[R] How to pass selection criteria in a function

[David Winsemius]

On Dec 1, 2010, at 1:12 PM, Charles C. Berry wrote:

> On Wed, 1 Dec 2010, CMcCarthy at bmcc.cuny.edu wrote:
>
>> Hi,
>> Suppose I have the following data
>>
>> name     score
>> Abel        88
>> Baker      54
>> Charlie    77
>>
>> stored a  table called myData.
>>
>>
>> I want to write a function that will create a table which is a  
>> subset of myData containing those have a score > 75.
>>
>> I know I can do this with the following command:
>> subset(myData, score > 75)
>>
>> But I would like to do this via a function, something like:
>>
>> newTable <- function( data,  criteria){
>>                     subset( data, criteria) }
>>
>> and then calling:     newTable(myData, score > 75)
>>
>> But this doesn't work. I am sure there is a simple way to do this,
>> but I am stuck! Please help. Thanks!
>
> Simple? Maybe not so much!
>
> You are trying to pass objects without evaluating them. subset is  
> rather special in the way it works. Here is one way:
>
>> foo <- function(x,...){
> + mc <- match.call()
> + mc[[1]] <- as.name("subset")
> + eval(mc)
> + }
>> foo(iris, Petal.Width>2.4 )
>    Sepal.Length Sepal.Width Petal.Length Petal.Width   Species
> 101          6.3         3.3          6.0         2.5 virginica
> 110          7.2         3.6          6.1         2.5 virginica
> 145          6.7         3.3          5.7         2.5 virginica

I've never really understood how to use browser() to develop or  
resolve problems with function definitions. Chuck's comments pointing  
to the subset function offered another opportunity to teach myself  
something new. Here's the console transcript:

 > newTable <- function( data, criteria){  crit <-  
substitute(criteria);browser()}
# Obviously not a complete function
 > newT <- newTable(myData, score < 75)
Called from: newTable(myData, score < 75) #so far so good
Browse[1]> criteria                  # I realize now that this should  
have been "crit"
Error: object 'score' not found      # took another look at subset()  
code
Browse[1]> newTable <- function( data, criteria){  crit <-  
substitute(criteria); logvec <- eval(crit, data, parent.frame());  
browser()}
Browse[1]> criteria              # still didn't catch on that "crit"  
was the local object to examine
Error: object 'score' not found  # just slow I guess.
In addition: Warning message:
restarting interrupted promise evaluation
Browse[1]> c      # was worried the the redefinition might not take  
effect at the top-level

# Try # 2
 > newTable <- function( data, criteria){  crit <-  
substitute(criteria); logvec <- eval(crit, data, parent.frame());  
browser()}
 > newT <- newTable(myData, score < 75)
Called from: newTable(myData, score < 75)
Browse[1]> logvec
[1] FALSE  TRUE FALSE                # now we're getting results
Browse[1]> return(data[logvec, ])    # see if the naive next step works
 > newT
    name score
2 Baker    54                        # very promising
 > newTable <- function( data, criteria){  crit <-  
substitute(criteria); logvec <- eval(crit, data, parent.frame());  
return(data[logvec, ])}
 > newT <- newTable(myData, score < 75)
 > newT                                # SUCCESS
    name score
2 Baker    54

There is not much in the way of error checking, but it seems to be a  
reasonable start (and looks to offer an example, albeit with a some  
newbie errors, of an extremely useful R tool.)

>
> Reading the code at the top of lm shows how this kind of strategy  
> can be used.
>>
>
> Charles C. Berry                            Dept of Family/ 
> Preventive Medicine

Thanks, Chuck.

-- 

David Winsemius, MD
West Hartford, CT