tinystr/

int_ops.rs

// This file is part of ICU4X. For terms of use, please see the file
// called LICENSE at the top level of the ICU4X source tree
// (online at: https://github.com/unicode-org/icu4x/blob/main/LICENSE ).

use crate::asciibyte::AsciiByte;

/// Internal helper struct that performs operations on aligned integers.
/// Supports strings up to 4 bytes long.
#[repr(transparent)]
pub struct Aligned4(u32);

impl Aligned4 {
    /// # Panics
    /// Panics if N is greater than 4
    #[inline]
    pub const fn from_ascii_bytes<const N: usize>(src: &[AsciiByte; N]) -> Self {
        let mut bytes = [0; 4];
        let mut i = 0;
        // The function documentation defines when panics may occur
        #[expect(clippy::indexing_slicing)]
        while i < N {
            bytes[i] = src[i] as u8;
            i += 1;
        }
        Self(u32::from_ne_bytes(bytes))
    }

    pub const fn len(&self) -> usize {
        let word = self.0;
        #[cfg(target_endian = "little")]
        let len = (4 - word.leading_zeros() / 8) as usize;
        #[cfg(target_endian = "big")]
        let len = (4 - word.trailing_zeros() / 8) as usize;
        len
    }

    pub const fn is_ascii_alphabetic(&self) -> bool {
        let word = self.0;
        // Each of the following bitmasks set *the high bit* (0x8) to 0 for valid and 1 for invalid.
        // `mask` sets all NUL bytes to 0.
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        // `lower` converts the string to lowercase. It may also change the value of non-alpha
        // characters, but this does not matter for the alphabetic test that follows.
        let lower = word | 0x2020_2020;
        // `alpha` sets all alphabetic bytes to 0. We only need check for lowercase characters.
        let alpha = !(lower + 0x1f1f_1f1f) | (lower + 0x0505_0505);
        // The overall string is valid if every character passes at least one test.
        // We performed two tests here: non-NUL (`mask`) and alphabetic (`alpha`).
        (alpha & mask) == 0
    }

    pub const fn is_ascii_alphanumeric(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        let numeric = !(word + 0x5050_5050) | (word + 0x4646_4646);
        let lower = word | 0x2020_2020;
        let alpha = !(lower + 0x1f1f_1f1f) | (lower + 0x0505_0505);
        (alpha & numeric & mask) == 0
    }

    pub const fn is_ascii_numeric(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        let numeric = !(word + 0x5050_5050) | (word + 0x4646_4646);
        (numeric & mask) == 0
    }

    pub const fn is_ascii_lowercase(&self) -> bool {
        let word = self.0;
        // For efficiency, this function tests for an invalid string rather than a valid string.
        // A string is ASCII lowercase iff it contains no uppercase ASCII characters.
        // `invalid_case` sets all uppercase ASCII characters to 0 and all others to 1.
        let invalid_case = !(word + 0x3f3f_3f3f) | (word + 0x2525_2525);
        // The string is valid if it contains no invalid characters (if all high bits are 1).
        (invalid_case & 0x8080_8080) == 0x8080_8080
    }

    pub const fn is_ascii_titlecase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_lowercase
        let invalid_case = if cfg!(target_endian = "little") {
            !(word + 0x3f3f_3f1f) | (word + 0x2525_2505)
        } else {
            !(word + 0x1f3f_3f3f) | (word + 0x0525_2525)
        };
        (invalid_case & 0x8080_8080) == 0x8080_8080
    }

    pub const fn is_ascii_uppercase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_lowercase
        let invalid_case = !(word + 0x1f1f_1f1f) | (word + 0x0505_0505);
        (invalid_case & 0x8080_8080) == 0x8080_8080
    }

    pub const fn is_ascii_alphabetic_lowercase(&self) -> bool {
        let word = self.0;
        // `mask` sets all NUL bytes to 0.
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        // `lower_alpha` sets all lowercase ASCII characters to 0 and all others to 1.
        let lower_alpha = !(word + 0x1f1f_1f1f) | (word + 0x0505_0505);
        // The overall string is valid if every character passes at least one test.
        // We performed two tests here: non-NUL (`mask`) and lowercase ASCII character (`alpha`).
        (lower_alpha & mask) == 0
    }

    pub const fn is_ascii_alphabetic_titlecase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic_lowercase
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        let title_case = if cfg!(target_endian = "little") {
            !(word + 0x1f1f_1f3f) | (word + 0x0505_0525)
        } else {
            !(word + 0x3f1f_1f1f) | (word + 0x2505_0505)
        };
        (title_case & mask) == 0
    }

