add! does not work to add two TaylorN

[dpsanders]

julia> x, y = set_variables("x y")
2-element Array{TaylorSeries.TaylorN{Float64},1}:
  1.0 x + 𝒪(‖x‖⁷)
  1.0 y + 𝒪(‖x‖⁷)

julia> s = (x+y)^3
 1.0 x³ + 3.0 x² y + 3.0 x y² + 1.0 y³ + 𝒪(‖x‖⁷)

julia> TaylorSeries.add!(s, 3x)
ERROR: MethodError: no method matching add!(::TaylorSeries.TaylorN{Float64}, ::TaylorSeries.TaylorN{Float64})
Closest candidates are:
  add!(::TaylorSeries.TaylorN, ::TaylorSeries.TaylorN, ::Int64) at /Users/dpsanders/.julia/v0.6/TaylorSeries/src/arithmetic.jl:121
  add!(::TaylorSeries.TaylorN, ::TaylorSeries.TaylorN, ::TaylorSeries.TaylorN, ::Int64) at /Users/dpsanders/.julia/v0.6/TaylorSeries/src/arithmetic.jl:129
  add!(::TaylorSeries.TaylorN, ::TaylorSeries.TaylorN, ::Union{Complex, Real}, ::Int64) at /Users/dpsanders/.julia/v0.6/TaylorSeries/src/arithmetic.jl:133
  ...

[dpsanders]

Maybe I'm misunderstanding what add! is for, but I actually need this!

[lbenet]

The basic idea of add!(res, a, b, k) is to modify res::TaylorN (for your example), by computing the addition of a and b (both TaylorNs) for the coefficient k, so it computes res[k]=a[k]+b[k] as an in-place operation. There are different methods for add!.

I am not sure what are you after, but the following works:

julia> TaylorSeries.add!(s, 3x, 1)

julia> s
 3.0 x + 1.0 x³ + 3.0 x² y + 3.0 x y² + 1.0 y³ + 𝒪(‖x‖⁷)

[dpsanders]

I want to add a TaylorSeries t in place to another one s (modifying s) via add!(s, t).

[dpsanders]

I don't understand what the coefficient k is for, e.g.

julia> s = (x + y)^2

julia> TaylorSeries.add!(s, x, 1)

julia> s
 1.0 x + 1.0 x² + 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

julia> TaylorSeries.add!(s, x, 2)

julia> s
 1.0 x + 𝒪(‖x‖⁷)

[lbenet]

The coefficient k is the only you compute. So you need something like

for k in 0:s.order
    TaylorSeries.add!(s, s, t, k)
end
s

For example:

julia> s = (x + y)^2
 1.0 x² + 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

julia> t = (x-y)^2
 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

julia> for k in 0:s.order
           TaylorSeries.add!(s, s, t, k)
       end

julia> s
 2.0 x² + 2.0 y² + 𝒪(‖x‖⁷)

julia> t
 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

[dpsanders]

OK, thanks. Would it make sense to add a method to add! to do this? On 23 Feb 2018 2:37 pm, "Luis Benet" <[email protected]> wrote: The coefficient k is the only you compute. So you need something like for k in 0:s.order TaylorSeries.add!(s, s, t, k)end s For example: julia> s = (x + y)^2 1.0 x² + 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷) julia> t = (x-y)^2 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷) julia> for k in 0:s.order TaylorSeries.add!(s, s, t, k) end julia> s 2.0 x² + 2.0 y² + 𝒪(‖x‖⁷) julia> t 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#156 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AALtTpBbGsyAqgUYNZL_EDXjgEKETSVvks5tXyGSgaJpZM4SRabQ> .

[lbenet]

I'm not sure what are you after, but in the example above, this is what you get with s+t.

[dpsanders]

s = s+t (or equivalently s += t) doesn't work in place and so is not possible if s is a field of an immutable object. Maybe we should allow s .= s+t for inplace add.

…

On 23 Feb 2018 3:54 pm, "Luis Benet" ***@***.***> wrote: I'm not sure what are you after, but in the example above, this is what you get with s+t. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#156 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AALtTtx5vxluUNd_j1F2MrVfgb9CVbB8ks5tXzOMgaJpZM4SRabQ> .

[lbenet]

It does work, though not as an in-place operation.

julia> using TaylorSeries

julia> x, y = set_variables("x y")
2-element Array{TaylorSeries.TaylorN{Float64},1}:
  1.0 x + 𝒪(‖x‖⁷)
  1.0 y + 𝒪(‖x‖⁷)

julia> s = (x + y)^2
 1.0 x² + 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

julia> t = (x-y)^2
 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷)

julia> s = s+t
 2.0 x² + 2.0 y² + 𝒪(‖x‖⁷)

[dpsanders]

My question is how to do that inplace, though.

…

On 23 Feb 2018 4:03 pm, "Luis Benet" ***@***.***> wrote: It does work, though not as an in-place operation. julia> using TaylorSeries julia> x, y = set_variables("x y")2-element Array{TaylorSeries.TaylorN{Float64},1}: 1.0 x + 𝒪(‖x‖⁷) 1.0 y + 𝒪(‖x‖⁷) julia> s = (x + y)^2 1.0 x² + 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷) julia> t = (x-y)^2 1.0 x² - 2.0 x y + 1.0 y² + 𝒪(‖x‖⁷) julia> s = s+t 2.0 x² + 2.0 y² + 𝒪(‖x‖⁷) — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#156 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AALtTg7VEPuEXIumvTpvRkjRUSzHzPKcks5tXzWWgaJpZM4SRabQ> .

[lbenet]

Currently there is no specific method that does that. I guess you could add add!(s,t,u) (with no integer coefficient) to do the for-loop illustrated above.

The current use of all those in-place functions (add!, mul!, sin!, ...) is to optimize the allocations in TaylorIntegration.jl. I've got pretty far with that, but there are few issues which are unresolved; see this PR.