Problems to Install Bottleneck on Amazon EC2 Instance

76 views
Skip to first unread message

willia...@gmail.com

unread,
Oct 16, 2013, 7:43:25 PM10/16/13
to bottl...@googlegroups.com
Hi there,

I've been trying to install Bottleneck on a Ubuntu 12.04 and Ubuntu 13.04 both 64 Bits instances on EC2 Amazon but couldn't make it work so far.

Even though the module gets installed and is recognized by python, only the following functions appear as options to be imported:

bench      benchmark  dtypes     slow       src        test       version 

The other ones do not appear.

Running a bottleneck.test() I got 55 errors such as:

======================================================================
ERROR: Failure: AttributeError ('module' object has no attribute 'move_nanmax')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 231, in generate
    for test in g():
  File "/home/ubuntu/downloads/Bottleneck-0.5.0/bottleneck/tests/move_test.py", line 112, in test_move_nanmax
    yield unit_maker, bn.move_nanmax, bn.slow.move_nanmax, 5
AttributeError: 'module' object has no attribute 'move_nanmax'

======================================================================
ERROR: Failure: AttributeError ('module' object has no attribute 'partsort')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 231, in generate
    for test in g():
  File "/home/ubuntu/downloads/Bottleneck-0.5.0/bottleneck/tests/partsort_test.py", line 65, in test_partsort
    yield unit_maker, bn.partsort, bn.slow.partsort
AttributeError: 'module' object has no attribute 'partsort'

======================================================================
ERROR: Failure: AttributeError ('module' object has no attribute 'argpartsort')
----------------------------------------------------------------------
Traceback (most recent call last):
  File "/usr/lib/python2.7/dist-packages/nose/loader.py", line 231, in generate
    for test in g():
  File "/home/ubuntu/downloads/Bottleneck-0.5.0/bottleneck/tests/partsort_test.py", line 69, in test_argpartsort
    yield unit_maker, bn.argpartsort, bn.slow.argpartsort
AttributeError: 'module' object has no attribute 'argpartsort'

----------------------------------------------------------------------
Ran 55 tests in 0.026s

FAILED (errors=55)
Out[2]: <nose.result.TextTestResult run=55 errors=55 failures=0>

I tried to install either version 0.7 and 0.5 but in both cases I had the same issue.

The output after trying to install both bottlenecks are attached to this message (in the file Bottleneck I tried to install Bottleneck 0.7.0 without the sudo command and got access denied, afterwards I ran the same command with a sudo)

Python version is 2.7.3, 64Bits. 

To install numy, scipy and everything else I ran the command:

sudo apt-get install python-numpy python-scipy python-matplotlib ipython ipython-notebook python-pandas python-sympy python-nose

Does anyone know what might be going on? I supposed there's some dependency causing the issue but I couldn't find what it could be.

Thanks in advance!

Will

Bottleneck.txt
Bottleneck 0.5.txt

Keith Goodman

unread,
Oct 17, 2013, 12:38:05 PM10/17/13
to bottl...@googlegroups.com
On Wed, Oct 16, 2013 at 4:43 PM, <willia...@gmail.com> wrote:

I've been trying to install Bottleneck on a Ubuntu 12.04 and Ubuntu 13.04 both 64 Bits instances on EC2 Amazon but couldn't make it work so far.

This is a known issue. So far no one has reported a solution. Here's a discussion (with many false turns):

 https://github.com/kwgoodman/bottleneck/issues/54


Even though the module gets installed and is recognized by python, only the following functions appear as options to be imported:

bench      benchmark  dtypes     slow       src        test       version 

The other ones do not appear.

That means func and move (the bottleneck cython functions) cannot be imported. Which, in turn, probably means they were not compiled.

I forget the lessons learned (if any) from issue 54. But I wonder if cutting bottleneck down to two functions, like nansum and move_nansum, would compile on a virtual machine.
 
Reply all
Reply to author
Forward
0 new messages