BUG: fix loading from compressed .npz (Issue 4093)

In Python 2.6.x the number of bytes read from a zip file is 2^16, which
is less than the 2^18 requested by lib.format.py. This change handles
the case where fp.read() returns fewer than the requested number of bytes.

1 files changed, 9 insertions, 6 deletions

diff --git a/numpy/lib/format.py b/numpy/lib/format.py
index 4ac1427b4..c667273bc 100644
--- a/numpy/lib/format.py
+++ b/numpy/lib/format.py
@@ -476,13 +476,16 @@ def read_array(fp):
             max_read_count = BUFFER_SIZE // min(BUFFER_SIZE, dtype.itemsize)
 
             array = numpy.empty(count, dtype=dtype)
-
-            for i in range(0, count, max_read_count):
+            extra_data = ''
+            i = 0
+            while i < count: 
                 read_count = min(max_read_count, count - i)
-
-                data = fp.read(int(read_count * dtype.itemsize))
-                array[i:i+read_count] = numpy.frombuffer(data, dtype=dtype,
-                                                         count=read_count)
+                data = extra_data + fp.read(int(read_count * dtype.itemsize))
+                actual_count = len(data) // dtype.itemsize
+                extra_data = data[actual_count*dtype.itemsize:] 
+                array[i:i+actual_count] = numpy.frombuffer(data, dtype=dtype,
+                                                         count=actual_count)
+                i += actual_count
 
         if fortran_order:
             array.shape = shape[::-1]