Checking if two dictionaries of key-val pairs with values as array are equal

Question

I have two dictionaries as:

a = {0: array([4.5, 5. ]), 1: array([3.5, 4.5]), 2: array([1., 1.])}
b = {0: array([4., 5. ]), 1: array([3, 4]), 2: array([1.5, 1.])}

How do I check if these two dictionaries are equal? I tried:

a==b

ValueError: The truth value of an array with more than one element is ambiguous. Use a.any() or a.all()

Also,

(a==b).all()
a.all()==b.all()

All return errors.

See https://stackoverflow.com/questions/10580676/comparing-two-numpy-arrays-for-equality-element-wise for how to test if arrays are equal. — Barmar, Jun 01 '19 at 21:54
And see https://stackoverflow.com/questions/53348959/python3-determine-if-two-dictionaries-are-equal to test if dictionaries are equal — Barmar, Jun 01 '19 at 21:55
Put the two of them together to test if a dictionary containing arrays is equal. — Barmar, Jun 01 '19 at 21:56
@BryceKille Yes the order does matter, they are co-ordinates to points. — Rasika, Jun 01 '19 at 21:58
@Barmar I am unable to combine those two. I would appreciate any help if you could. — Rasika, Jun 01 '19 at 21:59
If they're coordinates, why don't you use tuples instead of numpy arrays? — Barmar, Jun 01 '19 at 22:00
I have mentioned what all I tried. And I need them to be numpy array hence I have kept them so. — Rasika, Jun 01 '19 at 22:05
All you showed was how to do it with arrays, you didn't combine it with the code for dictionaries. — Barmar, Jun 02 '19 at 00:14

hpaulj · Accepted Answer · 2019-06-01T22:57:55.820

== with dictionaries compares keys and values. But the values are arrays. array1 == array2 produces a boolean array, which does not play well with the yes/no expectations of the dictionary test (the ambiguity error).

A way around that is to compare the values individually. np.allclose is the best test for float arrays. Assuming that the keys match, the following list comprehension works nicely:

In [177]: array=np.array                                                             
In [178]: a = {0: array([4.5, 5. ]), 1: array([3.5, 4.5]), 2: array([1., 1.])} 
     ...: b = {0: array([4., 5. ]), 1: array([3, 4]), 2: array([1.5, 1.])}           
In [179]: a                                                                          
Out[179]: {0: array([4.5, 5. ]), 1: array([3.5, 4.5]), 2: array([1., 1.])}
In [180]: b                                                                          
Out[180]: {0: array([4., 5.]), 1: array([3, 4]), 2: array([1.5, 1. ])}
In [181]: [np.allclose(a[k],b[k]) for k in a]                                        
Out[181]: [False, False, False]
In [182]: [np.allclose(a[k],a[k]) for k in a]                                        
Out[182]: [True, True, True]

There should be another lay of testing, for equal keys.

However allclose does not work if the arrays differ in shape:

In [183]: c = {0: array([4., 5., 0 ]), 1: array([3, 4]), 2: array([1, 1.])}          
In [185]: [np.allclose(a[k],c[k]) for k in a]                                        
....
ValueError: operands could not be broadcast together with shapes (2,) (3,)

The comparison task will be a lot simpler if you know how the dictionaries might differ. Can they differ in keys? Can they differ in types of the values (array vs a list vs a number)? If values are arrays, can they differ in shape?

Checking if two dictionaries of key-val pairs with values as array are equal

1 Answers1