How to merge dictionaries in np.ndarray into one dictionary?

Question

Following my question on Convert Dataframe to Nested Dictionary in Python, I have been trying to convert a Pandas dataframe into a nested dictionary.

Current challenge is merging multiple dictionaries inside a np.ndarray into one.

My current output dkl is

array([{0.7863340563991272: 0.0002639915522703274},
       {0.7863340563991272: 0.0006863780359028511}], dtype=object)

Which when checked for type results numpy.ndarray.

I need to get dkl to be

{0.7863340563991272: 0.0002639915522703274,
 0.7863340563991272: 0.0006863780359028511}

I am failing to get the desired results. I tried to turn dkl into a l1 = list(dkl), resulting in:

[{0.7863340563991272: 0.0002639915522703274},
 {0.7863340563991272: 0.0006863780359028511}]

I am now able to get the two elements of the list above which are it1 = {0.7863340563991272: 0.0002639915522703274} and it2 = {0.7863340563991272: 0.0006863780359028511}, which I am trying to merge into one dictionary. I tried:

 dict(it1.items() + it2.items())

resulting in:

TypeError                                 Traceback (most recent call last)
<ipython-input-150-4422fd31b345> in <module>
----> 1 dict(it1.items() + it2.items())

TypeError: unsupported operand type(s) for +: 'dict_items' and 'dict_items'

In short, I need to get from 1 to 2. Update: Ref to mention in comments that this shape of output is not possible in dict, I am tyring to understand how this comes about:

Reason being that I have a dataset I need to replicate that is structured as shown above in my desired output, and if not possible with dict, I have no idea how it's been created...

That is not possible, you can't have a dictionary where the same key appears twice with different values. — mkrieger1, Jul 24 '20 at 00:13
In general, this should answer your question: [How do I merge two dictionaries in a single expression in Python (taking union of dictionaries)?](https://stackoverflow.com/questions/38987/how-do-i-merge-two-dictionaries-in-a-single-expression-in-python-taking-union-o) — mkrieger1, Jul 24 '20 at 00:16
P.S. Perhaps you should pick smaller numbers like 1, 2, 3, 4 to demonstrate with. — Mateen Ulhaq, Jul 24 '20 at 00:17
In earlier Python versions, `dict.items()` returned a list, which you could concatenate using the `+` operator. Presumably you were following some advice which was only valid for an older Python version. — mkrieger1, Jul 24 '20 at 00:17
@mkrieger1 I saw this in another post that it is not possible, so then how is it If I type in my desired output, and query the 'type' , it says 'dict' ? will show example in edited question. — El_1988, Jul 24 '20 at 00:23

hpaulj · Answer 1 · 2020-07-24T00:29:30.387

In [10]: dkl=np.array([{0.7863340563991272: 0.0002639915522703274}, 
    ...:        {0.7863340563991272: 0.0006863780359028511}], dtype=object)                          
In [11]: dkl                                                                                         
Out[11]: 
array([{0.7863340563991272: 0.0002639915522703274},
       {0.7863340563991272: 0.0006863780359028511}], dtype=object)

An object dtype array is for (most) practical purposes a list. Your list(dkl) does that explicitly, or tolist():

In [12]: alist = dkl.tolist()                                                                        
In [13]: alist                                                                                       
Out[13]: 
[{0.7863340563991272: 0.0002639915522703274},
 {0.7863340563991272: 0.0006863780359028511}]

One way way to merge dictionaries is update:

In [14]: d1 = alist[0]                                                                               
In [15]: for d in alist[1:]: d1.update(d)                                                            
In [16]: d1                                                                                          
Out[16]: {0.7863340563991272: 0.0006863780359028511}

In your case both have the same key, so the result is last dictionary in the list.

{0.7863340563991272: 0.0002639915522703274,
 0.7863340563991272: 0.0006863780359028511}

is impossible since keys of a dictionary are unique.

Other possible mergers include adding values with matching keys, or a list with values. I'm not sure I should spend time on those since you don't seem to know what you really need.

You try adding the items; using the py3 list(items), we get a list of lists of tuples:

In [20]: [list(d.items()) for d in dkl]                                                              
Out[20]: 
[[(0.7863340563991272, 0.0002639915522703274)],
 [(0.7863340563991272, 0.0006863780359028511)]]

where the tuples contain the key and value. But apart from concatenating I don't know what you want to do with those tuples.

In sum, you have a bunch of dictionaries, which may or may not have matching keys. There's nothing special, numpy-wise here.

thanks for your response, so the basic idea is that there is no way to create in python the dataset in the shape I'm looking for. Aim for this is: A model ingests data that looks like this, but has not been provided with documentation, so I need to reverse identify how its input (my desired output) been created from a base panda dataframe. The data when converted from dataframe to nested dictionary looks like what I've shared as desired output. Thus my confusion. — El_1988, Jul 24 '20 at 00:32

How to merge dictionaries in np.ndarray into one dictionary?

1 Answers1