rbind_all should raise exception if arguments aren't data.frames

[coolbutuseless]

I mistakenly tried to rbind_all() on a list of vectors, and did not get what I was expecting.

rbind_all() only seems to work with data.frames (as with the rest of dplyr)

Perhaps raise an exception in the case where arguments aren't data.frames? Or is the result I go expected somehow?

e.g.

> ll <- list(c(1,2,3,4, 5), c(6, 7, 8, 9, 10))
> do.call(rbind, ll)
     [,1] [,2] [,3] [,4] [,5]
[1,]    1    2    3    4    5
[2,]    6    7    8    9   10
> rbind_all(ll)
   c(1, 2, 3, 4, 5) c(6, 7, 8, 9, 10)
1                 1                NA
2                 2                NA
3                 3                NA
4                 4                NA
5                 5                NA
6                NA                 6
7                NA                 7
8                NA                 8
9                NA                 9
10               NA                10
> 

[romainfrancois]

What happens internally is a call to as.data.frame because I'm using a ListOf<DataFrame>. and:

> lapply( ll, as.data.frame )
[[1]]
  X[[1L]]
1       1
2       2
3       3
4       4
5       5

[[2]]
  X[[2L]]
1       6
2       7
3       8
4       9
5      10

Perhaps we can be stricter and indeed only allow data frames in. Although I've seen example of code using matrices as arguments to rbind_*

[hadley]

@romainfrancois I think in generally rbind_all should be pretty strict, and only accept data frames. We might later add a generic version that is more flexible.

[romainfrancois]

Done with the addition of the StrictListOf class template. Maybe there is space for a better error message:

> ll <- list(c(1,2,3,4, 5), c(6, 7, 8, 9, 10))
> rbind_all(ll)
Error: object at index 0 not compatible with class Rcpp::DataFrame_Impl<Rcpp::PreserveStorage>

[hadley]

Could we make the error message: "Object at index 1 not a data.frame"?

Also I think this broken NULL handling: https://travis-ci.org/hadley/dplyr/jobs/21027956#L1781

[romainfrancois]

I'll make the error message 1-index-based and more R friendly.

About the NULL handling, I forgot this requirement. So we ignore NULL, but otherwise we only accept data frames. I'll update in the morning.