[Python] Difference between np.argsort and index_natsorted natural sort

I have the below dataframe and I want to sort naturally by 'time' column.

df = pd.DataFrame({
"time": ['0hr', '128hr', '72hr', '48hr', '96hr'],
"value": [10, 20, 30, 40, 50]
})

df.sort_values(
by="time",
key=lambda x: index_natsorted(df["time"])

)


I use the natsort.index_natsorted function which is supposed to return the 'time' column in the sequence of ['0hr','48hr','72hr','96hr','128hr'], because the index_natsorted, applied to the given 'time' column, returns a tuple of [0, 3, 2, 4, 1], which suppose to sort the column in the above manner, with 0 in tuple referring to the first element '0hr' of the 'time' column, and the subsequent 3 in tuple referring to the fourth element '48hr' of the 'time' column, and so on...

However the output I got is:

output:
time value
0hr 10
96hr 50
72hr 30
128hr 20
48hr 40



Could someone explain to me why the index_natsorted function will return the 'time' column in this sequence?

Similar to np.argsort, the docs of index_natsorted reads that it " Sorts a sequence naturally, but returns a list of sorted the indexes and not the sorted list itself." So I am curious as to why in this case its returned indices will not sort the elements naturally.

Continue reading...