Quelle  gb18030.rs

// Copyright Mozilla Foundation. See the COPYRIGHT
// file at the top-level directory of this distribution.
//
// Licensed under the Apache License, Version 2.0 <LICENSE-APACHE or
// https://www.apache.org/licenses/LICENSE-2.0> or the MIT license
// <LICENSE-MIT or https://opensource.org/licenses/MIT>, at your
// option. This file may not be copied, modified, or distributed
// except according to those terms.

use super::*;
use crate::data::*;
use crate::gb18030_2022::*;
use crate::handles::*;
use crate::variant::*;
// Rust 1.14.0 requires the following despite the asterisk above.
use super::in_inclusive_range16;
use super::in_range16;

enum Gb18030Pending {
    None,
    One(u8),
    Two(u8, u8),
    Three(u8, u8, u8),
}

impl Gb18030Pending {
    fn is_none(&self) -> bool {
        match *self {
            Gb18030Pending::None => true,
            _ => false,
        }
    }

    fn count(&self) -> usize {
        match *self {
            Gb18030Pending::None => 0,
            Gb18030Pending::One(_) => 1,
            Gb18030Pending::Two(_, _) => 2,
            Gb18030Pending::Three(_, _, _) => 3,
        }
    }
}

pub struct Gb18030Decoder {
    first: Option<u8>,
    second: Option<u8>,
    third: Option<u8>,
    pending: Gb18030Pending,
    pending_ascii: Option<u8>,
}

impl Gb18030Decoder {
    pub fn new() -> VariantDecoder {
        VariantDecoder::Gb18030(Gb18030Decoder {
            first: None,
            second: None,
            third: None,
            pending: Gb18030Pending::None,
            pending_ascii: None,
        })
    }

    pub fn in_neutral_state(&self) -> bool {
        self.first.is_none()
            && self.second.is_none()
            && self.third.is_none()
            && self.pending.is_none()
            && self.pending_ascii.is_none()
    }

    fn extra_from_state(&self, byte_length: usize) -> Option<usize> {
        byte_length.checked_add(
            self.pending.count()
                + match self.first {
                    None => 0,
                    Some(_) => 1,
                }
                + match self.second {
                    None => 0,
                    Some(_) => 1,
                }
                + match self.third {
                    None => 0,
                    Some(_) => 1,
                }
                + match self.pending_ascii {
                    None => 0,
                    Some(_) => 1,
                },
        )
    }

    pub fn max_utf16_buffer_length(&self, byte_length: usize) -> Option<usize> {
        // ASCII: 1 to 1 (worst case)
        // gbk: 2 to 1
        // ranges: 4 to 1 or 4 to 2
        checked_add(1, self.extra_from_state(byte_length))
    }

    pub fn max_utf8_buffer_length_without_replacement(&self, byte_length: usize) -> Option<usize> {
        // ASCII: 1 to 1
        // gbk: 2 to 2 or 2 to 3
        // ranges: 4 to 2, 4 to 3 or 4 to 4
        // 0x80: 1 to 3 (worst case)
        self.max_utf8_buffer_length(byte_length)
    }

    pub fn max_utf8_buffer_length(&self, byte_length: usize) -> Option<usize> {
        checked_add(1, checked_mul(3, self.extra_from_state(byte_length)))
    }

