
diff --git a/numbers b/numbers
index 5f5fbe7..d184fef 100644
--- a/numbers
+++ b/numbers
@@ -1,3 +1,3 @@
-1
+4
2