I have an numpy array of complex numbers. So I want to insert zero at start of the array,
and shift the rest of the array one place forward.
example:
Simplest way:
a = np.array([1 + 2j, 5 + 7j])
a = np.insert(a, 0, 0)
Then:
>>> a
array([ 0.+0.j, 1.+2.j, 5.+7.j])
Note that this creates a new array, it does not actually insert the 0
into the original array.
There are several alternatives to np.insert, all of which also create a new array:
In [377]: a
Out[377]: array([ 1.+2.j, 5.+7.j])
In [378]: np.r_[0, a]
Out[378]: array([ 0.+0.j, 1.+2.j, 5.+7.j])
In [379]: np.append(0, a)
Out[379]: array([ 0.+0.j, 1.+2.j, 5.+7.j])
In [380]: np.concatenate([[0], a])
Out[380]: array([ 0.+0.j, 1.+2.j, 5.+7.j])
In [381]: np.hstack([0, a])
Out[381]: array([ 0.+0.j, 1.+2.j, 5.+7.j])
In [382]: np.insert(a, 0, 0)
Out[382]: array([ 0.+0.j, 1.+2.j, 5.+7.j])
Additionally, if you want to add n numbers of rows with zero values. you can create a zero array and use hstack:
zero_array = np.zeros([n])
new = np.hstack([zero_array,old_array])
An alternative is "horizontal stack" (also creates a new array):
np.hstack((0,a))
I timed all the five different methods to insert an element at the beginning of an array. Here are the results:
In [20]: %timeit np.hstack([1, [1, 2, 3]])
10000 loops, best of 3: 30.4 µs per loop
In [21]: %timeit np.insert([1, 2, 3], 0, 1)
10000 loops, best of 3: 46.6 µs per loop
In [22]: %timeit np.r_[[1], [1, 2, 3]]
10000 loops, best of 3: 32.8 µs per loop
In [28]: %timeit np.append(1, [1, 2, 3])
10000 loops, best of 3: 23.4 µs per loop
In [29]: %timeit np.concatenate([[1], [1, 2, 3]])
The slowest run took 6.43 times longer than the fastest. This could mean that an intermediate result is being cached.
100000 loops, best of 3: 8.79 µs per loop
Also, if you have an n-dimensional array, you need to specify the axis as well, otherwise it gets flattened out:
np.insert(my_array, 0, myvalue, axis=1)