R floats are silently converted into Rust ints

[clauswilke]

If a Rust function expecting an int is handed a float from R, then the float is silently converted to int, stripping any decimal values in the process. See e.g. here:

extendr/tests/extendrtests/tests/testthat/test-type-conversion.R

Line 10 in 9752c7f

expect_equal(.Call(wrap__int_scalar, 4.4), 4L) # is this deliberate? seems dangerous

I think this should instead raise an "unable to convert" error. Silent conversion from float to int is dangerous.

[Ilia-Kosenkov]

Just a summary of inconsistent conversion behavior. If i32 is expected, numeric is silently converted to i32.
On the other hand, when Option<i32> is expected, numeric become illegal and only pure integer (with literal L) are allowed, as well as NA.

rust_code <- "
#[extendr]
fn slice_test(v : &[i32], i : i32) -> i32 {
    v[i as usize]
}

#[extendr]
fn slice_test_na(v : &[i32], i : Option<i32>) -> Option<i32> {
    match i {
        Some (index) => Some(v[index as usize]),
        None => None
    }
}
"

rextendr::rust_source(code = rust_code)
#> build directory: C:\Users\[redacted]\AppData\Local\Temp\Rtmp2Z8D1v\filea087d24386d
# OK
slice_test(1:10, 1)
#> [1] 2
# Should not be OK
slice_test(1:10, 1.5)
#> [1] 2
# Not OK
slice_test(1:10, NA)
#> Error in slice_test(1:10, NA): unable to convert R object to primitive

# Should be OK but is not?
slice_test_na(1:10, 1)
#> Error in slice_test_na(1:10, 1): expected an integer scalar
# But this is
slice_test_na(1:10, 1L)
#> [1] 2
# This is correct
slice_test_na(1:10, 1.5)
#> Error in slice_test_na(1:10, 1.5): expected an integer scalar
# This is also correct
slice_test_na(1:10, NA)
#> [1] NA

^{Created on 2021-01-20 by the reprex package (v0.3.0)}

[clauswilke]

Removing the "good first issue" label as I think this is a complex and somewhat tricky topic. The problem is not the implementation but the correct conceptual design.