    pub const fn is_ascii_alphabetic_uppercase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic_lowercase
        let mask = (word + 0x7f7f_7f7f) & 0x8080_8080;
        let upper_alpha = !(word + 0x3f3f_3f3f) | (word + 0x2525_2525);
        (upper_alpha & mask) == 0
    }

    pub const fn to_ascii_lowercase(&self) -> [AsciiByte; 4] {
        let word = self.0;
        let mask = ((word + 0x3f3f_3f3f) & !(word + 0x2525_2525) & 0x8080_8080) >> 2;
        let result = word | mask;
        // SAFETY: The existing word is ASCII, and it is bitwise-ORed with `mask`,
        // a subset of 0x20_20_20_... (0x80_... >> 2), which keeps the word
        // in the ASCII range (all bytes <= 127).
        unsafe { AsciiByte::to_ascii_byte_array(&result.to_ne_bytes()) }
    }

    pub const fn to_ascii_titlecase(&self) -> [AsciiByte; 4] {
        let word = self.0.to_le();
        let mask = ((word + 0x3f3f_3f1f) & !(word + 0x2525_2505) & 0x8080_8080) >> 2;
        let result = (word | mask) & !(0x20 & mask);
        // SAFETY: The existing word is ASCII, and it is bitwise-ORed with `mask`, a subset of 0x20_20_20_...
        // (0x80_... >> 2), which keeps the word in the ASCII range (all bytes <= 127). It is then ANDed
        // which cannot introduce new bits.
        unsafe { AsciiByte::to_ascii_byte_array(&u32::from_le(result).to_ne_bytes()) }
    }

    pub const fn to_ascii_uppercase(&self) -> [AsciiByte; 4] {
        let word = self.0;
        let mask = !(((word + 0x1f1f_1f1f) & !(word + 0x0505_0505) & 0x8080_8080) >> 2);
        let result = word & mask;
        // SAFETY: The existing word is ASCII, and it is bitwise-ANDed with `mask`,
        // which cannot introduce new bits, so it stays ASCII.
        unsafe { AsciiByte::to_ascii_byte_array(&result.to_ne_bytes()) }
    }
}

/// Internal helper struct that performs operations on aligned integers.
/// Supports strings up to 8 bytes long.
#[repr(transparent)]
pub struct Aligned8(u64);

impl Aligned8 {
    /// # Panics
    /// Panics if N is greater than 8
    #[inline]
    pub const fn from_ascii_bytes<const N: usize>(src: &[AsciiByte; N]) -> Self {
        let mut bytes = [0; 8];
        let mut i = 0;
        // The function documentation defines when panics may occur
        #[expect(clippy::indexing_slicing)]
        while i < N {
            bytes[i] = src[i] as u8;
            i += 1;
        }
        Self(u64::from_ne_bytes(bytes))
    }

    pub const fn len(&self) -> usize {
        let word = self.0;
        #[cfg(target_endian = "little")]
        let len = (8 - word.leading_zeros() / 8) as usize;
        #[cfg(target_endian = "big")]
        let len = (8 - word.trailing_zeros() / 8) as usize;
        len
    }

    pub const fn is_ascii_alphabetic(&self) -> bool {
        let word = self.0;
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        let lower = word | 0x2020_2020_2020_2020;
        let alpha = !(lower + 0x1f1f_1f1f_1f1f_1f1f) | (lower + 0x0505_0505_0505_0505);
        (alpha & mask) == 0
    }

    pub const fn is_ascii_alphanumeric(&self) -> bool {
        let word = self.0;
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        let numeric = !(word + 0x5050_5050_5050_5050) | (word + 0x4646_4646_4646_4646);
        let lower = word | 0x2020_2020_2020_2020;
        let alpha = !(lower + 0x1f1f_1f1f_1f1f_1f1f) | (lower + 0x0505_0505_0505_0505);
        (alpha & numeric & mask) == 0
    }

    pub const fn is_ascii_numeric(&self) -> bool {
        let word = self.0;
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        let numeric = !(word + 0x5050_5050_5050_5050) | (word + 0x4646_4646_4646_4646);
        (numeric & mask) == 0
    }

    pub const fn is_ascii_lowercase(&self) -> bool {
        let word = self.0;
        let invalid_case = !(word + 0x3f3f_3f3f_3f3f_3f3f) | (word + 0x2525_2525_2525_2525);
        (invalid_case & 0x8080_8080_8080_8080) == 0x8080_8080_8080_8080
    }

