Improve structured array doc regarding extracting key-value pairs

[yasirroni]

Issue with current documentation:

From structured array doc, there is no a clear and concise tutorial to get key names or even the key-value pairs.

Currentyly, list of key tutorial only tell it implicitly:

d = np.dtype([('x', 'i8'), ('y', 'f4')])
d.names
# ('x', 'y')

Idea or request for content:

Tell user that the keys are under dtype, and to get key-value pairs, user need to do the following:

for key in x.dtype.names:
    value = x[key]

Also might be related, please

1. support dict() conversion:

   dict_ = dict(x)
   
   # equal to
   dict_ = {}
   for key in x.dtype.names:
       dict_[key] = x[key]

2. support .items() method to iterate over key-value pairs:

   for key, value in x.items():
       key, value
   
   # equal to
   for key in x.dtype.names:
       value = x[key]

I might be able to submit PR to implement that functionality.

[chethanreddy123]

Hey @yasirroni I would like to work on this, can you please guide me with some more insights on the issue, this is my first open-source contribution.

[Mukulikaa]

Hi @yasirroni, can you please elaborate on your issue? If you refer to this section in the doc, it says that the names and fields attributes of the dtype object that will give you the keys and key-value pairs for the structured datatype. I'm not sure if that's what you are looking for?

[rossbar]

IMO this is indeed covered in the section linked above. I'd be reticent to expand too much re: recommending translating between structured dtypes and dicts, as there may be better data structures suited to the cases, e.g. Dataframes.

[yasirroni]

@chethanreddy123 Hi, I'm not NumPy mainteainer, please ask them for that.

@Mukulikaa Hi, it seems that you don't understand the problem and you better try my piece of code and re-read the docs before further discussion.

@rossbar Hi, sadly, that section did not cover the way to retrieve the key-value, pairs. My use case is that I got structured numpy array from scipy.io.loadmat that give me a sturctured numpy array, but each array did not have any relationship with each other. They even have non matching size and shape! Thus, converting to dataframes is useless. From the doc, I know that I can get the list of keys from x.dtype.names, but the get all the data with the name, I need to use:

x = np.array([('Rex', 9, 81.0), ('Fido', 3, 27.0)],
             dtype=[('name', 'U10'), ('age', 'i4'), ('weight', 'f4')])

for key in x.dtype.names:
    val = x[key]
    print(key, val)

There should be a doc about that, or a better more, an easier way to do that.

Current docs only cover the name retireval and data retrieval separately.

[yasirroni]

import numpy as np
x = np.array([('Rex', 9, 81.0), ('Fido', 3, 27.0)],
             dtype=[('name', 'U10'), ('age', 'i4'), ('weight', 'f4')])
x
# array([('Rex', 9, 81.), ('Fido', 3, 27.)],
#       dtype=[('name', '<U10'), ('age', '<i4'), ('weight', '<f4')])

x.names
# AttributeError: 'numpy.ndarray' object has no attribute 'names'

x.fields
# AttributeError: 'numpy.ndarray' object has no attribute 'fields'

x['name']
# array(['Rex', 'Fido'], dtype='<U10')

for key in x.dtype.names:
    val = x[key]
    print(key, val)
# name ['Rex' 'Fido']
# age [9 3]
# weight [81. 27.]