[R] `mutate(x2=ifelse(x=='',NA,x))` Error: Function 'if_else' has no kernel matching input types

[asfimport]

 

q <- data.table(x=c('','30111976','01011976'))
q %>% write_dataset('q')
q2 <- 'q' %>% open_dataset %>% mutate(x2=ifelse(x=='',NA,x)) %>% collect
Error in `collect()`:
! NotImplemented: Function 'if_else' has no kernel matching input types (bool, bool, string)
Run `rlang::last_error()` to see where the error occurred.

In Functions available in Arrow dplyr queries it is stated that ifelse() is available, but the error above suggests otherwise.

Update, "str_replace" is another function that is supposedly available in 10.0.0 but is not (or does not seem to be):

q2 <- 'q' %>% open_dataset %>% mutate(x2=x %>% str_replace('',NA)) %>% collect
Error: Expression x %>% str_replace("", NA) not supported in Arrow
Call collect() first to pull data into R.

 

Reporter: Lucas Mation / @lucasmation
Assignee: Neal Richardson / @nealrichardson

Related issues:

• [R] Cast scalars to type of field in Expression building (is fixed by)

_{Note: This issue was originally created as ARROW-18244. Please see the migration documentation for further details.}

[asfimport]

Neal Richardson / @nealrichardson:
I commented about ifelse on ARROW-18241 with a workaround for the current version and a suggestion that it should work as you expect in the development version. If you can install a nightly version of the package and confirm, that would be great.

str_replace does work if the replacement is a string, but NA is logical, and this fails in regular stringr too:

> data.frame(x="a") %>% mutate(y = str_replace(x, "a", NA))
Error in `mutate()`:
! Problem while computing `y = str_replace(x, "a", NA)`.
Caused by error:
! `replacement` must be a character vector
Run `rlang::last_error()` to see where the error occurred.

Use NA_character_ to specify a string.

[asfimport]

Lucas Mation / @lucasmation:
@nealrichardson OK. Then, I guess we should close this issue, since ifelse if fixed in the dev version. 

But on str_replace (and maybe this should be a ticket of it's own) it does not replicate R's behaviuour in this case

q <- data.table(x=c('','1','2'))
q %>% write_dataset('q')

#in R

q %>% mutate(x2=x %>% str_replace('^
s\*$',NA_character_))

   x   x2
1:   <NA>
2: 1    1
3: 2    2

#in arrow

q2 <- 'q' %>% open_dataset %>% mutate(x2=x %>% str_replace('^
s\*$',NA_character_)) %>% collect

q2

   x x2
1:     
2: 1  1
3: 2  2

[asfimport]

Neal Richardson / @nealrichardson:
I'm not sure that's a valid use of string replacement since that's replacing substrings into the elements of the character vector, not necessarily replacing the entire element (ifelse is what you want for that), but we can detect and reject that case. Could you please open a new issue for that? Thanks!

[asfimport]

Lucas Mation / @lucasmation:
Sure, created on ticket [#18250]