[R] Add bindings for pmin() and pmax()

[asfimport]

ARROW-12751 adds a row-wise min/max kernel to the C++ library. Add R bindings so users can call pmin() and pmax() in dplyr verbs. Use ElementWiseAggregateOptions to ensure that NA values are handled the same as in base::pmin() and base::pmax().

Reporter: Ian Cook / @ianmcook
Assignee: Mauricio 'Pachá' Vargas Sepúlveda / @pachadotdev

Related issues:

• [C++] Add variadic row-wise min/max kernels (least/greatest) (depends upon)

PRs and other links:

• GitHub Pull Request #10615

_{Note: This issue was originally created as ARROW-12967. Please see the migration documentation for further details.}

[asfimport]

Mauricio 'Pachá' Vargas Sepúlveda / @pachadotdev:
PR sent !
this caught my attention, one should expect double instead of non-representable, but the R function works well

write_dataset(df, "df")

df2 <- open_dataset("df")

df2 %>% 
  mutate(
    max_val_1 = pmax(val1, val2, val3),
    max_val_2 = pmax(val1, val2, val3, na.rm = T),
    min_val_1 = pmin(val1, val2, val3),
    min_val_2 = pmin(val1, val2, val3, na.rm = T)
  ) 

FileSystemDataset (query)
city: string
val1: double
val2: double
val3: double
max_val_1: double (element_wise_max(val1, val2, val3, {NON-REPRESENTABLE OPTIONS}))
max_val_2: double (element_wise_max(val1, val2, val3, {NON-REPRESENTABLE OPTIONS}))
min_val_1: double (element_wise_min(val1, val2, val3, {NON-REPRESENTABLE OPTIONS}))
min_val_2: double (element_wise_min(val1, val2, val3, {NON-REPRESENTABLE OPTIONS}))

See $.data for the source Arrow object

[asfimport]

Ian Cook / @ianmcook:
Issue resolved by pull request 10615
#10615