sorug 0.5.0

//! Minimal IDNA / Punycode for WHATWG host parsing.
//!
//! Implements RFC 3492 Bootstring (Punycode) encoding plus a deliberately small
//! UTS #46 mapping surface — enough for the WPT `urltestdata` suite without
//! shipping ICU / the `idna` crate at runtime.
//!
//! Membership classifiers (`CheckBidi`, `CheckJoiners`, disallowed, Ext-B, …)
//! use compile-time range tables from [`super::unicode_ranges`] generated by
//! `build.rs` from **vendored** Unicode UCD files under `data/ucd/` (pinned in
//! `UNICODE_VERSION`) merged with `data/idna_overlay.txt` (Node/WPT deltas).
//! Value-producing remaps stay in [`map_char`].
//!
//! WHATWG domain parser (`beStrict = false`):
//! - ASCII input → lowercase pass-through (even if ACE labels are invalid)
//! - Common IDN (Latin-1 lowercase / CJK / kana / Hangul) → Punycode only
//!   ([`Scan::SimpleIdentity`]; skips NFC + CheckBidi/Joiners)
//! - Other non-ASCII → map (if needed) → NFC → validity checks → Punycode

use alloc::borrow::Cow;
use alloc::string::String;
use alloc::vec::Vec;

use unicode_normalization::UnicodeNormalization;
use unicode_normalization::char::canonical_combining_class;

use super::percent::AppendBuf;
use super::unicode_ranges::{
    is_arabic_ext_b_compatible_cp, is_arabic_ext_b_era_mark_cp, is_arabic_ext_b_era_ok_cp,
    is_arabic_ext_b_letter_cp, is_bidi_ignored_mark, is_bidi_number_cp, is_disallowed_cp,
    is_joining_letter_cp, is_legacy_arabic_cp, is_ltr_letter_block, is_rtl_alphabetic_block,
    is_rtl_non_letter, is_uts46_ignored_cp, maps_nfkc_via_space_cp, uts46_needs_map_cp,
};

/// Convert a domain to ASCII per WHATWG (non-strict).
pub(crate) fn to_ascii(domain: &str) -> Result<Cow<'_, str>, ()> {
    if domain.is_empty() {
        return Err(());
    }

    if domain.is_ascii() {
        let lower = lowercase_ascii(domain);
        if contains_forbidden_domain_code_point(lower.as_ref()) {
            return Err(());
        }
        return Ok(lower);
    }

    let mut out = String::with_capacity(domain.len().saturating_add(16));
    to_ascii_into(domain, &mut out)?;
    if out.is_empty() || contains_forbidden_domain_code_point(&out) {
        return Err(());
    }
    Ok(Cow::Owned(out))
}

/// Append the WHATWG ToASCII form of `domain` to `out`.
///
/// `domain` must be non-empty and non-ASCII (ASCII callers should use the
/// borrow-friendly [`to_ascii`] path). On error, `out` may contain a partial write.
pub(crate) fn to_ascii_into(domain: &str, out: &mut impl AppendBuf) -> Result<(), ()> {
    debug_assert!(!domain.is_empty());
    debug_assert!(!domain.is_ascii());

    match scan_non_ascii(domain)? {
        // Common IDN (Latin-1 / CJK / kana / Hangul): Punycode only — no NFC or
        // CheckBidi/CheckJoiners (those code points cannot fail those checks).
        Scan::SimpleIdentity => encode_domain_labels_simple(domain, out)?,
        Scan::Identity => encode_domain_labels(domain, out)?,
        Scan::NeedsMap => {
            let mapped = uts46_map_checked(domain)?;
            if mapped.is_empty() {
                return Err(());
            }
            encode_domain_labels(&mapped, out)?;
        }
    }

    Ok(())
}

/// Like [`to_ascii_into`], but validates the written slice and rolls back `out`
/// to `start_len` on failure. Used when writing directly into the URL buffer.
pub(crate) fn to_ascii_append_validated(
    domain: &str,
    out: &mut super::serialization::SerializationBuf<'_>,
) -> Result<(), ()> {
    if domain.is_empty() {
        return Err(());
    }

    let start = out.len();

    if domain.is_ascii() {
        if domain.bytes().all(|b| !b.is_ascii_uppercase()) {
            if contains_forbidden_domain_code_point(domain) {
                return Err(());
            }
            out.push_str(domain);
        } else {
            let lower = domain.to_ascii_lowercase();
            if contains_forbidden_domain_code_point(&lower) {
                return Err(());
            }
            out.push_str(&lower);
        }
    } else if to_ascii_into(domain, out).is_err() {
        out.truncate(start);
        return Err(());
    } else {
        let written = &out.as_str()[start..];
        if written.is_empty() || contains_forbidden_domain_code_point(written) {
            out.truncate(start);
            return Err(());
        }
    }

