From 0a5189054e4404fe13de7b0e10be5858648e0924 Mon Sep 17 00:00:00 2001
From: NAITOH Jun <naitoh@gmail.com>
Date: Sat, 20 Jan 2024 07:03:11 +0900
Subject: [PATCH] Reduce calls to `Source#buffer`(`StringScanner#rest`)

[Why]
`Source#buffer` calling `StringScanner#rest`.
`StringScanner#rest` is slow.
Reduce calls to `Source#buffer`.

## Benchmark

```
RUBYLIB= BUNDLER_ORIG_RUBYLIB= /Users/naitoh/.rbenv/versions/3.3.0/bin/ruby -v -S benchmark-driver /Users/naitoh/ghq/github.com/naitoh/rexml/benchmark/parse.yaml
ruby 3.3.0 (2023-12-25 revision 5124f9ac75) [arm64-darwin22]
Calculating -------------------------------------
                         before       after  before(YJIT)  after(YJIT)
                 dom     10.673      10.776        16.184       16.334 i/s -     100.000 times in 9.369090s 9.280170s 6.178928s 6.122305s
                 sax     28.385      28.930        42.270       42.533 i/s -     100.000 times in 3.522933s 3.456645s 2.365771s 2.351098s
                pull     32.568      34.040        50.351       52.088 i/s -     100.000 times in 3.070514s 2.937748s 1.986055s 1.919817s
              stream     30.688      31.483        42.226       44.754 i/s -     100.000 times in 3.258605s 3.176358s 2.368228s 2.234450s

Comparison:
                              dom
         after(YJIT):        16.3 i/s
        before(YJIT):        16.2 i/s - 1.01x  slower
               after:        10.8 i/s - 1.52x  slower
              before:        10.7 i/s - 1.53x  slower

                              sax
         after(YJIT):        42.5 i/s
        before(YJIT):        42.3 i/s - 1.01x  slower
               after:        28.9 i/s - 1.47x  slower
              before:        28.4 i/s - 1.50x  slower

                             pull
         after(YJIT):        52.1 i/s
        before(YJIT):        50.4 i/s - 1.03x  slower
               after:        34.0 i/s - 1.53x  slower
              before:        32.6 i/s - 1.60x  slower

                           stream
         after(YJIT):        44.8 i/s
        before(YJIT):        42.2 i/s - 1.06x  slower
               after:        31.5 i/s - 1.42x  slower
              before:        30.7 i/s - 1.46x  slower

```

- YJIT=ON : 1.01x - 1.06x faster
- YJIT=OFF : 1.01x - 1.04x faster
---
 lib/rexml/parsers/baseparser.rb | 9 +++++----
 1 file changed, 5 insertions(+), 4 deletions(-)

diff --git a/lib/rexml/parsers/baseparser.rb b/lib/rexml/parsers/baseparser.rb
index 65bad260..be9c5677 100644
--- a/lib/rexml/parsers/baseparser.rb
+++ b/lib/rexml/parsers/baseparser.rb
@@ -349,8 +349,9 @@ def pull_event
         end
         begin
           @source.read if @source.buffer.size<2
-          if @source.buffer[0] == ?<
-            if @source.buffer[1] == ?/
+          pattern = @source.buffer[0,2]
+          if pattern[0] == ?<
+            if pattern[1] == ?/
               @nsstack.shift
               last_tag = @tags.pop
               md = @source.match( CLOSE_MATCH, true )
@@ -364,7 +365,7 @@ def pull_event
                 raise REXML::ParseException.new(message, @source)
               end
               return [ :end_element, last_tag ]
-            elsif @source.buffer[1] == ?!
+            elsif pattern[1] == ?!
               md = @source.match(/\A(\s*[^>]*>)/um)
               #STDERR.puts "SOURCE BUFFER = #{source.buffer}, #{source.buffer.size}"
               raise REXML::ParseException.new("Malformed node", @source) unless md
@@ -383,7 +384,7 @@ def pull_event
               end
               raise REXML::ParseException.new( "Declarations can only occur "+
                 "in the doctype declaration.", @source)
-            elsif @source.buffer[1] == ??
+            elsif pattern[1] == ??
               return process_instruction
             else
               # Get the next tag