用基于自身的条件语句索引一个 numpy 数组是什么意思？

Question

I'm working on a large library for network analysis and have come across a perplexing line whose calling convention I'm not familiar with.

    monitors = [1,2,3,4]
    nmonitors = 7 # This value is passed arbitrarily to the function

    while len(monitors) < nmonitors:
        remaining = np.arange(len(routers)) # len(routers) here is == 5
        for i in monitors:
            remaining = remaining[remaining != i]
        monitors.append(np.random.choice(remaining))

The line in questions in inside the loop which indexes the remaining array by a conditional based on i and itself. After some debugging it seems to be doing more than just evaluating a bool and indexing the array using that boolean value?

Would anyone be familiar with this syntax/convention and able to point me to the relevant part of numpy documentation or explain? I've been searching for hours with no results still, thank you.

Answer 1

There no special syntax, just a combination of generating a boolean array with a conditional test, and indexing an array with a boolean.

A sample array:

In [125]: arr = np.arange(4)
In [126]: arr
Out[126]: array([0, 1, 2, 3])

Indexing with a boolean:

In [127]: arr[[True,False,True,False]]
Out[127]: array([0, 2])

Creating a boolean with a test:

In [128]: (arr%2)==0
Out[128]: array([ True, False,  True, False])
In [129]: arr[(arr%2)==0]
Out[129]: array([0, 2])

Or with a test like your example:

In [131]: arr!=2
Out[131]: array([ True,  True, False,  True])
In [132]: arr[arr!=2]
Out[132]: array([0, 1, 3])

So that inner loop is removing all elements equal to monitors from remaining , leaving only [0] ? The larger loop is buggy, but that has nothing to do with the "syntax" of the boolean indexing.

Answer 2

It returns a new array of Boolean values, the same dimension of the original remaining . For each original element of remaining , wherever that element is not equal to i , the equivalent index in the new array is true .