Handle assignment into an environment

[clarkfitzg]

In the following I'd like to know that the environment e was updated:

> code = quote(assign("x", 20, envir = e))
> CodeDepends::getInputs(code)
An object of class "ScriptNodeInfo"

...
Slot "outputs":
[1] "x"

Slot "updates":
character(0)

Slot "sideEffects":
character(0)
...

This came up attempting to analyze the body of the koRpus::set.kRp.env() function which sets a bunch of variables in a private package environment koRpus:::.koRpus.env.

[gmbecker]

(now with proper formatting, stupid email reply...)

This is doable. We can make it the default (although environments are are
pass-by-reference so they violate assumptions we make all over the place
when, e.g., creating threads, tree-shaking, differentiating inputs vs
outputs/updates, etc).

The default assign() handler is assignfunhandler:

assignfunhandler = function(e, collector, basedir, input,
                            formulaInputs, update, pipe = FALSE,
                            nseval = FALSE, ...){
    if(is.symbol(e[[2]])) {
        warning("assign() used with symbol as first argument. Unable to
statically resolve what name the value will be assigned to")
        return()
    } else { ## character containing the name to assign to
        collector$calls("assign")
        if(is.character(e[[2]]))
           collector$set(e[[2]]) ##variable
        else
           collector$set(structure(as.character(NA), names =
deparse(e[[2]])))
        getInputs(e[[3]], collector = collector, basedir = basedir, input =
TRUE,
                  formulaInputs = formulaInputs, update = update, pipe =
pipe,
                  nseval = nseval, ... )
    }

}

We can add a check to see if the envir argument is set and if so, add it to
updates:

> code = "e = new.env(); assign('x', 5L, envir = e)"
> scr = readScript(txt = code)
> assignfunhandler2 = *<snip, see below>*
> col = inputCollector(assign = assignfunhandler2)
> getInputs(scr, col)
An object of class "ScriptInfo"
[[1]]
An object of class "ScriptNodeInfo"
Slot "files":
character(0)

Slot "strings":
character(0)

Slot "libraries":
character(0)

Slot "inputs":
character(0)

Slot "outputs":
[1] "e"

Slot "updates":
character(0)

Slot "functions":
new.env
  FALSE

Slot "removes":
character(0)

Slot "nsevalVars":
character(0)

Slot "sideEffects":
character(0)

Slot "code":
e = new.env()


[[2]]
An object of class "ScriptNodeInfo"
Slot "files":
character(0)

Slot "strings":
character(0)

Slot "libraries":
character(0)

Slot "inputs":
character(0)

Slot "outputs":
[1] "x"

*Slot "updates":*
*[1] "e"*

Slot "functions":
assign
 FALSE

Slot "removes":
character(0)

Slot "nsevalVars":
character(0)

Slot "sideEffects":
character(0)

Slot "code":
assign("x", 5L, envir = e)

We can think about making this the default, but you can experiment with the
behavior now, as above.

Here is the definition of assignhandler2 in a useful (cut/paste-able) form:

assignfunhandler2 = function(e, collector, basedir, input,
                            formulaInputs, update, pipe = FALSE,
                            nseval = FALSE, ...){
    if(is.symbol(e[[2]])) {
        warning("assign() used with symbol as first argument. Unable to
statically resolve what name the value will be assigned to")
        return()
    } else { ## character containing the name to assign to
        collector$calls("assign")
        if(is.character(e[[2]]))
           collector$set(e[[2]]) ##variable
        else
            collector$set(structure(as.character(NA), names =
deparse(e[[2]])))
        eind = numeric()
        if("envir" %in% names(e)) {
            eind = match("envir", names(e))
            collector$update(asVarName(e[[eind]]))
        }
        lapply(setdiff(seq(3, length(e)), eind), function(i) {
            getInputs(e[[i]], collector = collector, basedir = basedir,
input = TRUE,
                  formulaInputs = formulaInputs, update = update, pipe =
pipe,
                  nseval = nseval, ... )})
      }

}

--
Gabriel Becker, PhD
Associate Scientist (Bioinformatics)
Genentech Research

[clarkfitzg]

Thanks @gmbecker, I'll use this. I imagine the pass by reference semantics complicates things significantly.