updated: 2022-09-28
#tools/python
What's wrong with this initialization of list?
array = [{}] * 3
for i in range(3):
array[i][f"{i}"] = i
print(array)
Opartor * is useful to specify the size of list filled with a value. But this operaor creates a view if the element is mutable, like dict and list. This means that array[0] and array[1] are pointers referencing the same object, so that manipulating array[0] changes array[1].
import numpy as np
array = np.zeros(5)
indices = np.array([1,1,2,2,3,3,3])
array[indices]+=1 # increment
print(array) # [0,1,1,1,0,0] or [0,2,2,3,0,0]?
Numpy increment does not double increment the array.
You have a list of person names like this:
names = ["Elizabeth", "John", "Elizabeth", "Aaron", "John", "John"]
Write one-line code to assign a unique integer ID for each different person name. You can use numpy or scipy.
import numpy as np
names = ["Elizabeth", "John", "Elizabeth", "Aaron", "John", "John"]
np.unique(names, return_inverse=True)[1]
Write a one line code to shuffle the rows of pandas's dataframe using pandas.
df.sample(frac = 1)
Numpy and pandas offer an easy way to draw a random sample in a set of samples. For instance,
numpy.random.choice(data, size = 1)
pandas.DataFrame(data).sample(1)
sample one data from data (which can be a list or numpy array).
This is convenient, though, extremely slow, especially when they are executed repeatedly. An easy word around is to sample index instead of the data itself:
index = numpy.random.randint(len(data))
data[index]
which is 100x faster (see here)
∏P