optimize diffing input with long matching prefix, see #15

The benchmark in this commit speeds up from ~360us to ~210us.

Author: utkarshkukreti

benches/benches.rs

@@ -3,7 +3,7 @@ extern crate diff;
 
 use criterion::Criterion;
 
-criterion::criterion_group!(benches, bench_slice);
+criterion::criterion_group!(benches, bench_slice, bench_chars);
 criterion::criterion_main!(benches);
 
 fn bench_slice(c: &mut Criterion) {
@@ -42,3 +42,12 @@ fn bench_slice(c: &mut Criterion) {
         b.iter(|| ::diff::slice(&left, &right));
     });
 }
+
+fn bench_chars(c: &mut Criterion) {
+    c.bench_function("1024 byte string, last 256 different", |b| {
+        let left = "?".repeat(768) + &"_".repeat(256);
+        let right = "?".repeat(768) + &"!".repeat(256);
+        assert_eq!(left.len(), right.len());
+        b.iter(|| ::diff::chars(&left, &right));
+    });
+}

src/lib.rs

@@ -67,7 +67,7 @@ where
     let table: Vec<Vec<u32>> = {
         let mut table = vec![vec![0; right_diff_size + 1]; left_diff_size + 1];
         let left_skip = left.clone().skip(leading_equals).take(left_diff_size);
-        let right_skip = right.clone().skip(leading_equals).take(right_diff_size);
+        let right_skip = advance_by(right.clone(), leading_equals).take(right_diff_size);
 
         for (i, l) in left_skip.clone().enumerate() {
             for (j, r) in right_skip.clone().enumerate() {
@@ -126,3 +126,11 @@ where
 
     total_diff
 }
+
+#[inline]
+fn advance_by<I: Iterator>(mut iter: I, by: usize) -> I {
+    for _ in 0..by {
+        iter.next();
+    }
+    iter
+}