    gb18030_decoder_functions!(
        {
            // If first is between 0x81 and 0xFE, inclusive,
            // subtract offset 0x81.
            let non_ascii_minus_offset = non_ascii.wrapping_sub(0x81);
            if non_ascii_minus_offset > (0xFE - 0x81) {
                if non_ascii == 0x80 {
                    handle.write_upper_bmp(0x20ACu16);
                    continue 'outermost;
                }
                return (DecoderResult::Malformed(1, 0),
                        source.consumed(),
                        handle.written());
            }
            non_ascii_minus_offset
        },
        {
            // Two-byte (or error)
            if first_minus_offset >= 0x20 {
                // Not the gbk ideograph range above GB2312
                let trail_minus_offset = second.wrapping_sub(0xA1);
                if trail_minus_offset <= (0xFE - 0xA1) {
                    // GB2312
                    let hanzi_lead = first_minus_offset.wrapping_sub(0x2F);
                    if hanzi_lead < (0x77 - 0x2F) {
                        // Level 1 Hanzi, Level 2 Hanzi
                        // or one of the 5 PUA code
                        // points in between.
                        let hanzi_pointer = mul_94(hanzi_lead) + trail_minus_offset as usize;
                        let upper_bmp = GB2312_HANZI[hanzi_pointer];
                        handle.write_upper_bmp(upper_bmp)
                    } else if first_minus_offset == 0x20 {
                        // Symbols (starting with ideographic space)
                        let bmp = GB2312_SYMBOLS[trail_minus_offset as usize];
                        handle.write_bmp_excl_ascii(bmp)
                    } else if first_minus_offset == 0x25 && ((trail_minus_offset.wrapping_sub(63) as usize) < GB2312_SYMBOLS_AFTER_GREEK.len()) {
                        handle.write_bmp_excl_ascii(GB2312_SYMBOLS_AFTER_GREEK[trail_minus_offset.wrapping_sub(63) as usize])
                    } else if first_minus_offset == 0x27 && (trail_minus_offset as usize) < GB2312_PINYIN.len() {
                        handle.write_bmp_excl_ascii(GB2312_PINYIN[trail_minus_offset as usize])
                    } else if first_minus_offset > 0x76 {
                        // Bottom PUA
                        let pua = (0xE234 + mul_94(first_minus_offset - 0x77) + trail_minus_offset as usize) as u16;
                        handle.write_upper_bmp(pua)
                    } else {
                        let bmp = gb2312_other_decode((mul_94(first_minus_offset - 0x21) + (trail_minus_offset as usize)) as u16);
                        handle.write_bmp_excl_ascii(bmp)
                    }
                } else {
                    // gbk range on the left
                    let mut trail_minus_offset = second.wrapping_sub(0x40);
                    if trail_minus_offset > (0x7E - 0x40) {
                        let trail_minus_range_start = second.wrapping_sub(0x80);
                        if trail_minus_range_start > (0xA0 - 0x80) {
                            if second < 0x80 {
                                return (DecoderResult::Malformed(1, 0),
                                        unread_handle_second.unread(),
                                        handle.written());
                            }
                            return (DecoderResult::Malformed(2, 0),
                                    unread_handle_second.consumed(),
                                    handle.written());
                        }
                        trail_minus_offset = second - 0x41;
                    }
                    // Zero-base lead
                    let left_lead = first_minus_offset - 0x20;
                    let left_pointer = left_lead as usize * (190 - 94) +
                                       trail_minus_offset as usize;
                    let gbk_left_ideograph_pointer = left_pointer.wrapping_sub((0x29 - 0x20) * (190 - 94));
                    if gbk_left_ideograph_pointer < (((0x7D - 0x29) * (190 - 94)) - 5) {
                        let upper_bmp = gbk_left_ideograph_decode(gbk_left_ideograph_pointer as u16);
                        handle.write_upper_bmp(upper_bmp)
                    } else if left_pointer < ((0x29 - 0x20) * (190 - 94)) {
                        let bmp = gbk_other_decode(left_pointer as u16);
                        handle.write_bmp_excl_ascii(bmp)
                    } else {
                        let bottom_pointer = left_pointer - (((0x7D - 0x20) * (190 - 94)) - 5);
                        let upper_bmp = GBK_BOTTOM[bottom_pointer];
                        handle.write_upper_bmp(upper_bmp)
                    }
                }
            } else {
                // gbk ideograph range above GB2312
                let mut trail_minus_offset = second.wrapping_sub(0x40);
                if trail_minus_offset > (0x7E - 0x40) {
                    let trail_minus_range_start = second.wrapping_sub(0x80);
                    if trail_minus_range_start > (0xFE - 0x80) {
                        if second < 0x80 {
                            return (DecoderResult::Malformed(1, 0),
                                    unread_handle_second.unread(),
                                    handle.written());
                        }
                        return (DecoderResult::Malformed(2, 0),
                                unread_handle_second.consumed(),
                                handle.written());
                    }
                    trail_minus_offset = second - 0x41;
                }
                let pointer = first_minus_offset as usize * 190usize +
                              trail_minus_offset as usize;
                let upper_bmp = gbk_top_ideograph_decode(pointer as u16);
                handle.write_upper_bmp(upper_bmp)
            }
        },
        {
            // If third is between 0x81 and 0xFE, inclusive,
            // subtract offset 0x81.
            let third_minus_offset = third.wrapping_sub(0x81);
            if third_minus_offset > (0xFE - 0x81) {
                // We have an error. Let's inline what's going
                // to happen when `second` is
                // reprocessed. (`third` gets unread.)
                // `second` is guaranteed ASCII, so let's
                // put it in `pending_ascii`. Recompute
                // `second` from `second_minus_offset`.
                self.pending_ascii = Some(second_minus_offset + 0x30);
                // Now unread `third` and designate the previous
                // `first` as being in error.
                return (DecoderResult::Malformed(1, 1),
                        unread_handle_third.unread(),
                        handle.written());
            }
            third_minus_offset
        },
        {
            // If fourth is between 0x30 and 0x39, inclusive,
            // subtract offset 0x30.
            //
            // If we have an error, we'll inline what's going
            // to happen when `second` and `third` are
            // reprocessed. (`fourth` gets unread.)
            // `second` is guaranteed ASCII, so let's
            // put it in `pending_ascii`. Recompute
            // `second` from `second_minus_offset` to
            // make this block reusable when `second`
            // is not in scope.
            //
            // `third` is guaranteed to be in the range
            // that makes it become the new `self.first`.
            //
            // `fourth` gets unread and the previous
            // `first` gets designates as being in error.
            let fourth_minus_offset = fourth.wrapping_sub(0x30);
            if fourth_minus_offset > (0x39 - 0x30) {
                self.pending_ascii = Some(second_minus_offset + 0x30);
                self.pending = Gb18030Pending::One(third_minus_offset);
                return (DecoderResult::Malformed(1, 2),
                        unread_handle_fourth.unread(),
                        handle.written());
            }
            let pointer = (first_minus_offset as usize * (10 * 126 * 10)) +
                          (second_minus_offset as usize * (10 * 126)) +
                          (third_minus_offset as usize * 10) +
                          fourth_minus_offset as usize;
            if pointer <= 39419 {
                // BMP
                if pointer == 7457 {
                    handle.write_upper_bmp(0xE7C7)
                } else {
                    handle.write_bmp_excl_ascii(gb18030_range_decode(pointer as u16))
                }
            } else if pointer >= 189_000 && pointer <= 1_237_575 {
                // Astral
                handle.write_astral((pointer - (189_000usize - 0x1_0000usize)) as u32)
            } else {
                return (DecoderResult::Malformed(4, 0),
                        unread_handle_fourth.consumed(),
                        handle.written());
            }
        },
        self,
        non_ascii,
        first_minus_offset,
        second,
        second_minus_offset,
        unread_handle_second,
        third,
        third_minus_offset,
        unread_handle_third,
        fourth,
        fourth_minus_offset,
        unread_handle_fourth,
        source,
        handle,
        'outermost);
}

