R - filter data with dplyr using a test condition

Question

I'm trying to use dplyr to filter my data based on a test condition, but this test condition can change depending on other variables.

Using the built in sample dataset cars:

data(cars)

I'd like to do something like this:

if (foo == 0) {
  test <- speed > 15
} else {
  test <- dist < 50
}
filter(cars, test)

This doesn't work. I can get it to work if I alter it to something like this:

if (foo == 0) {
  test <- 'cars$speed > 15'
} else {
  test <- 'cars$dist < 50'
}
filter(cars, eval(parse(text = test)))

But

1. Having to type out cars$speed and cars$dist seems to defeat the purpose of using the filter function.
2. According to this SO answer, using the eval(parse(text = ...)) construction is not recommended.

Is there a better way of achieving this?

Answer 1

You could do this:

filter(cars, if(foo==0){speed>15}else{dist<50})

Test by comparing with the simple filter:

> foo =0
> identical(filter(cars, speed>15), filter(cars, if(foo==0){speed>15}else{dist<50}))
[1] TRUE
> foo =1
> identical(filter(cars, dist<50), filter(cars, if(foo==0){speed>15}else{dist<50}))
[1] TRUE

It might just be easier and neater to put the filter statement inside the curly brackets:

if (foo == 0) {
  filter(cars, speed > 15)
} else {
  filter(cars, dist < 50)
}

Note if you want to assign the result somewhere, the if returns the value:

> ff = if (foo == 0) {
       filter(cars, speed > 15)
     } else {
       filter(cars, dist < 50)
     }
> identical(ff, filter(cars, speed>15))
[1] FALSE
> identical(ff, filter(cars, dist<50))
[1] TRUE
> foo
[1] 1

Answer 2

This works for me:

library(dplyr)

if (foo == 0) {
  test <- cars$speed > 15
} else {
  test <- cars$dist < 50
}

filter(cars, test)

I don't see a problem on using cars$speed and cars$dist just because you are using filter. Also, do you really need to use filter? There's an alternative to do this using base R. Replace the last line by:

cars[test,]