CODEWITHSUNDEEP
pythonpython-2.7numpy
Hassaan
Hassaan

Reputation: 13

Unhashable type error

Trying to do this:

ls = np.empty
ls =getColA()
rs = np.empty
rs=getColG()
x = dict(itertools.izip(ls,rs))

and getting this error:

TypeError: unhashable type: 'numpy.ndarray'

would appreciate if anyone could help me out.

Upvotes: 0

Views: 1754

Answers (3)

hpaulj
hpaulj

Reputation: 231665

There several things that don't make sense

ls = np.empty
ls =getColA()

np.empty is a function. You assign that to ls; then you assign the result of getColA() to ls. That wipes out the first assignment. Do you think the first assignment defines the ls variable as an array? That's not how Python works. Variables don't have a type.

x = dict(itertools.izip(ls,rs))

You didn't tell us what getColA() returns, but the error indicates that it is an array. 1d? 2d?

This sort of dictionary building works with 1d arrays (I'm using PY3)

In [497]: dict(itertools.zip_longest(np.arange(3),np.arange(10,13)))
Out[497]: {0: 10, 1: 11, 2: 12}

but not with 2d

In [498]: dict(itertools.zip_longest(np.arange(6).reshape(2,3),np.ones((2,3))))
...
TypeError: unhashable type: 'numpy.ndarray'

That's because the zip on 2d returns the rows of the arrays, and the rows themselves are arrays:

In [499]: list(itertools.zip_longest(np.arange(6).reshape(2,3),np.ones((2,3))
     ...: ))
Out[499]: 
[(array([0, 1, 2]), array([ 1.,  1.,  1.])),
 (array([3, 4, 5]), array([ 1.,  1.,  1.]))]

Why do you want to use an array as dictionary key?

Upvotes: 0

User2342342351
User2342342351

Reputation: 2174

A dict must have the key set as a hashable type.

You are trying to create a type using a non-hashable type (numpy.ndarray).

This is why you get this error.

In you case, you could do:

x = dict(itertools.izip(tuple(ls),rs))

Upvotes: 0

Kasravnd
Kasravnd

Reputation: 107347

When you are using itertools.izip(ls,rs) the first column will be contain numpy arrays which are not hashable to stand as dictionary keys.

You have to convert them to a hashable iterable like tuple instead. Also if you want to concatenate two numpy arrays along their second axis it's better to use np.column_stack() instead of itertools.izip().

Also It's not clear that what the getColA object is and what it's returning. Maybe you can modify that callable object in order to make a more proper result.

Upvotes: 2

Related Questions