From 76cac4c05a7be61a94a709b8b850118ad0bfa684 Mon Sep 17 00:00:00 2001
From: Yusuke Endoh <mame@ruby-lang.org>
Date: Wed, 21 Oct 2020 13:29:19 +0900
Subject: [ruby/reline] Improve the performance of `get_mbchar_width`

It is about three times faster to use one big regexp instead of
sequential matching.

https://github.com/ruby/reline/commit/e36f6c0707
---
 lib/reline/unicode.rb                  | 40 +++++++++++++++++++++-------------
 lib/reline/unicode/east_asian_width.rb | 26 +++++++++++-----------
 2 files changed, 38 insertions(+), 28 deletions(-)

(limited to 'lib')

diff --git a/lib/reline/unicode.rb b/lib/reline/unicode.rb
index cd8c27e85b..df2f6719a4 100644
--- a/lib/reline/unicode.rb
+++ b/lib/reline/unicode.rb
@@ -72,20 +72,32 @@ class Reline::Unicode
     }.join
   end
 
+  require 'reline/unicode/east_asian_width'
+
+  MBCharWidthRE = /
+    (?<width_2_1>
+      [#{ EscapedChars.map {|c| "\\x%02x" % c.ord }.join }] (?# ^ + char, such as ^M, ^H, ^[, ...)
+    )
+  | (?<width_3>^\u{2E3B}) (?# THREE-EM DASH)
+  | (?<width_0>^\p{M})
+  | (?<width_2_2>
+      #{ EastAsianWidth::TYPE_F }
+    | #{ EastAsianWidth::TYPE_W }
+    )
+  | (?<width_1>
+      #{ EastAsianWidth::TYPE_H }
+    | #{ EastAsianWidth::TYPE_NA }
+    | #{ EastAsianWidth::TYPE_N }
+    )
+  /x
+
   def self.get_mbchar_width(mbchar)
-    case mbchar.encode(Encoding::UTF_8)
-    when *EscapedChars # ^ + char, such as ^M, ^H, ^[, ...
-      2
-    when /^\u{2E3B}/ # THREE-EM DASH
-      3
-    when /^\p{M}/
-      0
-    when EastAsianWidth::TYPE_A
-      Reline.ambiguous_width
-    when EastAsianWidth::TYPE_F, EastAsianWidth::TYPE_W
-      2
-    when EastAsianWidth::TYPE_H, EastAsianWidth::TYPE_NA, EastAsianWidth::TYPE_N
-      1
+    m = mbchar.encode(Encoding::UTF_8).match(MBCharWidthRE)
+    case
+    when m[:width_2_1], m[:width_2_2] then 2
+    when m[:width_3] then 3
+    when m[:width_0] then 0
+    when m[:width_1] then 1
     else
       nil
     end
@@ -591,5 +603,3 @@ class Reline::Unicode
     [byte_size, width]
   end
 end
-
-require 'reline/unicode/east_asian_width'
diff --git a/lib/reline/unicode/east_asian_width.rb b/lib/reline/unicode/east_asian_width.rb
index 7483c78936..89bc9d9435 100644
--- a/lib/reline/unicode/east_asian_width.rb
+++ b/lib/reline/unicode/east_asian_width.rb
@@ -1,16 +1,16 @@
 class Reline::Unicode::EastAsianWidth
   # This is based on EastAsianWidth.txt
-  # http://www.unicode.org/Public/13.0.0/ucd/EastAsianWidth.txt
+  # EastAsianWidth.txt
 
   # Fullwidth
-  TYPE_F = /^([#{ %W(
+  TYPE_F = /^[#{ %W(
     \u{3000}
     \u{FF01}-\u{FF60}
     \u{FFE0}-\u{FFE6}
-  ).join }])/
+  ).join }]/
 
   # Halfwidth
-  TYPE_H = /^([#{ %W(
+  TYPE_H = /^[#{ %W(
     \u{20A9}
     \u{FF61}-\u{FFBE}
     \u{FFC2}-\u{FFC7}
@@ -18,10 +18,10 @@ class Reline::Unicode::EastAsianWidth
     \u{FFD2}-\u{FFD7}
     \u{FFDA}-\u{FFDC}
     \u{FFE8}-\u{FFEE}
-  ).join }])/
+  ).join }]/
 
   # Wide
-  TYPE_W = /^([#{ %W(
+  TYPE_W = /^[#{ %W(
     \u{1100}-\u{115F}
     \u{231A}-\u{231B}
     \u{2329}-\u{232A}
@@ -136,10 +136,10 @@ class Reline::Unicode::EastAsianWidth
     \u{1FAD0}-\u{1FAD6}
     \u{20000}-\u{2FFFD}
     \u{30000}-\u{3FFFD}
-  ).join }])/
+  ).join }]/
 
   # Narrow
-  TYPE_NA = /^([#{ %W(
+  TYPE_NA = /^[#{ %W(
     \u{0020}-\u{007E}
     \u{00A2}-\u{00A3}
     \u{00A5}-\u{00A6}
@@ -147,10 +147,10 @@ class Reline::Unicode::EastAsianWidth
     \u{00AF}
     \u{27E6}-\u{27ED}
     \u{2985}-\u{2986}
-  ).join }])/
+  ).join }]/
 
   # Ambiguous
-  TYPE_A = /^([#{ %W(
+  TYPE_A = /^[#{ %W(
     \u{00A1}
     \u{00A4}
     \u{00A7}-\u{00A8}
@@ -330,10 +330,10 @@ class Reline::Unicode::EastAsianWidth
     \u{E0100}-\u{E01EF}
     \u{F0000}-\u{FFFFD}
     \u{100000}-\u{10FFFD}
-  ).join }])/
+  ).join }]/
 
   # Neutral
-  TYPE_N = /^([#{ %W(
+  TYPE_N = /^[#{ %W(
     \u{0000}-\u{001F}
     \u{007F}-\u{00A0}
     \u{00A9}
@@ -1160,5 +1160,5 @@ class Reline::Unicode::EastAsianWidth
     \u{1FBF0}-\u{1FBF9}
     \u{E0001}
     \u{E0020}-\u{E007F}
-  ).join }])/
+  ).join }]/
 end
-- 
cgit v1.2.3