From 2cb3e403e802250308ef05e37e33105ad68eaef6 Mon Sep 17 00:00:00 2001
From: Dan Davison <dandavison7@gmail.com>
Date: Mon, 18 May 2020 09:25:46 -0400
Subject: Performance: create regexp as a top-level static

This seems to be a huge performance improvement: `make benchmark`
changes from 1.7s to 0.9s.
---
 src/edits.rs | 8 ++++++--
 1 file changed, 6 insertions(+), 2 deletions(-)

(limited to 'src/edits.rs')

diff --git a/src/edits.rs b/src/edits.rs
index 1788afd5..2e16d2be 100644
--- a/src/edits.rs
+++ b/src/edits.rs
@@ -1,5 +1,6 @@
 use regex::Regex;
 
+use lazy_static::lazy_static;
 use unicode_width::UnicodeWidthStr;
 
 use crate::align;
@@ -75,13 +76,16 @@ where
     (annotated_minus_lines, annotated_plus_lines)
 }
 
+lazy_static! {
+    static ref TOKENIZATION_REGEXP: Regex = Regex::new(r#"[\t ,;.:()\[\]<>/'"-]+"#).unwrap();
+}
+
 /// Split line into tokens for alignment. The alignment algorithm aligns sequences of substrings;
 /// not individual characters.
 fn tokenize(line: &str) -> Vec<&str> {
-    let separators = Regex::new(r#"[\t ,;.:()\[\]<>/'"-]+"#).unwrap();
     let mut tokens = Vec::new();
     let mut offset = 0;
-    for m in separators.find_iter(line) {
+    for m in TOKENIZATION_REGEXP.find_iter(line) {
         tokens.push(&line[offset..m.start()]);
         // Align separating text as multiple single-character tokens.
         for i in m.start()..m.end() {
-- 
cgit v1.2.3