Avoiding the infamous "eval(parse())" construct

Question

Ok, so I'm running some loops to process data stored in list objects. Ever mindful of the infamous fortune admonishment not to use eval(parse(mystring)), I came up with this:

Rgames> bar
$foo
$foo$fast
[1] 1 2 3 4 5

$foo$slow
[1]  6  7  8  9 10


$oof
$oof[[1]]
[1]  6  7  8  9 10

$oof[[2]]
[1] 1 2 3 4 5


Rgames> rab<-'bar'
Rgames> do.call('$',list(as.name(rab),'oof'))
[[1]]
[1]  6  7  8  9 10

[[2]]
[1] 1 2 3 4 5

Typically I'd be selecting a list (of which bar is one such) and then one element of the list (e.g. oof) which contains my data. The code above does the same thing as eval(parse(text=paste(rab,'$','oof',sep=''))) .
I'm doing all this specifically because I want to use the lists' names rather than [[x]] notation as a safety mechanism (because not all list objects have their contents in the same order).

Should I stick with the advice from DWin in R: eval(parse(...)) is often suboptimal ?

Lilla

Answer 1

Using get and [[:

bar <- list(foo = list(fast = 1:5, slow = 6:10),
            oof = list(6:10, 1:5))

rab <- 'bar'

get(rab)[['oof']]
# [[1]]
# [1]  6  7  8  9 10
# 
# [[2]]
# [1] 1 2 3 4 5

Andy

Answer 2

If the name of your top list is going to change and be accessed by a variable with the name then it is best to put those lists into another list, then you can access the list you want using [[. Also read fortune(312) and the help on ?'[['.

You can then access the pieces in a different ways (detailed on the help page ?'[[').

mylist <- list()
mylist$bar <- bar

mylist[[rab]][['oof']]
#or
mylist[[ c(rab,'oof') ]]

Transcurrent

This is a very good point, but it does assume I'm organized enough :-) to know all the lists I'll be using in advance. – Lilla

@CarlWitthoft, any time that you go to use a list that is not already in your master list you can add it using get. – Transcurrent

Or do globalenv()[["mylist"]] - lists and environments work very similarly. – Yokoyama

@hadley, I considered making the same suggestion, but it should be noted (this is to others reading this, not Hadley) that that can get very dangerous if new users start overusing 'globalenv()` and similar rather than thinking through better approaches. – Transcurrent