-
Notifications
You must be signed in to change notification settings - Fork 125
Open
Description
This is related to #21.
The infix operators seem to break with what you would expect normal math operator precedence to have. I'm just wondering if this is the intended behavior or not.
Here's an example.
# module M = Owl.Mat;;
# module Op = Owl.Dense.Matrix.Operator;;
# let x = M.linspace 0. 10. 6
val x : M.mat =
C0 C1 C2 C3 C4 C5
R0 0 2 4 6 8 10
(* Okay *)
# Op.(x *$ 2. +$ 1.);;
- : (float, Bigarray.float64_elt) Owl_dense_matrix_generic.t =
C0 C1 C2 C3 C4 C5
R0 1 5 9 13 17 21
(* Not what you would expect... *)
# Op.(2. $* x +$ 1.);;
- : (float, Bigarray.float64_elt) Owl_dense_matrix_generic.t =
C0 C1 C2 C3 C4 C5
R0 2 6 10 14 18 22
I assume this is because of OCaml's rules about operator precedence and associativity. Particularly the fact that operators starting with +
(like +$
) have a higher precedence than those starting with $
(like $*
).
So then you have a case where scalar * matrix
multiplication ($*
) has a lower precedence than matrix * scalar
multiplication (*$
).
In particular, you can see that, scalar * matrix
multiplication ($*
) has a lower precedence than matrix + scalar
(+$
) addition.
Just for reference, here is numpy with the same example.
>>> import numpy as np
>>> x = np.linspace(0, 10, num=6)
>>> 2 * x + 1
array([ 1., 5., 9., 13., 17., 21.])
>>> x * 2 + 1
array([ 1., 5., 9., 13., 17., 21.])
Metadata
Metadata
Assignees
Labels
No labels