error using all()/numpy [TypeError: cannot perform reduce withflexible type]

[Marc Oldenhof]

Hello all,

I'm pretty new to Python, but use it a lot lately. I'm getting a crazy
error trying to do operations on a string list after importing numpy.
Minimal example:

[start Python]

Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32 bit
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.

a=['1','2','3']
all(a) True
from numpy import *
all(a)

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
982, in all
return _wrapit(a, 'all', axis, out)
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
37, in _wrapit
result = getattr(asarray(obj),method)(*args, **kwds)
TypeError: cannot perform reduce with flexible type

[end of example]

It seems that Python calls numpy's "all" instead of the standard one,
is that right? If so, how can I call the standard "all" after the
numpy import? ["import numpy" is not a desirable option, I use a lot
of math in my progs]

Is there another way around this error?

Marc

 

[Marc Oldenhof]

Hello all,

I'm pretty new to Python, but use it a lot lately. I'm getting a crazy
error trying to do operations on a string list after importing numpy.
Minimal example:

[start Python]

Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32 bit
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.

a=['1','2','3']
all(a) True
from numpy import *
all(a)

Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
  File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
982, in all
    return _wrapit(a, 'all', axis, out)
  File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
37, in _wrapit
    result = getattr(asarray(obj),method)(*args, **kwds)
TypeError: cannot perform reduce with flexible type

[end of example]

It seems that Python calls numpy's "all" instead of the standard one,
is that right? If so, how can I call the standard "all" after the
numpy import? ["import numpy" is not a desirable option, I use a lot
of math in my progs]

Is there another way around this error?

Marc

Never mind, I found a way. For reference:
True

Works!

Marc

 

[Gary Herron]

Hello all,

I'm pretty new to Python, but use it a lot lately. I'm getting a crazy
error trying to do operations on a string list after importing numpy.
Minimal example:

[start Python]

Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32 bit
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.

a=['1','2','3']
all(a)

True

from numpy import *
all(a)

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
982, in all
return _wrapit(a, 'all', axis, out)
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
37, in _wrapit
result = getattr(asarray(obj),method)(*args, **kwds)
TypeError: cannot perform reduce with flexible type

[end of example]

It seems that Python calls numpy's "all" instead of the standard one,
is that right? If so, how can I call the standard "all" after the
numpy import? ["import numpy" is not a desirable option, I use a lot
of math in my progs]

Is there another way around this error?

Yes, there are several solutions, but before that I'll say that "from
.... import *" is frowned upon for just this reason. You have no
control (and often no idea) what * ends up importing, and if any of
those names overwrite an existing name (as you've found here), you may
not notice. (It's not quite fair to say "Python calls numpy's "all".
*You* call it after you chose to replace Python's "all" with numpy's "all".)

Solutions:



Save Python's "all" first under another name:
original_all = all
from numpy import all
Now you can call all(...) or original_all(...).



The built-in (as they are called) are always available through __builtins__:
from numpy import *
all(...) # is numpy's all
__builtins__.all(...) # Is the original all




Don't import *, but rather import only those things you need.
from numpy import array, dot, float32, int32, ...
and if you need numpy's all
from numpy import all as numpy_all


Gary Herron

 

[Peter Otten]

I'm pretty new to Python, but use it a lot lately. I'm getting a crazy
error trying to do operations on a string list after importing numpy.
Minimal example:

[start Python]

Python 2.5.1 (r251:54863, Apr 18 2007, 08:51:08) [MSC v.1310 32 bit
(Intel)] on win32
Type "help", "copyright", "credits" or "license" for more information.

a=['1','2','3']
all(a) True
from numpy import *
all(a)

Traceback (most recent call last):
File "<stdin>", line 1, in <module>
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
982, in all
return _wrapit(a, 'all', axis, out)
File "C:\Python25\lib\site-packages\numpy\core\fromnumeric.py", line
37, in _wrapit
result = getattr(asarray(obj),method)(*args, **kwds)
TypeError: cannot perform reduce with flexible type

[end of example]

It seems that Python calls numpy's "all" instead of the standard one,
is that right? If so, how can I call the standard "all" after the
numpy import? ["import numpy" is not a desirable option, I use a lot
of math in my progs]

Is there another way around this error?

That's not an error; star-import is a rebinding operation, just like
assignments and the def statement.
True

Peter

 

[I V]

It seems that Python calls numpy's "all" instead of the standard one, is
that right? If so, how can I call the standard "all" after the numpy
import? ["import numpy" is not a desirable option, I use a lot of math
in my progs]

I think the ideal solution is to try and persuade yourself that typing
"numpy." in front of some functions now and then is not that big a price
to pay to avoid name collisions; using an editor with code completion
would probably help in that task.

You could also try:

import numpy as n

which reduces the typing a bit and limits you to one possible name
collision, or

from numpy import array, gradient #and whatever else you need

or

import numpy

array = numpy.array
gradient = numpy.gradient # Then you can access the names you use a lot
# directly, while accessing stuff you use less
# frequently via numpy.whatever

in which case you'll know exactly which names you're overwriting. Peter's
solution, of explicitly deleting some names that you've imported, strikes
me as less good, because you might find the same problem recurs later, if
the numpy developers add new names to the module.

 

[Marc Oldenhof]

<snip my post>

Thanks for the reactions, I'll use the "from numpy import <individual
stuff>" from now on

Marc