    Ok(())
}

#[derive(Copy, Clone, Debug, Eq, PartialEq)]
enum Scan {
    /// Non-ASCII but every code point is in the simple-identity set (Latin-1
    /// lowercase / CJK / kana / Hangul). Skip NFC + CheckBidi/Joiners.
    SimpleIdentity,
    /// No UTS #46 mapping / case folding required — encode labels as-is.
    Identity,
    /// Must run [`uts46_map_checked`] before encoding.
    NeedsMap,
}

/// Single pass: reject disallowed code points and detect whether mapping is needed.
fn scan_non_ascii(domain: &str) -> Result<Scan, ()> {
    let mut needs_map = false;
    let mut all_simple = true;
    for c in domain.chars() {
        if is_disallowed_idna(c) {
            return Err(());
        }
        if c.is_ascii() {
            if c.is_ascii_uppercase() {
                needs_map = true;
                all_simple = false;
            }
            continue;
        }
        if is_simple_identity_cp(c) {
            continue;
        }
        all_simple = false;
        if !needs_map && char_needs_uts46_map(c) {
            needs_map = true;
        }
    }
    Ok(if needs_map {
        Scan::NeedsMap
    } else if all_simple {
        Scan::SimpleIdentity
    } else {
        Scan::Identity
    })
}

/// Code points that are already UTS #46 identity, NFC, LTR, and joiner-safe.
/// Covers the Criterion IDNA bench (`türkçe`) and common CJK/kana/Hangul IDNs.
///
/// Must not include combining marks: skipping NFC for those would both accept
/// leading Mn (should reject) and produce distinct ACE labels for precomposed
/// vs decomposed kana (homograph risk).
#[inline]
fn is_simple_identity_cp(c: char) -> bool {
    let u = c as u32;
    // Latin-1 lowercase letters + ß (non-transitional keeps ß).
    // Excludes × (U+00D7) and ÷ (U+00F7) by range gaps.
    if (0x00E0..=0x00F6).contains(&u) || (0x00F8..=0x00FF).contains(&u) || u == 0x00DF {
        return true;
    }
    // CJK Unified Ideographs + Extension A (all Lo; no combining marks).
    if (0x3400..=0x4DBF).contains(&u) || (0x4E00..=0x9FFF).contains(&u) {
        return true;
    }
    // Hiragana / Katakana — exclude unassigned U+3040/U+3097/U+3098 and Mn U+3099/U+309A.
    if (0x3041..=0x3096).contains(&u)
        || (0x309B..=0x309F).contains(&u)
        || (0x30A0..=0x30FF).contains(&u)
    {
        return true;
    }
    // Hangul syllables (Unicode ends at U+D7A3; U+D7A4..=U+D7AF are unassigned).
    (0xAC00..=0xD7A3).contains(&u)
}

#[inline]
fn char_needs_uts46_map(c: char) -> bool {
    // Static membership (ignored / rewritten / compatibility) from generated tables.
    if uts46_needs_map_cp(c) || maps_nfkc_via_space(c) || letterlike_nfkc_ascii(c).is_some() {
        return true;
    }
    if c.is_ascii() {
        return c.is_ascii_uppercase();
    }
    // Hot IDN letters: no map / case fold / NFKC.
    if is_simple_identity_cp(c) {
        return false;
    }
    // Latin-1 uppercase → needs fold (before expensive NFKC).
    let u = c as u32;
    if (0x00C0..=0x00D6).contains(&u) || (0x00D8..=0x00DE).contains(&u) {
        return true;
    }
    // Detect U+0345 without allocating an NFD `String`.
    let mut has_ypogegrammeni = false;
    unicode_normalization::char::decompose_canonical(c, |d| {
        if d == '\u{0345}' {
            has_ypogegrammeni = true;
        }
    });
    if has_ypogegrammeni {
        return true;
    }
    let mut it = c.nfkc();
    match (it.next(), it.next()) {
        (Some(ch), None) if ch == c => needs_unicode_case_fold(c),
        _ => true, // expands or remaps under NFKC
    }
}

