@erik thank you. I know of Base.Cartesian, but I was not able to make it work for my use case.
@steven yes, that is what i was afraid of (I was trying to shy away from recursion). On Monday, September 28, 2015 at 8:23:48 PM UTC+2, Erik Schnetter wrote: > > Guiseppe > > In addition to using recursion, you can also use a macro to generate the > code. > > However, this functionality is available in the "Cartesian" module which > is part of Base Julia. You are probably looking for "@nloops". > > -erik > > > On Mon, Sep 28, 2015 at 11:24 AM, Giuseppe Ragusa <giusepp...@gmail.com > <javascript:>> wrote: > >> I am having problems (serious problems!) to deal with algorithms that >> boil down do nested for loops but whose number of loops is not known at >> compile time. As an example consider this: >> >> ``` >> function tmp() >> r = 3 >> n = 100 >> λ = zeros(r) >> u = rand(n, r) >> counter = 0 >> Y = Array(Float64, n) >> >> @inbounds for j_1 = 0:k >> >> for i = 1:n >> Y[i] = pow(u[i, 1], j_1) >> end >> >> λ[1] = j_1 >> for j_2 = 0:(k-j_1) >> λ[2] = j_2 >> for i = 1:n >> Y[i] *= pow(u[i, 2], j_2) >> end >> >> for j_3 = 0:(k-j_1-j_2) >> λ[3] = j_3 >> for i = 1:n >> Y[i] *= pow(u[i, 3], j_3) >> end >> counter += 1 >> println(λ, " ", " => ", j_1, j_2, j_3, " counter =>", >> counter) >> end >> end >> end >> end >> ``` >> >> This is what I want when `r = 3`. For larger `r = 4` there would be an >> additional loop. etc. Now, everything is complicated by the fact that `r` >> is not know at compile time. >> >> I have tried to generate the loops using the macros in Base.Cartesian >> and then using generated functions to accomplish what I want, but I cannot >> get what I want. The main difficulty is that Base.Cartesian I can't get the >> ranges I want). >> >> Does anybody have suggestions on how to deal with this? >> >> >> > > > -- > Erik Schnetter <schn...@gmail.com <javascript:>> > http://www.perimeterinstitute.ca/personal/eschnetter/ >
