Programming with env= misses GForce with parametrically-supplied functions

[MichaelChirico]

DT=data.table(a=1:2, b=3:4)
DT[, f(b), env = list(f=var), verbose=TRUE]

(currently errors, which #6029 addresses)

Has output:

Argument 'j' after substitute: (function (x, y = NULL, na.rm = FALSE, use)  {     if (missing(use))          use <- if (na.rm)              "na.or.complete"         else "everything"     na.method <- pmatch(use, c("all.obs", "complete.obs", "pairwise.complete.obs", "everything", "na.or.complete"))     if (is.na(na.method))          stop("invalid 'use' argument")     if (is.data.frame(x))          x <- as.matrix(x)     else stopifnot(is.atomic(x))     if (is.data.frame(y))          y <- as.matrix(y)     else stopifnot(is.atomic(y))     .Call(C_cov, x, y, na.method, FALSE) })(b)

Obviously there's no way GForce is going to pick up that this could be gvar once it analyzes that convoluted j.

Are we stuck here? Is there no way to patch in the name var so that the outcome of env substitution becomes j=var(b) and then [ can just do normal processing to find GForce applies?

[jangorecki]

You just pass function as name, because that what you actually want to do. Variable to provide programmatically is the name of the function, not a function itself.

[MichaelChirico]

You just pass function as name, because that what you actually want to do. Variable to provide programmatically is the name of the function, not a function itself.

could you give an example? GForce works on neither of these:

DT[, f(b), env=list(f='var'), verbose=TRUE]
# Argument 'j' after substitute: var(b)
# Detected that j uses these columns: [b]
# [1] 0.5

DT[, f(b), env=list(f=as.name('var')), verbose=TRUE]
# Argument 'j' after substitute: var(b)
# Detected that j uses these columns: [b]
# [1] 0.5

[jangorecki]

Try adding by so GForce can work :)

[MichaelChirico]

🤦‍♂️ oops

Confirming that works:

DT[, f(b), env = list(f='var'), verbose=TRUE, by=a]
# ...
GForce optimized j to 'gvar(b)' (see ?GForce)
# ...

DT[, f(b), env = list(f=as.name('var')), verbose=TRUE, by=a]
# ...
GForce optimized j to 'gvar(b)' (see ?GForce)
# ...

[MichaelChirico]

@jangorecki can I task you with making a note about the importance of passing string/name to [ in the programming vignette? It seems very important to call out. I glanced at the vignette but it did not jump out at me where to add this note, exactly, I guess you have a better intuition for this as vignette author.

Should we also add a regression test about this to close this issue (if there's not one already)?