Implement if_else

[thisisnic]

Fixes #156

There is an issue with return type in DuckDB with logical values, but it won't affect the TPC-H query calculations, so I was going to leave that as something to come back to.

library(substrait)
library(dplyr)

# using Arrow
mtcars %>%
  arrow_substrait_compiler() %>%
  mutate(am_chr = if_else(am == 0, "automatic", "manual")) %>%
  select(am, am_chr) %>%
  collect()
#> # A tibble: 32 × 2
#>       am am_chr   
#>    <dbl> <chr>    
#>  1     1 manual   
#>  2     1 manual   
#>  3     1 manual   
#>  4     0 automatic
#>  5     0 automatic
#>  6     0 automatic
#>  7     0 automatic
#>  8     0 automatic
#>  9     0 automatic
#> 10     0 automatic
#> # … with 22 more rows

# using DuckDB
mtcars %>%
  duckdb_substrait_compiler() %>%
  mutate(am_chr = if_else(am == 0, "automatic", "manual")) %>%
  select(am, am_chr) %>%
  collect()
#> # A tibble: 32 × 2
#>       am am_chr   
#>    <dbl> <chr>    
#>  1     1 manual   
#>  2     1 manual   
#>  3     1 manual   
#>  4     0 automatic
#>  5     0 automatic
#>  6     0 automatic
#>  7     0 automatic
#>  8     0 automatic
#>  9     0 automatic
#> 10     0 automatic
#> # … with 22 more rows

## issue with TRUE/FALSE in DuckDB
# using Arrow
mtcars %>%
  arrow_substrait_compiler() %>%
  mutate(automatic = if_else(am == 0, TRUE, FALSE)) %>%
  select(am, automatic) %>%
  collect()
#> # A tibble: 32 × 2
#>       am automatic
#>    <dbl> <lgl>    
#>  1     1 FALSE    
#>  2     1 FALSE    
#>  3     1 FALSE    
#>  4     0 TRUE     
#>  5     0 TRUE     
#>  6     0 TRUE     
#>  7     0 TRUE     
#>  8     0 TRUE     
#>  9     0 TRUE     
#> 10     0 TRUE     
#> # … with 22 more rows

# using DuckDB
mtcars %>%
  duckdb_substrait_compiler() %>%
  mutate(automatic = if_else(am == 0, TRUE, FALSE)) %>%
  select(am, automatic) %>%
  collect()
#> # A tibble: 32 × 2
#>       am automatic
#>    <dbl>     <int>
#>  1     1         0
#>  2     1         0
#>  3     1         0
#>  4     0         1
#>  5     0         1
#>  6     0         1
#>  7     0         1
#>  8     0         1
#>  9     0         1
#> 10     0         1
#> # … with 22 more rows

[paleolimbot]

When writing new tests, it would be nice to move away from example_data (so that when I'm reviewing these tests I don't have to remember what example_data to know if the expected answer is correct).

[thisisnic]

I quite like having the same dataset to reuse across multiple tests. I guess it could make sense to leave it exclusively for the compare_dplyr_bindings() functions, where you don't need to know what the dataset looks like?

[paleolimbot]

I think that's a good rule of thumb...those tests should pass no matter what example_data is (provided that it has a reasonable type coverage) so it doesn't help much when reading the tests to recreate a dataset every time.

[paleolimbot]

After your previous PR, you can put this in substrait_funs since it's the same for arrow and duckdb, correct?

[thisisnic]

Indeed

[paleolimbot]

Looks great!