// XXX Experiment with inline directives
fn gbk_encode_non_unified(bmp: u16) -> Option<(usize, usize)> {
    // Try ideographic punctuation first as it's the most likely case.
    // Throwing in the check for full-width currencies and tilde is probably
    // more size-efficient here than elsewhere.
    if in_inclusive_range16(bmp, 0x2014, 0x3017) || in_inclusive_range16(bmp, 0xFF04, 0xFFE1) {
        if let Some(pos) = position(&GB2312_SYMBOLS[..], bmp) {
            return Some((0xA1, pos + 0xA1));
        }
    }
    // Ext A
    if in_range16(bmp, 0x3400, 0x4E00) {
        return position(&GBK_BOTTOM[21..100], bmp).map(|pos| {
            (
                0xFE,
                pos + if pos < (0x3F - 16) {
                    0x40 + 16
                } else {
                    0x41 + 16
                },
            )
        });
    }
    // Compatibility ideographs
    if in_range16(bmp, 0xF900, 0xFB00) {
        return position(&GBK_BOTTOM[0..21], bmp).map(|pos| {
            if pos < 5 {
                // end of second to last row
                (0xFD, pos + (190 - 94 - 5 + 0x41))
            } else {
                // last row
                (0xFE, pos + (0x40 - 5))
            }
        });
    }
    // Handle everything below U+02CA, which is in GBK_OTHER.
    if bmp < 0x02CA {
        if in_range16(bmp, 0x00E0, 0x0262) && bmp != 0x00F7 {
            // Pinyin except U+1E3F
            if let Some(pos) = position(&GB2312_PINYIN[..], bmp) {
                return Some((0xA8, pos + 0xA1));
            }
        } else if in_inclusive_range16(bmp, 0x00A4, 0x00F7)
            || in_inclusive_range16(bmp, 0x02C7, 0x02C9)
        {
            // Diacritics and Latin 1 symbols
            if let Some(pos) = position(&GB2312_SYMBOLS[3..(0xAC - 0x60)], bmp) {
                return Some((0xA1, pos + 0xA1 + 3));
            }
        }
        return None;
    }

