string_decoder: rewrite implementation

This commit provides a rewrite of StringDecoder that both improves
performance (for non-single-byte encodings) and understandability.

Additionally, StringDecoder instantiation performance has increased
considerably due to inlinability and more efficient encoding name
checking.

PR-URL: https://github.com/nodejs/node/pull/6777
Reviewed-By: James M Snell <jasnell@gmail.com>
Reviewed-By: Anna Henningsen <anna@addaleax.net>
Reviewed-By: Ben Noordhuis <info@bnoordhuis.nl>

2 files changed, 72 insertions, 22 deletions

diff --git a/benchmark/string_decoder/string-decoder-create.js b/benchmark/string_decoder/string-decoder-create.js
new file mode 100644
index 0000000000..17c0f6750d
--- /dev/null
+++ b/benchmark/string_decoder/string-decoder-create.js
@@ -0,0 +1,22 @@
+'use strict';
+const common = require('../common.js');
+const StringDecoder = require('string_decoder').StringDecoder;
+
+const bench = common.createBenchmark(main, {
+  encoding: [
+    'ascii', 'utf8', 'utf-8', 'base64', 'ucs2', 'UTF-8', 'AscII', 'UTF-16LE'
+  ],
+  n: [25e6]
+});
+
+function main(conf) {
+  const encoding = conf.encoding;
+  const n = conf.n | 0;
+
+  bench.start();
+  for (var i = 0; i < n; ++i) {
+    const sd = new StringDecoder(encoding);
+    !!sd.encoding;
+  }
+  bench.end(n);
+}
diff --git a/benchmark/string_decoder/string-decoder.js b/benchmark/string_decoder/string-decoder.js
index 705ef84103..7a86fb6ef5 100644
--- a/benchmark/string_decoder/string-decoder.js
+++ b/benchmark/string_decoder/string-decoder.js
@@ -1,51 +1,79 @@
 'use strict';
-var common = require('../common.js');
-var StringDecoder = require('string_decoder').StringDecoder;
+const common = require('../common.js');
+const StringDecoder = require('string_decoder').StringDecoder;
 
-var bench = common.createBenchmark(main, {
-  encoding: ['ascii', 'utf8', 'base64-utf8', 'base64-ascii'],
-  inlen: [32, 128, 1024],
+const bench = common.createBenchmark(main, {
+  encoding: ['ascii', 'utf8', 'base64-utf8', 'base64-ascii', 'utf16le'],
+  inlen: [32, 128, 1024, 4096],
   chunk: [16, 64, 256, 1024],
-  n: [25e4]
+  n: [25e5]
 });
 
-var UTF_ALPHA = 'Blåbærsyltetøy';
-var ASC_ALPHA = 'Blueberry jam';
+const UTF8_ALPHA = 'Blåbærsyltetøy';
+const ASC_ALPHA = 'Blueberry jam';
+const UTF16_BUF = Buffer.from('Blåbærsyltetøy', 'utf16le');
 
 function main(conf) {
-  var encoding = conf.encoding;
-  var inLen = conf.inlen | 0;
-  var chunkLen = conf.chunk | 0;
-  var n = conf.n | 0;
+  const encoding = conf.encoding;
+  const inLen = conf.inlen | 0;
+  const chunkLen = conf.chunk | 0;
+  const n = conf.n | 0;
 
   var alpha;
-  var chunks = [];
+  var buf;
+  const chunks = [];
   var str = '';
-  var isBase64 = (encoding === 'base64-ascii' || encoding === 'base64-utf8');
+  const isBase64 = (encoding === 'base64-ascii' || encoding === 'base64-utf8');
   var i;
 
   if (encoding === 'ascii' || encoding === 'base64-ascii')
     alpha = ASC_ALPHA;
   else if (encoding === 'utf8' || encoding === 'base64-utf8')
-    alpha = UTF_ALPHA;
-  else
+    alpha = UTF8_ALPHA;
+  else if (encoding === 'utf16le') {
+    buf = UTF16_BUF;
+    str = Buffer.alloc(0);
+  } else
     throw new Error('Bad encoding');
 
-  var sd = new StringDecoder(isBase64 ? 'base64' : encoding);
+  const sd = new StringDecoder(isBase64 ? 'base64' : encoding);
 
   for (i = 0; i < inLen; ++i) {
     if (i > 0 && (i % chunkLen) === 0 && !isBase64) {
-      chunks.push(Buffer.from(str, encoding));
-      str = '';
+      if (alpha) {
+        chunks.push(Buffer.from(str, encoding));
+        str = '';
+      } else {
+        chunks.push(str);
+        str = Buffer.alloc(0);
+      }
+    }
+    if (alpha)
+      str += alpha[i % alpha.length];
+    else {
+      var start = i;
+      var end = i + 2;
+      if (i % 2 !== 0) {
+        ++start;
+        ++end;
+      }
+      str = Buffer.concat([
+        str,
+        buf.slice(start % buf.length, end % buf.length)
+      ]);
     }
-    str += alpha[i % alpha.length];
   }
-  if (str.length > 0 && !isBase64)
+
+  if (!alpha) {
+    if (str.length > 0)
+      chunks.push(str);
+  } else if (str.length > 0 && !isBase64)
     chunks.push(Buffer.from(str, encoding));
+
   if (isBase64) {
     str = Buffer.from(str, 'utf8').toString('base64');
     while (str.length > 0) {
-      var len = Math.min(chunkLen, str.length);
+      const len = Math.min(chunkLen, str.length);
       chunks.push(Buffer.from(str.substring(0, len), 'utf8'));
       str = str.substring(len);
     }