Indexing (non)missing values using first() inside of summarise() no longer works.

[iron0012]

I used to be able to use the following code to get the first non-missing value within a group, but it no longer works. Now I get "Error: Unsupported vector type language."
Rolling back to 0.4.3 confirms that it was the update to 0.5 that caused this change in behavior.

ids <- c(1,1,2,2,3,3)
data_column <- c(NA,1,2,3,3,NA)
df <- data.frame(ids,data_column)

df %>% 
group_by(ids) %>% 
summarise(data_column_first = first(data_column[!is.na(data_column)]))

What I wanted the code above to return is something like:

ids	data_column_first
1	1
2	2
3	3

I don't know if this is a bug, a feature, a problem with the configuration of my environment, or if it's just a specific manifestation of some more general bug or feature, but it also occurs when trying to use last() in the same context instead of first().

Also, if you were to tell me a better alternative way to accomplish what I'm trying to do, I wouldn't complain! :D

[py-b]

Hello,

I have the same kind of problem.
This doesn't seem to be related to NAs in the data.

[alexsanjoseph]

👍 Have the same issue with the 0.5.0 update

[julou]

same problem here with other single-table verbs (mutate, filter) since upgrading to dplyr0.5
the current workaround is obviously to use head, tail and [[…
A fix would be really helpful though!

[krlmlr]

Thanks, confirmed. Would you like to contribute a testthat test?

[hadley]

Minimal reprex:

library(dplyr, warn.conflicts = FALSE)

tibble(x = 1:3) %>%  summarise(f = first(x[-1]))
#> Error in summarise_impl(.data, dots): Unsupported vector type language

[hadley]

@krlmlr I assume this is the same root cause as #946 - we should not use the hybrid evaluator here. How hard is it to add a branch for these cases?