/// True when `c.to_lowercase()` is not the single code point `c`.
#[inline]
fn needs_unicode_case_fold(c: char) -> bool {
    let u = c as u32;
    // Fast path: already-lowercase Latin-1 letters (covers ü/ç/ö/… IDN bench).
    if (0x00E0..=0x00F6).contains(&u) || (0x00F8..=0x00FF).contains(&u) || u == 0x00DF {
        return false;
    }
    // Latin-1 uppercase → needs fold.
    if (0x00C0..=0x00D6).contains(&u) || (0x00D8..=0x00DE).contains(&u) {
        return true;
    }
    // General Unicode: only call std tables when outside the Latin-1 shortcuts.
    let mut iter = c.to_lowercase();
    match (iter.next(), iter.next()) {
        (Some(l), None) => l != c,
        _ => true,
    }
}

/// Minimal UTS #46 disallowed set covering WPT failures (noncharacters, spaces,
/// replacement char, line/paragraph separators). Full UTS46 tables are
/// intentionally not embedded — membership ranges are derived at build time
/// from vendored UCD (`data/ucd/`) plus `data/idna_overlay.txt`.
#[inline]
fn is_disallowed_idna(c: char) -> bool {
    if is_disallowed_cp(c) {
        return true;
    }
    let u = c as u32;
    // Unassigned Unicode planes 4..=13 (e.g. U+D71CB).
    if matches!(u >> 16, 4..=13) {
        return true;
    }
    // Noncharacters: any plane's U+FFFE / U+FFFF.
    (u & 0xFFFE) == 0xFFFE
}

fn lowercase_ascii(s: &str) -> Cow<'_, str> {
    if s.bytes().all(|b| !b.is_ascii_uppercase()) {
        Cow::Borrowed(s)
    } else {
        Cow::Owned(s.to_ascii_lowercase())
    }
}

/// Forbidden host code point, C0 control, `%`, or DEL.
#[inline]
fn contains_forbidden_domain_code_point(s: &str) -> bool {
    s.bytes().any(|b| {
        matches!(
            b,
            0x00..=0x1F
                | b' '
                | b'#'
                | b'/'
                | b':'
                | b'<'
                | b'>'
                | b'?'
                | b'@'
                | b'['
                | b'\\'
                | b']'
                | b'^'
                | b'|'
                | b'%'
                | 0x7F
        )
    })
}

/// Minimal CheckBidi (RFC 5893 subset used by UTS #46 CheckBidi=true):
/// - Reject labels mixing RTL script with LTR letters.
/// - RTL labels must start with an RTL letter and end with RTL or a digit (EN/AN).
fn label_fails_check_bidi(label: &str) -> bool {
    let mut has_rtl = false;
    let mut has_ltr = false;
    for c in label.chars() {
        // NSM / combining marks / ZWNJ are neither R nor L for CheckBidi mixing.
        if is_bidi_transparent(c) {
            continue;
        }
        if is_rtl_script(c) {
            has_rtl = true;
        } else if is_ltr_letter(c) {
            has_ltr = true;
        }
        if has_rtl && has_ltr {
            return true;
        }
    }
    if has_rtl && !has_ltr {
        let mut chars = label.chars();
        let first = chars.next();
        // Skip trailing NSM (e.g. Thaana vowel signs U+07A6..=U+07B0): browsers
        // / ICU treat the last *strong* char as the CheckBidi end (Node accepts
        // `\u{7a3}\u{7aa}` → xn--hrbo).
        let last_strong = label.chars().rev().find(|c| !is_bidi_transparent(*c));
        let last_ok = last_strong.is_some_and(|c| is_rtl_script(c) || is_bidi_number(c));
        if !first.is_some_and(is_rtl_script) || !last_ok {
            return true;
        }
    }
    false
}

#[inline]
fn is_bidi_transparent(c: char) -> bool {
    // NSM / combining marks from the generated table, plus ZWNJ/ZWJ (CheckJoiners).
    is_bidi_ignored_mark(c) || matches!(c, '\u{200C}' | '\u{200D}')
}

/// EN (ASCII digits) or AN (Arabic-Indic digits) — valid RTL-label trailers.
#[inline]
fn is_bidi_number(c: char) -> bool {
    c.is_ascii_digit() || is_bidi_number_cp(c)
}