    if in_inclusive_range16(bmp, 0xE78D, 0xE864) {
        // The array is sorted but short, so let's do linear search.
        if let Some(pos) = position(&GB18030_2022_OVERRIDE_PUA[..], bmp) {
            let pair = &GB18030_2022_OVERRIDE_BYTES[pos];
            return Some((pair[0].into(), pair[1].into()));
        }
    } else if bmp >= 0xFE17 {
        // Various brackets, all in full-width regions
        if let Some(pos) = position(&GB2312_SYMBOLS_AFTER_GREEK[..], bmp) {
            return Some((0xA6, pos + (0x9F - 0x60 + 0xA1)));
        }
    } else if bmp == 0x1E3F {
        // The one Pinyin placed elsewhere on the BMP
        return Some((0xA8, 0x7B - 0x60 + 0xA1));
    } else if in_range16(bmp, 0xA000, 0xD800) {
        // Since Korean has usage in China, let's spend a branch to fast-track
        // Hangul.
        return None;
    }
    // GB2312 other (except bottom PUA and PUA between Hanzi levels).
    if let Some(other_pointer) = gb2312_other_encode(bmp) {
        let other_lead = other_pointer as usize / 94;
        let other_trail = other_pointer as usize % 94;
        return Some((0xA2 + other_lead, 0xA1 + other_trail));
    }
    // At this point, we've handled all mappable characters above U+02D9 but
    // below U+2010. Let's check for that range in order to let lower BMP
    // characters used for minority languages in China avoid the subsequent
    // search that deals mainly with various symbols.
    if in_range16(bmp, 0x02DA, 0x2010) {
        return None;
    }
    // GBK other (except radicals and PUA in GBK_BOTTOM).
    if let Some(other_pointer) = gbk_other_encode(bmp) {
        let other_lead = other_pointer as usize / (190 - 94);
        let other_trail = other_pointer as usize % (190 - 94);
        let offset = if other_trail < 0x3F { 0x40 } else { 0x41 };
        return Some((other_lead + (0x81 + 0x20), other_trail + offset));
    }
    // CJK Radicals Supplement, PUA, and U+9FBx ideographs in GBK_BOTTOM
    if in_inclusive_range16(bmp, 0x2E81, 0x2ECA)
        || in_inclusive_range16(bmp, 0x9FB4, 0x9FBB)
        || in_inclusive_range16(bmp, 0xE816, 0xE855)
    {
        if let Some(pos) = position(&GBK_BOTTOM[21..], bmp) {
            let trail = pos + 16;
            let offset = if trail < 0x3F { 0x40 } else { 0x41 };
            return Some((0xFE, trail + offset));
        }
    }
    // GB2312 bottom PUA
    let bmp_minus_gb2312_bottom_pua = bmp.wrapping_sub(0xE234);
    if bmp_minus_gb2312_bottom_pua <= (0xE4C5 - 0xE234) {
        let pua_lead = bmp_minus_gb2312_bottom_pua as usize / 94;
        let pua_trail = bmp_minus_gb2312_bottom_pua as usize % 94;
        return Some((0x81 + 0x77 + pua_lead, 0xA1 + pua_trail));
    }
    // PUA between Hanzi Levels
    let bmp_minus_pua_between_hanzi = bmp.wrapping_sub(0xE810);
    if bmp_minus_pua_between_hanzi < 5 {
        return Some((0x81 + 0x56, 0xFF - 5 + bmp_minus_pua_between_hanzi as usize));
    }
    None
}

