SPARK-1421. Make MLlib work on Python 2.6

The reason it wasn't working was passing a bytearray to stream.write(), which is not supported in Python 2.6 but is in 2.7. (This array came from NumPy when we converted data to send it over to Java). Now we just convert those bytearrays to strings of bytes, which preserves nonprintable characters as well.

Author: Matei Zaharia <matei@databricks.com>

Closes apache#335 from mateiz/mllib-python-2.6 and squashes the following commits:

f26c59f [Matei Zaharia] Update docs to no longer say we need Python 2.7
a84d6af [Matei Zaharia] SPARK-1421. Make MLlib work on Python 2.6

docs/mllib-guide.md

@@ -38,6 +38,5 @@ depends on native Fortran routines. You may need to install the
 if it is not already present on your nodes. MLlib will throw a linking error if it cannot 
 detect these libraries automatically.
 
-To use MLlib in Python, you will need [NumPy](http://www.numpy.org) version 1.7 or newer
-and Python 2.7.
+To use MLlib in Python, you will need [NumPy](http://www.numpy.org) version 1.7 or newer.
 

docs/python-programming-guide.md

@@ -152,7 +152,7 @@ Many of the methods also contain [doctests](http://docs.python.org/2/library/doc
 # Libraries
 
 [MLlib](mllib-guide.html) is also available in PySpark. To use it, you'll need
-[NumPy](http://www.numpy.org) version 1.7 or newer, and Python 2.7. The [MLlib guide](mllib-guide.html) contains
+[NumPy](http://www.numpy.org) version 1.7 or newer. The [MLlib guide](mllib-guide.html) contains
 some example applications.
 
 # Where to Go from Here

python/pyspark/mllib/__init__.py

@@ -19,11 +19,7 @@
 Python bindings for MLlib.
 """
 
-# MLlib currently needs Python 2.7+ and NumPy 1.7+, so complain if lower
-
-import sys
-if sys.version_info[0:2] < (2, 7):
-    raise Exception("MLlib requires Python 2.7+")
+# MLlib currently needs and NumPy 1.7+, so complain if lower
 
 import numpy
 if numpy.version.version < '1.7':

python/pyspark/serializers.py

@@ -64,6 +64,7 @@
 from itertools import chain, izip, product
 import marshal
 import struct
+import sys
 from pyspark import cloudpickle
 
 
@@ -113,6 +114,11 @@ class FramedSerializer(Serializer):
     where C{length} is a 32-bit integer and data is C{length} bytes.
     """
 
+    def __init__(self):
+        # On Python 2.6, we can't write bytearrays to streams, so we need to convert them
+        # to strings first. Check if the version number is that old.
+        self._only_write_strings = sys.version_info[0:2] <= (2, 6)
+
     def dump_stream(self, iterator, stream):
         for obj in iterator:
             self._write_with_length(obj, stream)
@@ -127,7 +133,10 @@ def load_stream(self, stream):
     def _write_with_length(self, obj, stream):
         serialized = self.dumps(obj)
         write_int(len(serialized), stream)
-        stream.write(serialized)
+        if self._only_write_strings:
+            stream.write(str(serialized))
+        else:
+            stream.write(serialized)
 
     def _read_with_length(self, stream):
         length = read_int(stream)