4

I'm trying to create a new class that inherits from data frame:

> setClass('new.frame',
    representation(colour='character'),
    contains = 'data.frame')

This is an instance of that class, for testing:

> test_data = data.frame(cbind(runif(5), runif(5)))
> names(test_data) = c('X', 'Y')
> test_frame = new('new.frame', test_data, colour='red')

Just to make sure it looks all right...

> data.frame
Object of class "new.frame"
          X         Y
1 0.8766306 0.4741213
2 0.1221508 0.5117665
3 0.4838761 0.4973627
4 0.7858294 0.4064749
5 0.5147703 0.9135304
Slot "colour":
[1] "red"

... and to make sure that the inheritance worked

> is.data.frame(test_frame)
[1] TRUE
> getClass(class(test_frame))
Class "new.frame" [in ".GlobalEnv"]

Slots:

Name:                .Data              colour               names
Class:                list           character           character

Name:            row.names            .S3Class
Class: data.frameRowLabels           character

Extends: 
Class "data.frame", directly
Class "list", by class "data.frame", distance 2
Class "oldClass", by class "data.frame", distance 2
Class "vector", by class "data.frame", distance 3

Here is the problem I encountered when I tried to utilize the property of being a data frame:

> terms.formula(Y ~ X, data = test_frame)
Error in terms.formula(Y ~ X, data = test_frame) : 
  'data' argument is of the wrong type

I might have missed something silly. If so, thanks in advance for pointing it out.

If I'm correct about the problem here, is there anyway I can make terms.formula recognize the fact that I'm giving it a data.frame?

mnel
  • 113,303
  • 27
  • 265
  • 254
Centzon Totochtin
  • 157
  • 1
  • 3
  • 7

1 Answers1

1

Doing debug(terms.formula) and then running terms.formula(Y ~ X, data = test_frame) shows that your code is failing on lines 3 and 4 of the quoted code block:

if (!is.null(data) && !is.environment(data) && !is.data.frame(data)) 
    data <- as.data.frame(data, optional = TRUE)
terms <- .Internal(terms.formula(x, specials, data, keep.order, 
    allowDotAsName))

The problem must be that the call to .Internal(terms.formula()) expects a 'plain old' data.frame, and is instead being passed an object of class new.frame. As a workaround, why not just pass terms.formula() the type of object it expects (an unadorned data.frame)?

Here's one easy way to do that:

terms.formula(Y ~ X, data = unclass(test_frame))
# Y ~ X
# attr(,"variables")
# list(Y, X)
# attr(,"factors")
#   X
# Y 0
# X 1
# attr(,"term.labels")
# [1] "X"
# attr(,"order")
# [1] 1
# attr(,"intercept")
# [1] 1
# attr(,"response")
# [1] 1
# attr(,".Environment")
# <environment: R_GlobalEnv>
Josh O'Brien
  • 159,210
  • 26
  • 366
  • 455
  • I haven't dared brave tracing ( or understanding) how the sp package works, but a SpatialPointsDataFrame ( or SpatialxxxDataFrame) will happily work with terms.formula – mnel Oct 31 '12 at 11:52
  • 2
    @mnel -- That's because a `SpatialPointsDataFrame` fails the `is.data.frame()` test in the first line of the quoted code block above, triggering a call to `as.data.frame()` (and, really, `as.data.frame.SpatialPointsDataFrame()`). That returns an honest-to-goodness data.frame, as needed, so the call to `.Internal(terms.formula())` works just fine. (If you run `library(sp); example(meuse); is.data.frame(meuse); class(as.data.frame(meuse))`, you can see how this works.) – Josh O'Brien Oct 31 '12 at 12:11
  • Wonderful. Thanks Josh! and mnel too for the discussion! – Centzon Totochtin Oct 31 '12 at 18:13