if_any() effectively reducing.

[romainfrancois]

closes #5709

simplified version of @iago-pssjd example:

library(dplyr, warn.conflicts = FALSE)
d <- data.frame(x = c(0, 0, 10), y = c(0, 0, 10), z = c(10, 0, 10))
d %>% mutate(
  test = case_when(
    if_all(x:z, ~ . > 8) ~ "all big",
    if_any(x:z, ~ . > 8) ~ "some big",
    TRUE                 ~ "small"
  ))
#>    x  y  z     test
#> 1  0  0 10 some big
#> 2  0  0  0    small
#> 3 10 10 10  all big

^{Created on 2021-01-28 by the reprex package (v0.3.0)}

[lionel-]

Nice use of if_any()!

[DavisVaughan]

Suggested change

	p_reduced[i] = p_reduced[i] == TRUE || p_df_j[i] == TRUE;
	p_reduced[i] = p_reduced[i] || p_df_j[i];

Can the comparison be removed?

[romainfrancois]

I think not because of NA_LOGICAL but I have to check

[romainfrancois]

if I make these two changes, I get:

library(dplyr, warn.conflicts = FALSE)

df <- expand.grid(
  x = c(TRUE, FALSE, NA), y = c(TRUE, FALSE, NA)
)
filter(df, x, y)
#>      x    y
#> 1 TRUE TRUE
filter(df, x & y)
#>      x    y
#> 1 TRUE TRUE
filter(df, if_all(c(x,y), identity))
#>      x    y
#> 1 TRUE TRUE
#> 2   NA TRUE
#> 3 TRUE   NA
#> 4   NA   NA

filter(df, x | y)
#>       x     y
#> 1  TRUE  TRUE
#> 2 FALSE  TRUE
#> 3    NA  TRUE
#> 4  TRUE FALSE
#> 5  TRUE    NA
filter(df, if_any(c(x,y), identity))
#>       x     y
#> 1  TRUE  TRUE
#> 2 FALSE  TRUE
#> 3    NA  TRUE
#> 4  TRUE FALSE
#> 5    NA FALSE
#> 6  TRUE    NA
#> 7 FALSE    NA
#> 8    NA    NA

^{Created on 2021-01-28 by the reprex package (v0.3.0)}

otherwise, I get:

library(dplyr, warn.conflicts = FALSE)

df <- expand.grid(
  x = c(TRUE, FALSE, NA), y = c(TRUE, FALSE, NA)
)
filter(df, x, y)
#>      x    y
#> 1 TRUE TRUE
filter(df, x & y)
#>      x    y
#> 1 TRUE TRUE
filter(df, if_all(c(x,y), identity))
#>      x    y
#> 1 TRUE TRUE

filter(df, x | y)
#>       x     y
#> 1  TRUE  TRUE
#> 2 FALSE  TRUE
#> 3    NA  TRUE
#> 4  TRUE FALSE
#> 5  TRUE    NA
filter(df, if_any(c(x,y), identity))
#>       x     y
#> 1  TRUE  TRUE
#> 2 FALSE  TRUE
#> 3    NA  TRUE
#> 4  TRUE FALSE
#> 5  TRUE    NA

^{Created on 2021-01-28 by the reprex package (v0.3.0)}