return

return (each block), keyword

At the end of an each block, the return statement specifies the value to be returned.

table T = extend.range(5)
 
factorial = 1
T.Factorial = each T scan auto
  keep factorial
  factorial = factorial * T.N
  return factorial
 
show table "Factorials" a1b5 with T.N, T.Factorial
1
2
3
4
5
6
7
8
9

This return statement supports tuple values.

table T = extend.range(5)
 
sum = 0
factorial = 1
T.Sum, T.Factorial = each T scan auto
  keep sum
  keep factorial
  sum = sum + T.N
  factorial = factorial * T.N
  return (sum, factorial)
 
show table "Factorials" a1b5 with T.N, T.Sum, T.Factorial
1
2
3
4
5
6
7
8
9
10
11
12

return each, shorthand syntax

At the end of an each block, the return each syntax returns a vector value obtained through the execution of the each block.

table AB[ab] = with
  [| as ab |]
  [| "A"   |]
  [| "B"   |]
 
table XY[xy] = with
  [| as xy |]
  [| "X"   |]
  [| "Y"   |]
 
table C = cross(AB, XY)
C.AplusB = each AB
  return each XY 
    return concat(ab into Scalar, xy into Scalar)
 
show table "A+B" a1b6 with C.AplusB, ab, xy
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16

In the above script, return each is used to return a vector that belongs to the table XY for every line of the table AB. This script is equivalent the more verbose variant given below:

table AB[ab] = with
  [| as ab |]
  [| "A"   |]
  [| "B"   |]
 
table XY[xy] = with
  [| as xy |]
  [| "X"   |]
  [| "Y"   |]
 
table C = cross(AB, XY)
C.AplusB = each AB
  XY.AplusB = each XY 
    return concat(ab into Scalar, xy into Scalar)
  return XY.AplusB
 
show table "A+B" a1b6 with C.AplusB, ab, xy
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17

return (autodiff block), keyword

At the end of an autodiff block, the return statement specifies the loss associated to the step of the stochastic gradient descent. The loss must be a number.

autodiff Scalar epochs: 500 learningRate: 0.1 with
  params a auto
  s = a * a
  return s
 
show scalar "" with a // 0.00
1
2
3
4
5
6

This return statement supports tuple values. However, only the first value gets interpreted as the loss. The other values, after the first, must been numbers too, and are interpreted as metrics for reporting only.

return (user-defined function), keyword

At the end of a def block, the return statement specifies the value to be returned by the function or the process.

def pure square(x : number) with
  return x * x
 
show scalar "3^2" with square(3) // '9'
1
2
3
4