[R] Controlling visibility of top-level functions

Duncan Murdoch murdoch.duncan at gmail.com
Sat Feb 1 01:19:56 CET 2014


On 14-01-31 5:09 PM, Rolf Turner wrote:
> On 01/02/14 08:13, Duncan Murdoch wrote:
>> On 31/01/2014 1:46 PM, Paul A. Steckler wrote:
>>> I'm fairly new to R, and have checked the R FAQ and done an
>>> RSiteSearch for help
>>> on this topic, to no avail.
>>>
>>> I want to write some R code that has functions at the top-level that
>>> are not visible when
>>> the code is loaded. So in
>>>
>>>     fun1 <- function(...)  { ... }
>>>
>>>     fun2 <- function(...)  { ... fun1 ...}
>>>
>>> I'd like fun2 to be callable, but have fun1 be invisible. That is, the
>>> scope of fun1 is
>>> limited to the file in which it's defined.
>>>
>>> In Python, I believe that prepending an underscore to a variable name
>>> limits its scope in this way.
>>> Is there a similar mechanism in R?
>>
>> There are a couple ways.
>>
>> The heavyweight way is to write a package that exports fun2 but not
>> fun1.  fun2 can see fun1, but the rest of the world can't.
>> Scope isn't limited to one file, any function in the package can see it.
>>
>> The lightweight way is to define fun1 and fun2 in a local scope, e.g.
>>
>> fun3 <- local({
>>
>>     fun1 <- function(...)  { ... }
>>
>>     fun2 <- function(...)  { ... fun1 ...}
>>     fun2
>> })
>>
>> This way fun1 and fun2 can see each other but nobody else can see them,
>> and fun3 is a copy of fun2 that is visible in the workspace.  You don't
>> need a third name, I just changed it to make the explanation easier.
>
> Would it not be more perspicuous to do:
>
> fun2 <- local({
>
>       fun1 <- function(...)  { ... }
>
>       function(...)  { ... fun1 ...}
> })
>
> Having fun3 kicking around seems to be an effect brought to you by the
> Department of Redundancy Department.

Sure, that's how I'd write it, but then it's harder to figure out how to 
say "fun1 and fun2 can see each other".  (Of course, fun1 and fun2 can 
see fun3 in my version, but in a different way, and not necessarily in a 
more complicated example.)

Duncan




More information about the R-help mailing list