    pub const fn is_ascii_titlecase(&self) -> bool {
        let word = self.0;
        let invalid_case = if cfg!(target_endian = "little") {
            !(word + 0x3f3f_3f3f_3f3f_3f1f) | (word + 0x2525_2525_2525_2505)
        } else {
            !(word + 0x1f3f_3f3f_3f3f_3f3f) | (word + 0x0525_2525_2525_2525)
        };
        (invalid_case & 0x8080_8080_8080_8080) == 0x8080_8080_8080_8080
    }

    pub const fn is_ascii_uppercase(&self) -> bool {
        let word = self.0;
        let invalid_case = !(word + 0x1f1f_1f1f_1f1f_1f1f) | (word + 0x0505_0505_0505_0505);
        (invalid_case & 0x8080_8080_8080_8080) == 0x8080_8080_8080_8080
    }

    pub const fn is_ascii_alphabetic_lowercase(&self) -> bool {
        let word = self.0;
        // `mask` sets all NUL bytes to 0.
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        // `lower_alpha` sets all lowercase ASCII characters to 0 and all others to 1.
        let lower_alpha = !(word + 0x1f1f_1f1f_1f1f_1f1f) | (word + 0x0505_0505_0505_0505);
        // The overall string is valid if every character passes at least one test.
        // We performed two tests here: non-NUL (`mask`) and lowercase ASCII character (`alpha`).
        (lower_alpha & mask) == 0
    }

    pub const fn is_ascii_alphabetic_titlecase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic_lowercase
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        let title_case = if cfg!(target_endian = "little") {
            !(word + 0x1f1f_1f1f_1f1f_1f3f) | (word + 0x0505_0505_0505_0525)
        } else {
            !(word + 0x3f1f_1f1f_1f1f_1f1f) | (word + 0x2505_0505_0505_0505)
        };
        (title_case & mask) == 0
    }

    pub const fn is_ascii_alphabetic_uppercase(&self) -> bool {
        let word = self.0;
        // See explanatory comments in is_ascii_alphabetic_lowercase
        let mask = (word + 0x7f7f_7f7f_7f7f_7f7f) & 0x8080_8080_8080_8080;
        let upper_alpha = !(word + 0x3f3f_3f3f_3f3f_3f3f) | (word + 0x2525_2525_2525_2525);
        (upper_alpha & mask) == 0
    }

    pub const fn to_ascii_lowercase(&self) -> [AsciiByte; 8] {
        let word = self.0;
        let mask = ((word + 0x3f3f_3f3f_3f3f_3f3f)
            & !(word + 0x2525_2525_2525_2525)
            & 0x8080_8080_8080_8080)
            >> 2;
        let result = word | mask;
        // SAFETY: The existing word is ASCII, and it is bitwise-ORed with `mask`,
        // a subset of 0x20_20_20_... (0x80_... >> 2), which keeps the word
        // in the ASCII range (all bytes <= 127).
        unsafe { AsciiByte::to_ascii_byte_array(&result.to_ne_bytes()) }
    }

    pub const fn to_ascii_titlecase(&self) -> [AsciiByte; 8] {
        let word = self.0.to_le();
        let mask = ((word + 0x3f3f_3f3f_3f3f_3f1f)
            & !(word + 0x2525_2525_2525_2505)
            & 0x8080_8080_8080_8080)
            >> 2;
        let result = (word | mask) & !(0x20 & mask);
        // SAFETY: The existing word is ASCII, and it is bitwise-ORed with `mask`, a subset of 0x20_20_20_...
        // (0x80_... >> 2), which keeps the word in the ASCII range (all bytes <= 127). It is then ANDed
        // which cannot introduce new bits.
        unsafe { AsciiByte::to_ascii_byte_array(&u64::from_le(result).to_ne_bytes()) }
    }

    pub const fn to_ascii_uppercase(&self) -> [AsciiByte; 8] {
        let word = self.0;
        let mask = !(((word + 0x1f1f_1f1f_1f1f_1f1f)
            & !(word + 0x0505_0505_0505_0505)
            & 0x8080_8080_8080_8080)
            >> 2);
        let result = word & mask;
        // SAFETY: The existing word is ASCII, and it is bitwise-ANDed with `mask`,
        // which cannot introduce new bits, so it stays ASCII.
        unsafe { AsciiByte::to_ascii_byte_array(&result.to_ne_bytes()) }
    }
}