No method matching getindex() using ODEProblem

General Usage
, ,
1

Hey there,

I am very new to Julia and I am working on an ODE Problem. Basically this is my set-up:

etaspan = (0.0, 1.0)
F0 = [1, -1e+10, q_, 0] 

function eventfcn(vcondition, value, eta, integrator)
    out[1] = (qmax-F(3))   
    out[2] = F(2)
    out[3] = F(4)
end

function affect!(integrator, idx)
    if (idx == 1 | idx == 2 | idx == 3)
      affect!(integrator) = terminate!(integrator)
    end
end

cb = VectorContinuousCallback(eventfcn,affect!,3)    

# odefun(eta,f) will be a vector containing 4 partial derivatives (ODEs), that I
# calculate before with the function fnct
odefun(eta, f) = fnct(eta, f, r, rho, a, a_, delta, delta_, sigma)[1]

# dyn will be a vector containing all dynamic parameters. I also get them through function fnct
dyn(eta, f) = fnct(eta, f, r, rho, a, a_, delta, delta_, sigma)[2]

QL = 0
QR = 1e+15

for iter = 1:50
    F0[4] = (QL+QR)/2
    prob = ODEProblem(odefun, F0, etaspan, dyn)
    sol = solve(prob, Tsit5(), callback = cb, RelTol = 1e-08, abstol = 1e-10)
end

However, when trying to run all of this, I get the following error message:

MethodError: no method matching getindex(::typeof(dyn), ::Int64)

As far as I understood, the ODEProblem will at one point call function getindex, but I don’t know why it doesn’t work. Any help would be highly appreciated! <3

2

That’s not a vector, it’s a function.

3

Hey Chris. It’s actually a vector. The function “fnct” gives me 2 vectors as an output. dyn will be the second of the 2. I just didn’t want to include the “fnct” function cause it would be too long and it’s not necessary for the purpose here…

4

No, dyn is function that takes two arguments eta and f. This is how you define it.

Are you trying to have time-dependent parameters? If so, you should shift the time dependence into odefun

5

No, it’s not a vector. If you wanted a vector you would write:

dyn = fnct(eta, f, r, rho, a, a_, delta, delta_, sigma)[2]

and that’s exactly your error.

6

Hey @ChrisRackauckas and @skleinbo. Thank you very much for your answers.

Indeed, the getindex() error message disappeared once I defined dyn as a vector. But from that I got some other problems and they related to what you mentioned @skleinbo. Function fnct(eta, f, …) does 2 things. It calculates fp, the vector of derivatives and it will update the parameters at every iteration, which is what you @skleinbo were referring to (just that in my model, everything evolves with eta rather than with time).

What I can’t figure out is how to integrate these dynamic parameters in my model. This is what happened when I defined dyn as a global inside the fnct function (instead of returning it):

MethodError: no method matching +(::Float64, ::Vector{Float64}) 
For element-wise addition, use broadcasting with dot syntax: scalar .+ array

Thanks again !

7

a .+ b when it’s between a Float64 and a Vector{Float64} to do it element-wise.