This code does not work:

library(multipol)

xyz <- function(px,py,pz,coef){

  A <- array(0, c(px+1,py+1,pz+1))

  P <- as.multipol(A)

  P[px,py,pz] <- coef

  P

}

> xyz(1,0,0,1)

 Error in eval(expr, p) : objet 'px' introuvable 

Show Traceback:

7.

eval(expr, p) 

6.

eval(expr, p) 

5.

eval.parent(ii, 3) 

4.

.handleTheOffset(mc, dim, offset, dn) 

3.

`[<-.multipol`(`*tmp*`, px, py, pz, value = 1) 

2.

`[<-`(`*tmp*`, px, py, pz, value = 1) 

1.

xyz(1, 0, 0, 1) 

I have found this workaround:

xyz <- function(px,py,pz,coef){

  A <- array(0, c(px+1,py+1,pz+1))

  P <- as.multipol(A)

  f <- function(){

    P[px,py,pz] <- coef

    P

  }

  f()

}

> xyz(1,0,0,1)

, , z^0



    y^0

x^0   0

x^1   1 

Nice? Hmm, wait. Now I have this function in a package:

representation <- function(P){

  monomials <- which(P!=0, arr.ind = TRUE) - 1

  kmax <- max(rowSums(monomials))

  allDegrees <- degreesTable(kmax)

  degrees <- row.match(as.data.frame(monomials), allDegrees)

  coefficients <- apply(monomials, 1, function(ijk){ 

    f <- function() P[ijk[1],ijk[2],ijk[3]]

    f()  

  })

  list(degrees=degrees, coefficients=coefficients)

}

> representation(P)

Error in eval(mc) : objet 'ijk' introuvable

However the trick works when it is not inside a function. This works:

coefficients <- apply(monomials, 1, function(ijk){ 

  f <- function() P[ijk[1],ijk[2],ijk[3]]

  f()  

})

The traceback:

8.

eval(mc) 

7.

eval(mc) 

6.

`[.multipol`(P, ijk[1], ijk[2], ijk[3]) at representation.R#38

5.

P[ijk[1], ijk[2], ijk[3]] at representation.R#38

4.

f() at representation.R#39

3.

FUN(newX[, i], ...) 

2.

apply(monomials, 1, function(ijk) {

    f <- function() P[ijk[1], ijk[2], ijk[3]]

    f()

}) at representation.R#37

1.

representation(P) 

How should I do? Any trick? I don't really understand code which uses eval.parent.

EDIT

I've found a solution:

  coefficients <- apply(monomials, 1, function(ijk){ 

    ijk <<- ijk

    P[ijk[1],ijk[2],ijk[3]]

  })

I'm still interested in a better solution and an explanation.

EDIT

Easier:

coefficients <- c(P[monomials])

But I'm still interested...