#[cfg(not(feature = "fast-gb-hanzi-encode"))]
#[inline(always)]
fn encode_hanzi(bmp: u16, _: u16) -> (u8, u8) {
    if let Some((lead, trail)) = gb2312_level1_hanzi_encode(bmp) {
        (lead, trail)
    } else if let Some(hanzi_pointer) = gb2312_level2_hanzi_encode(bmp) {
        let hanzi_lead = (hanzi_pointer / 94) + (0xD8);
        let hanzi_trail = (hanzi_pointer % 94) + 0xA1;
        (hanzi_lead as u8, hanzi_trail as u8)
    } else {
        let (lead, gbk_trail) = if bmp < 0x72DC {
            // Above GB2312
            let pointer = gbk_top_ideograph_encode(bmp) as usize;
            let lead = (pointer / 190) + 0x81;
            let gbk_trail = pointer % 190;
            (lead, gbk_trail)
        } else {
            // To the left of GB2312
            let gbk_left_ideograph_pointer = gbk_left_ideograph_encode(bmp) as usize;
            let lead = (gbk_left_ideograph_pointer / (190 - 94)) + (0x81 + 0x29);
            let gbk_trail = gbk_left_ideograph_pointer % (190 - 94);
            (lead, gbk_trail)
        };
        let offset = if gbk_trail < 0x3F { 0x40 } else { 0x41 };
        (lead as u8, (gbk_trail + offset) as u8)
    }
}

#[cfg(feature = "fast-gb-hanzi-encode")]
#[inline(always)]
fn encode_hanzi(_: u16, bmp_minus_unified_start: u16) -> (u8, u8) {
    gbk_hanzi_encode(bmp_minus_unified_start)
}

pub struct Gb18030Encoder {
    extended: bool,
}

impl Gb18030Encoder {
    pub fn new(encoding: &'static Encoding, extended_range: bool) -> Encoder {
        Encoder::new(
            encoding,
            VariantEncoder::Gb18030(Gb18030Encoder {
                extended: extended_range,
            }),
        )
    }

    pub fn max_buffer_length_from_utf16_without_replacement(
        &self,
        u16_length: usize,
    ) -> Option<usize> {
        if self.extended {
            u16_length.checked_mul(4)
        } else {
            // Need to add, because space check is done with the four-byte
            // assumption.
            checked_add(2, u16_length.checked_mul(2))
        }
    }

    pub fn max_buffer_length_from_utf8_without_replacement(
        &self,
        byte_length: usize,
    ) -> Option<usize> {
        if self.extended {
            // 1 to 1
            // 2 to 2
            // 3 to 2
            // 2 to 4 (worst)
            // 3 to 4
            // 4 to 4
            checked_add(2, byte_length.checked_mul(2))
        } else {
            // 1 to 1
            // 2 to 2
            // 3 to 2
            // Need to add, because space check is done with the four-byte
            // assumption.
            byte_length.checked_add(3)
        }
    }

