按照列表中值的顺序过滤排序的 NumPy 数组的行

Question

I am trying to iterate over a NumPy array to create a list of lists but the for loop used is appending to the list of lists in alphabetical order rather than by the order of occurrence.

Here is a portion of my NumPy array that I can use as an example:

tarifas = np.array([['Afganistán', '577.21', '0.9360168799091559', '1.01745744495737'],
                    ['Albania', '5450.0', '1.1439867079655244', '0.9195410037811979'],
                    ['Alemania', '49690', '1.0034542200895549', '0.9873874704432137'],
                    ['Angola', '3670.0', '0.931103978746121', '1.162652536895962'],
                    ['Antigua y Barbuda', '18170', '0.7795684991736309', '0.6399312443495023'],
                    ['Arabia Saudita', '23490', '1.0573676413333202', '0.7477763277701148'],
                    ['Argelia', '4650.0', '0.7969840140783656', '0.5123046862189027'],
                    ['Argentina', '9050.0', '1.3647162509775996', '0.48274125735042017'],
                    ['Armenia', '4450.0', '1.4545784506262867', '1.430465487479917'],
                    ['Australia', '57200', '0.7293018985322222', '1.1744384938116095'],
                    ['Austria', '52470', '1.2396562976033307', '0.8630735107719588'],
                    ['Azerbaiyán', '4780.0', '0.9111186496911305','0.534268284966654']])

I want to create a list of lists using another list to iterate over which would have the specific name of the countries I need to find in the array, ie

list_countries = ["Angola", "Austria", "Argentina", "Albania", "Armenia"]

Notice how the list is not in alphabetical order, therefore the list of lists should respect this order. The output after iteration should be the following:

new_list_of_countries = [['Angola' '3670.0' '0.931103978746121' '1.162652536895962'], 
                         ['Austria' '52470' '1.2396562976033307' '0.8630735107719588'],  
                         ['Argentina' '9050.0' '1.3647162509775996' '0.48274125735042017'], 
                         ['Albania' '5450.0' '1.1439867079655244' '0.9195410037811979'], 
                         ['Armenia' '4450.0' '1.4545784506262867' '1.430465487479917']]

Here is the code I used:

tarifas_paises_escogidos = []
for i in tarifas:
    for v in list_countries:
         if str(v) in str(i):
               tarifas_paises_escogidos.append(i)
print(np.array(tarifas_paises_escogidos))

Answer 1

Using list comprehension with sorted :

sorted([t for t in tarifas if t[0] in list_countries], 
        key=lambda x: list_countries.index(x[0]))

Output:

[['Angola', '3670.0', '0.931103978746121', '1.162652536895962'],
 ['Austria', '52470', '1.2396562976033307', '0.8630735107719588'],
 ['Argentina', '9050.0', '1.3647162509775996', '0.48274125735042017'],
 ['Albania', '5450.0', '1.1439867079655244', '0.9195410037811979'],
 ['Armenia', '4450.0', '1.4545784506262867', '1.430465487479917']]

One without using list comprehension:

tarifas_paises_escogidos = []
for t in tarifas:
    # for v in list_countries: You don't need this
    if t[0] in list_countries:
        tarifas_paises_escogidos.append(t)
print(tarifas_paises_escogidos)

which yields filtered but unsorted:

[['Albania', '5450.0', '1.1439867079655244', '0.9195410037811979'], 
 ['Angola', '3670.0', '0.931103978746121', '1.162652536895962'], 
 ['Argentina', '9050.0', '1.3647162509775996', '0.48274125735042017'], 
 ['Armenia', '4450.0', '1.4545784506262867', '1.430465487479917'], 
 ['Austria', '52470', '1.2396562976033307', '0.8630735107719588']]

Then you sort (and do assign it back:):

tarifas_paises_escogidos = sorted(tarifas_paises_escogidos, key=lambda x: list_countries.index(x[0]))

which makes the above output.

Insight:

In the lambda above, x almost means nothing. It just means that what ever input lambda gets, it is defined as x , and used for indexing (ie x[0] ).

It is identical as:

def some_func(x):
    return list_countries.index(x[0])

then used in sorted :

tarifas_paises_escogidos = sorted(tarifas_paises_escogidos, key=some_func)

But you may often find defining a function for just one use case quite inefficient. That's when lambda kicks in:).

Answer 2

Since the original NumPy array, tarifas , is sorted alphabetically, you can use np.searchsorted to get the indices corresponding to the list_countries :

indices = np.searchsorted(tarifas[:, 0], list_countries)
print(indices)
# [ 3 10  7  1  8]

and then use fancy indexing (indexing arrays using arrays) to get the desired result:

result = tarifas[indices]
print(result)
# [['Angola' '3670.0' '0.931103978746121' '1.162652536895962']
#  ['Austria' '52470' '1.2396562976033307' '0.8630735107719588']
#  ['Argentina' '9050.0' '1.3647162509775996' '0.48274125735042017']
#  ['Albania' '5450.0' '1.1439867079655244' '0.9195410037811979']
#  ['Armenia' '4450.0' '1.4545784506262867' '1.430465487479917']]

For big arrays this vectorized approach should be much faster than the solution using Python's for-loops from the Chris's answer .