Always pipe SAXParser to the /dev/null stream to prevent hitting Read…

…ableStream buffer limits if we don't have consumers (fixes #97)

docs/07_version_history.md

@@ -1,5 +1,8 @@
 # Version history
 
+## 2.1.2
+ * Fixed: SAX parser silently exits on big files (GH [#97](https://github.com/inikulin/parse5/issues/97))
+
 ## 2.1.1
  * Fixed: location info not attached for empty attributes (GH [#96](https://github.com/inikulin/parse5/issues/96))
  (by [@yyx990803](https://github.com/yyx990803)).

lib/sax/dev_null_stream.js

@@ -0,0 +1,14 @@
+'use strict';
+
+var WritableStream = require('stream').Writable,
+    util = require('util');
+
+var DevNullStream = module.exports = function () {
+    WritableStream.call(this);
+};
+
+util.inherits(DevNullStream, WritableStream);
+
+DevNullStream.prototype._write = function (chunk, encoding, cb) {
+    cb();
+};

lib/sax/index.js

@@ -1,6 +1,7 @@
 'use strict';
 
 var TransformStream = require('stream').Transform,
+    DevNullStream = require('./dev_null_stream'),
     inherits = require('util').inherits,
     Tokenizer = require('../tokenizer'),
     ParserFeedbackSimulator = require('./parser_feedback_simulator'),
@@ -59,6 +60,11 @@ var SAXParser = module.exports = function (options) {
 
     this.lastChunkWritten = false;
     this.stopped = false;
+
+    // NOTE: always pipe stream to the /dev/null stream to avoid
+    // `highWaterMark` hit even if we don't have consumers.
+    // (see: https://github.com/inikulin/parse5/issues/97#issuecomment-171940774)
+    this.pipe(new DevNullStream());
 };
 
 inherits(SAXParser, TransformStream);

package.json

@@ -1,7 +1,7 @@
 {
   "name": "parse5",
   "description": "WHATWG HTML5 specification-compliant, fast and ready for production HTML parsing/serialization toolset for Node.js",
-  "version": "2.1.1",
+  "version": "2.1.2",
   "author": "Ivan Nikulin <[email protected]> (https://github.com/inikulin)",
   "contributors": [
     "Alan Clarke (https://github.com/alanclarke)",

test/fixtures/sax_test.js

@@ -128,3 +128,14 @@ exports['SAX - Piping and .stop()'] = function (done) {
         done();
     });
 };
+
+exports['Regression-SAX-SAX parser silently exits on big files (GH-97)'] = function (done) {
+    var parser = new SAXParser();
+
+    fs
+        .createReadStream(path.join(__dirname, '../data/huge-page/huge-page.html'))
+        .pipe(parser);
+
+    //NOTE: This is a smoke test - in case of regression it will fail with timeout.
+    parser.once('finish', done);
+};