    ascii_compatible_encoder_functions!(
        {
            let bmp_minus_unified_start = bmp.wrapping_sub(0x4E00);
            if bmp_minus_unified_start < (0x9FA6 - 0x4E00) {
                // CJK Unified Ideographs
                // Can't fail now, since all are
                // mapped.
                let (lead, trail) = encode_hanzi(bmp, bmp_minus_unified_start);
                handle.write_two(lead, trail)
            } else if bmp == 0xE5E5 {
                // It's not optimal to check for the unmappable
                // and for euro at this stage, but getting
                // the out of the way makes the rest of the
                // code less messy.
                return (
                    EncoderResult::unmappable_from_bmp(bmp),
                    source.consumed(),
                    handle.written(),
                );
            } else if bmp == 0x20AC && !self.extended {
                handle.write_one(0x80u8)
            } else {
                match gbk_encode_non_unified(bmp) {
                    Some((lead, trail)) => handle.write_two(lead as u8, trail as u8),
                    None => {
                        if !self.extended {
                            return (
                                EncoderResult::unmappable_from_bmp(bmp),
                                source.consumed(),
                                handle.written(),
                            );
                        }
                        let range_pointer = gb18030_range_encode(bmp);
                        let first = range_pointer / (10 * 126 * 10);
                        let rem_first = range_pointer % (10 * 126 * 10);
                        let second = rem_first / (10 * 126);
                        let rem_second = rem_first % (10 * 126);
                        let third = rem_second / 10;
                        let fourth = rem_second % 10;
                        handle.write_four(
                            (first + 0x81) as u8,
                            (second + 0x30) as u8,
                            (third + 0x81) as u8,
                            (fourth + 0x30) as u8,
                        )
                    }
                }
            }
        },
        {
            if !self.extended {
                return (
                    EncoderResult::Unmappable(astral),
                    source.consumed(),
                    handle.written(),
                );
            }
            let range_pointer = astral as usize + (189_000usize - 0x1_0000usize);
            let first = range_pointer / (10 * 126 * 10);
            let rem_first = range_pointer % (10 * 126 * 10);
            let second = rem_first / (10 * 126);
            let rem_second = rem_first % (10 * 126);
            let third = rem_second / 10;
            let fourth = rem_second % 10;
            handle.write_four(
                (first + 0x81) as u8,
                (second + 0x30) as u8,
                (third + 0x81) as u8,
                (fourth + 0x30) as u8,
            )
        },
        bmp,
        astral,
        self,
        source,
        handle,
        copy_ascii_to_check_space_four,
        check_space_four,
        false
    );
}

// Any copyright to the test code below this comment is dedicated to the
// Public Domain. http://creativecommons.org/publicdomain/zero/1.0/

#[cfg(all(test, feature = "alloc"))]
mod tests {
    use super::super::testing::*;
    use super::super::*;

    fn decode_gb18030(bytes: &[u8], expect: &str) {
        decode(GB18030, bytes, expect);
    }

    fn encode_gb18030(string: &str, expect: &[u8]) {
        encode(GB18030, string, expect);
    }

    fn encode_gbk(string: &str, expect: &[u8]) {
        encode(GBK, string, expect);
    }

    #[test]
    fn test_gb18030_decode() {
        // Empty
        decode_gb18030(b"", &"");

        // ASCII
        decode_gb18030(b"\x61\x62", "\u{0061}\u{0062}");

        // euro
        decode_gb18030(b"\x80", "\u{20AC}");
        decode_gb18030(b"\xA2\xE3", "\u{20AC}");

        // two bytes
        decode_gb18030(b"\x81\x40", "\u{4E02}");
        decode_gb18030(b"\x81\x7E", "\u{4E8A}");
        decode_gb18030(b"\x81\x7F", "\u{FFFD}\u{007F}");
        decode_gb18030(b"\x81\x80", "\u{4E90}");
        decode_gb18030(b"\x81\xFE", "\u{4FA2}");
        decode_gb18030(b"\xFE\x40", "\u{FA0C}");
        decode_gb18030(b"\xFE\x7F", "\u{FFFD}\u{007F}");
        decode_gb18030(b"\xFE\x80", "\u{4723}");
        decode_gb18030(b"\xFE\xFE", "\u{E4C5}");

        // Changes between GB18030-2005 and GB18030-2022
        decode_gb18030(b"\xFE\x7E", "\u{9FB9}");
        decode_gb18030(b"\xA6\xDD", "\u{FE14}");

