Can diff work with arrays? `diff([cos(x),sin(x),x]` returns an error.

[johnalx]

I am trying to use SymEngine with arrays of expressions. See example below

using SymEngine
@vars x
diff( [cos(x), sin(x)], x)
ERROR: MethodError: no method matching diff(::Array{Basic,1}, ::Basic)

Even more important is to be able to generate a gradient matrix by differentiating each element of a vector expression with each symbolic variable.

@vars a b
expr = [ a^2+2*b^2, -a*b]
f = diff(expr, [a,b])
#  expected result 2×2 matrix: [ 2a -b; 4a -a]

[isuruf]

julia> using SymEngine

julia> @vars x y
(x, y)

julia> diff.([cos(x), sin(x)], x)
2-element Array{Basic,1}:
 -sin(x)
  cos(x)

[isuruf]

For the second example,

julia> @vars a b
(a, b)

julia> expr = [ a^2+2*b^2, -a*b]
2-element Array{Basic,1}:
 a^2 + 2*b^2
        -a*b

julia> diff.(expr, transpose([a, b]))
2×2 Array{Basic,2}:
 2*a  4*b
  -b   -a

[johnalx]

For the second example,

julia> @vars a b
(a, b)

julia> expr = [ a^2+2*b^2, -a*b]
2-element Array{Basic,1}:
 a^2 + 2*b^2
        -a*b

julia> diff.(expr, transpose([a, b]))
2×2 Array{Basic,2}:
 2*a  4*b
  -b   -a

Thank you. Care to explain why the dot notation is necessary here? I am new to Julia notation. I understand that in the context of matrix multiplication .* does it by element (like Matlab) and how * operates in Fortran. I am also familiar with the symbolic toolbox in Matlab which does not require a vectoring dot modifier for diff(expr,vars)

[jverzani]

The "dot" does broadcasting (cf. ?broadcast) and is more generally integrated into the language than MATLAB where it is mostly with operators (as I remember, it has been a while). Matlab has most functions vectorized, whereas in Julia, one indicates vectorization through broadcasting (the dot).

[Gyoshi]

Seems like this should be closed

[isuruf]

Closing as this is how Julia works