[Numpy-discussion] memmap, write through and flush

[Tom Kuiper]

There is something curious here.  The second flush() fails.  Can anyone 
explain this?

Tom

------------------- code snippet ------------------------
...
# create a memmap with dtype and shape that matches the data
fp = np.memmap(filename, dtype='float32', mode='w+', shape=(3,4))
print "Initial memory mapped array (mode 'w+'):\n",fp

# write data to memmap array
fp[:] = data[:]
fp.flush()
# append a row to the array
fp = np.append(fp, [[12,13,14,15]], 0)
print "Filled memory mapped array:\n",fp
fp.flush()
...

-----------------  output -----------------------

Filled memory mapped array:
[[  0.   1.   2.   3.]
 [  4.   5.   6.   7.]
 [  8.   9.  10.  11.]
 [ 12.  13.  14.  15.]]
---------------------------------------------------------------------------
AttributeError                            Traceback (most recent call last)

..../memmap_test.py in <module>()
     21 fp = np.append(fp, [[12,13,14,15]], 0)
     22 print "Filled memory mapped array:\n",fp
---> 23 fp.flush()
     24
     25 # deletion flushes memory changes to disk before removing the object

AttributeError: 'numpy.ndarray' object has no attribute 'flush'
WARNING: Failure executing file: <memmap_test.py>