        // These mappings remain in place the GB18030-2005 way despite GB18030-2022
        decode_gb18030(b"\x82\x35\x91\x32", "\u{9FB9}");
        decode_gb18030(b"\x84\x31\x83\x30", "\u{FE14}");

        // The difference from the original GB18030
        decode_gb18030(b"\xA3\xA0", "\u{3000}");
        decode_gb18030(b"\xA1\xA1", "\u{3000}");

        // 0xFF
        decode_gb18030(b"\xFF\x40", "\u{FFFD}\u{0040}");
        decode_gb18030(b"\xE3\xFF\x9A\x33", "\u{FFFD}\u{FFFD}"); // not \u{FFFD}\u{FFFD}\u{0033} !
        decode_gb18030(b"\xFF\x32\x9A\x33", "\u{FFFD}\u{0032}\u{FFFD}"); // not \u{FFFD}\u{0032}\u{FFFD}\u{0033} !
        decode_gb18030(b"\xFF\x40\x00", "\u{FFFD}\u{0040}\u{0000}");
        decode_gb18030(b"\xE3\xFF\x9A\x33\x00", "\u{FFFD}\u{FFFD}\u{0033}\u{0000}");
        decode_gb18030(
            b"\xFF\x32\x9A\x33\x00",
            "\u{FFFD}\u{0032}\u{FFFD}\u{0033}\u{0000}",
        );

        // Four bytes
        decode_gb18030(b"\x81\x30\x81\x30", "\u{0080}");
        decode_gb18030(b"\x81\x35\xF4\x37", "\u{E7C7}");
        decode_gb18030(b"\x81\x37\xA3\x30", "\u{2603}");
        decode_gb18030(b"\x94\x39\xDA\x33", "\u{1F4A9}");
        decode_gb18030(b"\xE3\x32\x9A\x35", "\u{10FFFF}");
        decode_gb18030(b"\xE3\x32\x9A\x36\x81\x30", "\u{FFFD}\u{FFFD}");
        decode_gb18030(b"\xE3\x32\x9A\x36\x81\x40", "\u{FFFD}\u{4E02}");
        decode_gb18030(b"\xE3\x32\x9A", "\u{FFFD}"); // not \u{FFFD}\u{0032}\u{FFFD} !
        decode_gb18030(b"\xE3\x32\x9A\x00", "\u{FFFD}\u{0032}\u{FFFD}\u{0000}");
    }

    #[test]
    fn test_gb18030_encode() {
        // Empty
        encode_gb18030("", b"");

        // ASCII
        encode_gb18030("\u{0061}\u{0062}", b"\x61\x62");

        // euro
        encode_gb18030("\u{20AC}", b"\xA2\xE3");

        // two bytes
        encode_gb18030("\u{4E02}", b"\x81\x40");
        encode_gb18030("\u{4E8A}", b"\x81\x7E");
        if !cfg!(miri) {
            // Miri is too slow
            encode_gb18030("\u{4E90}", b"\x81\x80");
            encode_gb18030("\u{4FA2}", b"\x81\xFE");
            encode_gb18030("\u{FA0C}", b"\xFE\x40");
            encode_gb18030("\u{E843}", b"\xFE\x7E");
            encode_gb18030("\u{4723}", b"\xFE\x80");
            encode_gb18030("\u{E4C5}", b"\xFE\xFE");
        }

        // The difference from the original GB18030
        encode_gb18030("\u{E5E5}", b"");
        encode_gb18030("\u{3000}", b"\xA1\xA1");

        // Four bytes
        encode_gb18030("\u{0080}", b"\x81\x30\x81\x30");
        encode_gb18030("\u{E7C7}", b"\x81\x35\xF4\x37");
        if !cfg!(miri) {
            // Miri is too slow
            encode_gb18030("\u{2603}", b"\x81\x37\xA3\x30");
            encode_gb18030("\u{1F4A9}", b"\x94\x39\xDA\x33");
            encode_gb18030("\u{10FFFF}", b"\xE3\x32\x9A\x35");
        }

