Function enable_ascii_fast_path

datafusion/functions/src/unicode/substr.rs:234–269

String characters are variable length encoded in UTF-8, `substr()` function's arguments are character-based, converting them into byte-based indices requires expensive decoding. However, checking if a string is ASCII-only is relatively cheap. If strings are ASCII only, use byte-based indices instead. A common pattern to call `substr()` is taking a small prefix of a long string, such as `substr(lo

(
    string_array: &V,
    start: &Int64Array,
    count: Option<&Int64Array>,
)

// In such case the overhead of ASCII-validation may not be worth it, so
// skip the validation for short prefix for now.
pub fn enable_ascii_fast_path<'a, V: StringArrayType<'a>>(
    string_array: &V,
    start: &Int64Array,
    count: Option<&Int64Array>,
) -> bool {
    let is_short_prefix = match count {
        Some(count) => {
            let short_prefix_threshold = 32.0;
            let n_sample = 10;

            // HACK: can be simplified if function has specialized
            // implementation for `ScalarValue` (implement without `make_scalar_function()`)
            let total_prefix_len = start
                .iter()
                .zip(count.iter())
                .take(n_sample)
                .map(|(start, count)| {
                    let start = start.unwrap_or(0);
                    let count = count.unwrap_or(0);
                    // To get substring, need to decode from 0 to start+count instead of start to start+count
                    start.saturating_add(count)
                })
                .fold(0i64, |acc, val| acc.saturating_add(val));

            (total_prefix_len as f64 / n_sample as f64) <= short_prefix_threshold
        }
        None => false,
    };

    if is_short_prefix {
        // Skip ASCII validation for short prefix
        false
    } else {
        string_array.is_ascii()
    }
}

fn string_view_substr(
    string_view_array: &StringViewArray,