Arithmetic and Broadcasting

Element-wise operation

let a = Matft.arange(start: 1, to: 9, by: 2, shape: [2,2])
let b = Matft.arange(start: 1, to: 5, by: 1, shape: [2,2])
print(a)
/*
mfarray = 
[[	1,		3],
[	5,		7]], type=Int, shape=[2, 2]
*/
print(b)
/*
mfarray = 
[[	1,		2],
[	3,		4]], type=Int, shape=[2, 2]
*/
print(a+b)
/*
mfarray = 
[[	2,		5],
[	8,		11]], type=Int, shape=[2, 2]
*/
print(a-b)
/*
mfarray = 
[[	0,		-1],
[	-2,		-3]], type=Int, shape=[2, 2]
*/
print(a*b)
/*
mfarray = 
[[	1,		6],
[	15,		28]], type=Int, shape=[2, 2]
*/
print(a/b) // true division like Numpy: integer arrays give Float
/*
mfarray = 
[[	1.0,		1.5],
[	1.6666666,		1.75]], type=Float, shape=[2, 2]
*/

Broadcasting

When the shapes of two arrays differ, the one is broadcast automatically as in Numpy.

let a = Matft.arange(start: 1, to: 9, by: 2, shape: [2,2])
let c = MfArray([-100,100])
print(a+c)
/*
mfarray = 
[[	-99,		103],
[	-95,		107]], type=Int, shape=[2, 2]
*/

Operators

Matft	Numpy	Description
+, -, *, /	+, -, *, /	element-wise arithmetic
-a	-a	negative
*&	@	matrix multiplication
*+	np.inner	inner product
*^	np.cross	cross product
===, !==	==, !=	element-wise comparison (returns a Bool MfArray)
<, <=, >, >=	<, <=, >, >=	element-wise comparison
==	np.array_equal	whether all elements are equal (returns Bool)

caution

== compares the whole arrays and returns a Swift Bool. Use === for the element-wise comparison (== in Numpy).