        // Edge cases
        encode_gb18030("\u{00F7}", b"\xA1\xC2");

        // GB18030-2022
        encode_gb18030("\u{9FB9}", b"\xFE\x7E");
        encode_gb18030("\u{FE14}", b"\xA6\xDD");
        encode_gb18030("\u{E843}", b"\xFE\x7E");
        encode_gb18030("\u{E791}", b"\xA6\xDD");

        // Non-change in GB18030-2022
        encode_gb18030("\u{E817}", b"\xFE\x52");
    }

    #[test]
    fn test_gbk_encode() {
        // Empty
        encode_gbk("", b"");

        // ASCII
        encode_gbk("\u{0061}\u{0062}", b"\x61\x62");

        // euro
        encode_gbk("\u{20AC}", b"\x80");

        // two bytes
        encode_gbk("\u{4E02}", b"\x81\x40");
        encode_gbk("\u{4E8A}", b"\x81\x7E");
        if !cfg!(miri) {
            // Miri is too slow
            encode_gbk("\u{4E90}", b"\x81\x80");
            encode_gbk("\u{4FA2}", b"\x81\xFE");
            encode_gbk("\u{FA0C}", b"\xFE\x40");
            encode_gbk("\u{E843}", b"\xFE\x7E");
            encode_gbk("\u{4723}", b"\xFE\x80");
            encode_gbk("\u{E4C5}", b"\xFE\xFE");
        }

        // The difference from the original gb18030
        encode_gbk("\u{E5E5}", b"");
        encode_gbk("\u{3000}", b"\xA1\xA1");

        // Four bytes
        encode_gbk("\u{0080}", b"€");
        encode_gbk("\u{E7C7}", b"");
        if !cfg!(miri) {
            // Miri is too slow
            encode_gbk("\u{2603}", b"☃");
            encode_gbk("\u{1F4A9}", b"💩");
            encode_gbk("\u{10FFFF}", b"􏿿");
        }

        // Edge cases
        encode_gbk("\u{00F7}", b"\xA1\xC2");

        // GB18030-2022
        encode_gb18030("\u{9FB9}", b"\xFE\x7E");
        encode_gb18030("\u{FE14}", b"\xA6\xDD");
        encode_gb18030("\u{E843}", b"\xFE\x7E");
        encode_gb18030("\u{E791}", b"\xA6\xDD");

        // Non-change in GB18030-2022
        encode_gb18030("\u{E817}", b"\xFE\x52");
    }

    #[test]
    #[cfg_attr(miri, ignore)] // Miri is too slow
    fn test_gb18030_decode_all() {
        let input = include_bytes!("test_data/gb18030_in.txt");
        let expectation = include_str!("test_data/gb18030_in_ref.txt");
        let (cow, had_errors) = GB18030.decode_without_bom_handling(input);
        assert!(!had_errors, "Should not have had errors.");
        assert_eq!(&cow[..], expectation);
    }

    #[test]
    #[cfg_attr(miri, ignore)] // Miri is too slow
    fn test_gb18030_encode_all() {
        let input = include_str!("test_data/gb18030_out.txt");
        let expectation = include_bytes!("test_data/gb18030_out_ref.txt");
        let (cow, encoding, had_errors) = GB18030.encode(input);
        assert!(!had_errors, "Should not have had errors.");
        assert_eq!(encoding, GB18030);
        assert_eq!(&cow[..], &expectation[..]);
    }

    #[test]
    fn test_gb18030_encode_from_utf16_max_length() {
        let mut output = [0u8; 20];
        let mut encoder = GB18030.new_encoder();
        {
            let needed = encoder
                .max_buffer_length_from_utf16_without_replacement(1)
                .unwrap();
            let (result, read, written) = encoder.encode_from_utf16_without_replacement(
                &[0x3000],
                &mut output[..needed],
                true,
            );
            assert_eq!(result, EncoderResult::InputEmpty);
            assert_eq!(read, 1);
            assert_eq!(written, 2);
            assert_eq!(output[0], 0xA1);
            assert_eq!(output[1], 0xA1);
        }
    }
}