Fix streams with split unicode sequences.

Author: RubenVerborgh

lib/N3Lexer.js

@@ -423,13 +423,26 @@ N3Lexer.prototype = {
     // Otherwise, the input must be a stream
     else {
       this._input = '';
+      this._pendingBuffer = null;
       if (typeof input.setEncoding === 'function')
         input.setEncoding('utf8');
       // Adds the data chunk to the buffer and parses as far as possible
       input.on('data', function (data) {
-        if (self._input !== null) {
-          self._input += data;
-          self._tokenizeToEnd(callback, false);
+        if (self._input !== null && data.length !== 0) {
+          // Prepend any previous pending writes
+          if (self._pendingBuffer) {
+            data = Buffer.concat([self._pendingBuffer, data]);
+            self._pendingBuffer = null;
+          }
+          // Hold if the buffer ends in an incomplete unicode sequence
+          if (data[data.length - 1] & 0x80) {
+            self._pendingBuffer = data;
+          }
+          // Otherwise, tokenize as far as possible
+          else {
+            self._input += data;
+            self._tokenizeToEnd(callback, false);
+          }
         }
       });
       // Parses until the end

test/N3StreamParser-test.js

@@ -31,7 +31,7 @@ describe('N3StreamParser', function () {
     it('should parse decimals that are split across chunks in the stream',
       shouldParse('<sub> <pred> 11.2 .'.match(/.{1,2}/g), 1));
 
-    it.skip('should parse non-breaking spaces that are split across chunks in the stream correctly', function (done) {
+    it('should parse non-breaking spaces that are split across chunks in the stream correctly', function (done) {
       var buffer = Buffer.from('<sub> <pred> " " .'),
           chunks = [buffer, buffer.slice(0, 15), buffer.slice(15, buffer.length)];
       shouldParse(chunks, 2, function (triples) {