Using if_else() results in error 'could not find function "fifelse"'

[larspijnappel]

I just installed dtplyr version 1.0.0 and it works like a charm. Unfortunately, when using the dplyr::if_else() function, it fails with the error message could not find function "fifelse".

library(dtplyr)
library(dplyr, warn.conflicts = FALSE)

mtcars %>% 
    lazy_dt() %>%
    mutate(gear2 = if_else(gear == 5, 10, gear)) %>%
    as_tibble()

[hadley]

Try installing the latest version of data.table and see if works then?

[larspijnappel]

Although my installed data.table package was already the latest version 1.12.6, I reinstalled it.
I also upgraded DT from v0.9 to v10.0.

But, after restarting, the issue remains..

[hadley]

It definitely works for me:

library(dtplyr)
library(dplyr, warn.conflicts = FALSE)

mtcars %>% 
  lazy_dt() %>%
  select(gear) %>% 
  mutate(gear2 = if_else(gear == 4, 10, gear)) %>%
  as_tibble()
#> # A tibble: 32 x 2
#>     gear gear2
#>    <dbl> <dbl>
#>  1     4    10
#>  2     4    10
#>  3     4    10
#>  4     3     3
#>  5     3     3
#>  6     3     3
#>  7     3     3
#>  8     4    10
#>  9     4    10
#> 10     4    10
#> # … with 22 more rows

I'm not sure what's going on. Any ideas @MichaelChirico?

[larspijnappel]

As pointed out by batpigandme in a related issue (#123 (comment)) the solution is to add library(data.table).

After re-reading your blog I noticed I missed this advice:

To use dtplyr, you must at least load dtplyr and dplyr. You might also want to load data.table so you can access the other goodies that it provides.

Sorry for overlooking this and thx for your help!

[MichaelChirico]

would it make sense to detect data.table isn't loaded & include it in the lazy execution message?

[hadley]

Or possibly inject the data table functions we use directly into the evaluation environment.

[MichaelChirico]

@hadley also might want to bump the data.table version dependency? fifelse/fcoalesce are from 1.12.4. Or otherwise branch on the package version to use fifelse only depending on the installed data.table version.

[MichaelChirico]

For the injection part, I guess that's what's happening already for other data.table functions, is there any reason we didn't do this injection part initially? I guess we have to manually populate that environment with functions?

[hadley]

Yeah, I think I just forgot about that