Breaking the **R**ules: mutable references

[JosiahParry]

This could be an advanced section with hazards all over it that discusses how to mutate R objects in place.

This is against the clone on write (Cow) rules of R. BUT if you created and own all of the objects, it may make sense to mutably handle them. in some cases.

[JosiahParry]

For example this will let you modify the first element of an integer vector in place

#[extendr]
fn changeme(x: Integers) {
    let mut x = x;
    x[0] = 100.into();
}

[JosiahParry]

#[extendr]
fn swap_list_element(x: List, y: Robj, idx: i32) {
    let mut x = x;
    let _ = x.set_elt(idx as usize, y);
}

(o <- list(x = 1, y = 2, z = "a"))
#> $x
#> [1] 1
#> 
#> $y
#> [1] 2
#> 
#> $z
#> [1] "a"

# get address
addr_1 <- tracemem(o)

# swap in place 
swap_list_element(o, letters, 0)

# print 
o
#> $x
#>  [1] "a" "b" "c" "d" "e" "f" "g" "h" "i" "j" "k" "l" "m" "n" "o" "p" "q" "r" "s"
#> [20] "t" "u" "v" "w" "x" "y" "z"
#> 
#> $y
#> [1] 2
#> 
#> $z
#> [1] "a"

# get address again
addr_2 <- tracemem(o)

# check that the addresses are the same 
identical(addr_1, addr_2)
#> [1] TRUE

^{Created on 2024-03-02 with reprex v2.0.2}

[JosiahParry]

This could be a cool place to recreate rlang::vec_poke_n() using Rust