[Python] (scalar) vector reading speedup via numpy (#4390)

* Add numpy accessor to python flatbuffers scalar vectors

* Update python tests to test numpy vector accessor

* Update appveyor CI to run Python tests, save generated code as artifact

* Update example generated python code

* Add numpy info to python usage docs

* Update test schema and python tests w/ multi-byte vector

* did not mean to push profiling code

* adding float64 numpy tests

python/flatbuffers/encode.py

@@ -15,13 +15,26 @@
 from . import number_types as N
 from . import packer
 from .compat import memoryview_type
+from .compat import import_numpy, NumpyRequiredForThisFeature
 
+np = import_numpy()
 
 def Get(packer_type, buf, head):
-    """ Get decodes a value at buf[head:] using `packer_type`. """
+    """ Get decodes a value at buf[head] using `packer_type`. """
     return packer_type.unpack_from(memoryview_type(buf), head)[0]
 
 
+def GetVectorAsNumpy(numpy_type, buf, count, offset):
+    """ GetVecAsNumpy decodes values starting at buf[head] as
+    `numpy_type`, where `numpy_type` is a numpy dtype. """
+    if np is not None:
+        # TODO: could set .flags.writeable = False to make users jump through
+        #       hoops before modifying...
+        return np.frombuffer(buf, dtype=numpy_type, count=count, offset=offset)
+    else:
+        raise NumpyRequiredForThisFeature('Numpy was not found.')
+
+
 def Write(packer_type, buf, head, n):
-    """ Write encodes `n` at buf[head:] using `packer_type`. """
+    """ Write encodes `n` at buf[head] using `packer_type`. """
     packer_type.pack_into(buf, head, n)