From 2be8e596e26ff974d50832f15a5df42d1348d3d5 Mon Sep 17 00:00:00 2001
From: Yuya Nishihara <yuya@tcha.org>
Date: Fri, 16 Aug 2024 16:02:36 +0900
Subject: [PATCH] diff: extract Diff::by_word() function

I'm going to split color-words diffs to by_line() and by_word() stages.

Perhaps, Diff::default_refinement() can be removed once all non-test callers
are migrated.
---
 cli/src/diff_util.rs |  8 ++------
 lib/src/diff.rs      | 12 ++++++++++++
 2 files changed, 14 insertions(+), 6 deletions(-)

diff --git a/cli/src/diff_util.rs b/cli/src/diff_util.rs
index 63a79210b..9b37f029d 100644
--- a/cli/src/diff_util.rs
+++ b/cli/src/diff_util.rs
@@ -25,7 +25,7 @@ use jj_lib::commit::Commit;
 use jj_lib::conflicts::{
     materialized_diff_stream, MaterializedTreeDiffEntry, MaterializedTreeValue,
 };
-use jj_lib::diff::{self, Diff, DiffHunk};
+use jj_lib::diff::{Diff, DiffHunk};
 use jj_lib::files::{DiffLine, DiffLineHunkSide, DiffLineIterator};
 use jj_lib::matchers::Matcher;
 use jj_lib::merge::MergedTreeValue;
@@ -1025,11 +1025,7 @@ fn inline_diff_hunks<'content>(
     let mut left_tokens: DiffTokenVec<'content> = vec![];
     let mut right_tokens: DiffTokenVec<'content> = vec![];
 
-    // Like Diff::default_refinement(), but doesn't try to match up contents by
-    // lines. We know left/right_contents have no matching lines.
-    let mut diff = Diff::for_tokenizer([left_content, right_content], diff::find_word_ranges);
-    diff.refine_changed_regions(diff::find_nonword_ranges);
-    for hunk in diff.hunks() {
+    for hunk in Diff::by_word([left_content, right_content]).hunks() {
         match hunk {
             DiffHunk::Matching(content) => {
                 for token in content.split_inclusive(|b| *b == b'\n') {
diff --git a/lib/src/diff.rs b/lib/src/diff.rs
index 5daa918e4..a56b3af78 100644
--- a/lib/src/diff.rs
+++ b/lib/src/diff.rs
@@ -493,6 +493,18 @@ impl<'input> Diff<'input> {
         Diff::for_tokenizer(inputs, find_line_ranges)
     }
 
+    /// Compares `inputs` word by word.
+    ///
+    /// The `inputs` is usually a changed hunk (e.g. a `DiffHunk::Different`)
+    /// that was the output from a line-by-line diff.
+    pub fn by_word<T: AsRef<[u8]> + ?Sized + 'input>(
+        inputs: impl IntoIterator<Item = &'input T>,
+    ) -> Self {
+        let mut diff = Diff::for_tokenizer(inputs, find_word_ranges);
+        diff.refine_changed_regions(find_nonword_ranges);
+        diff
+    }
+
     // TODO: At least when merging, it's wasteful to refine the diff if e.g. if 2
     // out of 3 inputs match in the differing regions. Perhaps the refine()
     // method should be on the hunk instead (probably returning a new Diff)?