rbind_list/rbind_all should accept lists as inputs

[hadley]

Not entirely sure how this should work, but often you get lists that are structure like a data frame (e.g. each element is a vector of the same length), but don't have the write attributes. It would be nice to be able to convert directly to one data frame, rather than having to convert each one first.

x <- list(list(a = 1, b = 2), list(a = 3, b = 4))
rbind_all(x)

[romainfrancois]

As a first approximation, we could decide on the R side that the lists are rbindable, make data frames and then just call the regular rbind_all, e.g.

> x %>% lapply( as.data.frame )  %>% rbind_all
Source: local data frame [2 x 2]

  a b
1 1 2
2 3 4

[romainfrancois]

The concern I guess is "is the data copied" when passed through as.data.frame. Apparently not:

> x %>% sapply( function(.) sapply(., pryr::address ) )
  [,1]             [,2]
a "0x7f9f1c381c58" "0x7f9f1c381cb8"
b "0x7f9f1c381c88" "0x7f9f1c381ce8"
>
> x %>% lapply(as.data.frame) %>% sapply( function(.) sapply(., pryr::address ) )
  [,1]             [,2]
a "0x7f9f1c381c58" "0x7f9f1c381cb8"
b "0x7f9f1c381c88" "0x7f9f1c381ce8"

A benefit however is that we can leverage as.data.frame recycling:

> x <- list(list(a = 1, b = 2), list(a = 3, b = 4:6))
> x %>% lapply( as.data.frame )  %>% rbind_all
Source: local data frame [4 x 2]

  a b
1 1 2
2 3 4
3 3 5
4 3 6

[romainfrancois]

@kevinushey perhaps that means we need as_data_frame which would be to as.data.frame what data_frame is to data.frame

[hadley]

Good point. I'll write an minimal as_data_frame(). I think an R implementation should be fine

[hadley]

Speed difference is crazy! (and as.data.frame will only get slower as the number of columns increases)

> l2 <- replicate(26, sample(letters), simplify = FALSE)
> names(l2) <- letters
> microbenchmark::microbenchmark(
+   as_data_frame(l2),
+   as.data.frame(l2)
+ )
Unit: microseconds
              expr      min        lq    median       uq      max neval
 as_data_frame(l2)   48.952   62.2085   71.6335   88.539  170.927   100
 as.data.frame(l2) 3734.746 4062.8385 4378.8375 4979.390 8336.778   100

[lionel-]

looks awesome!

I have a question regarding coding style. When I started programming I followed Hadley's style but I had a hard time figuring out what to do with class predicates and coercion functions. I first named them is_class() and as_class() but then I checked the code of various rstudio and hadleyverse packages where the convention seems to be is.class() and as.class().

So, wouldn't dplyr's API be more consistent if that function was named as.data_frame()?

[lionel-]

I just posted an issue about this on r-pkgs

hadley/r-pkgs#150