#[inline]
fn is_rtl_script(c: char) -> bool {
    // Strong RTL letters. Arabic/Hebrew *marks* (Mn), e.g. U+0613, must not
    // trip CheckBidi mixing even though Rust's `is_alphabetic()` is true for some.
    // Neutrals like U+06DE stay out so `\u{6de}e` still encodes.
    if is_bidi_ignored_mark(c) {
        return false;
    }
    if is_rtl_non_letter(c) {
        return true;
    }
    c.is_alphabetic() && is_rtl_alphabetic_block(c)
}

/// Node/ICU rejects Arabic Extended-B letters mixed with older Arabic letters or
/// other RTL scripts. Allowed companions: other Ext-B, U14+ Ext-A (U+08B5,
/// U+08C8..), marks/NSM, EN/AN digits, and neutrals.
fn label_fails_arabic_ext_b_mix(label: &str) -> bool {
    let mut has_ext_b = false;
    let mut has_incompatible = false;
    for c in label.chars() {
        if is_bidi_transparent(c) {
            continue;
        }
        if is_arabic_ext_b_letter(c) {
            has_ext_b = true;
            continue;
        }
        if c == '\u{0888}' {
            // Ext-B Sk symbol — allowed with Ext-B letters; counts as Ext-B context.
            has_ext_b = true;
            continue;
        }
        if is_arabic_ext_b_compatible(c) || is_bidi_number(c) || c.is_ascii() {
            continue;
        }
        // Other strong RTL / legacy Arabic letters / LTR letters are incompatible.
        if is_rtl_script(c) || is_legacy_arabic_letter(c) || is_ltr_letter(c) {
            has_incompatible = true;
        }
        if has_ext_b && has_incompatible {
            return true;
        }
    }
    has_ext_b && has_incompatible
}

#[inline]
fn is_arabic_ext_b_letter(c: char) -> bool {
    is_arabic_ext_b_letter_cp(c)
}

/// Ext-A letters that Node allows beside Arabic Extended-B (Unicode 14+).
#[inline]
fn is_arabic_ext_b_compatible(c: char) -> bool {
    is_arabic_ext_b_compatible_cp(c)
}

/// U14 Ext-B / early Ext-A Mn: Node allows with Ext-B / U14 Ext-A / LTR, not with
/// classic Arabic / Thaana / other legacy RTL.
#[inline]
fn is_arabic_ext_b_era_mark(c: char) -> bool {
    is_arabic_ext_b_era_mark_cp(c)
}

fn label_fails_ext_b_era_mark_mix(label: &str) -> bool {
    if !label.chars().any(is_arabic_ext_b_era_mark) {
        return false;
    }
    for c in label.chars() {
        if is_bidi_transparent(c) || is_arabic_ext_b_era_mark(c) {
            continue;
        }
        if is_arabic_ext_b_letter(c)
            || c == '\u{0888}'
            || is_arabic_ext_b_era_ok_cp(c)
            || is_bidi_number(c)
            || c.is_ascii()
            || is_ltr_letter(c)
        {
            continue;
        }
        if is_rtl_script(c) || is_legacy_arabic_letter(c) {
            return true;
        }
    }
    false
}

#[inline]
fn is_legacy_arabic_letter(c: char) -> bool {
    is_legacy_arabic_cp(c)
}

/// Minimal CheckJoiners (UTS #46):
/// - ZWNJ (U+200C): only between two joining letters (Arabic/Syriac; NSM skipped).
/// - ZWJ (U+200D): ContextJ — preceding code point has CCC=Virama (9) (Node/ICU).
fn label_fails_check_joiners(label: &str) -> bool {
    // UTF-8 for U+200C/U+200D is E2 80 8C / E2 80 8D — skip Vec alloc when absent.
    if !label_may_contain_zwj_or_zwnj(label) {
        return false;
    }
    let chars: Vec<char> = label.chars().collect();
    for (i, &c) in chars.iter().enumerate() {
        if c == '\u{200D}' {
            // RFC 5892 §A.2: Before(cp) is the immediately preceding character.
            match chars.get(i.wrapping_sub(1)).filter(|_| i > 0) {
                Some(&b) if canonical_combining_class(b) == 9 => {}
                _ => return true,
            }
            continue;
        }
        if c != '\u{200C}' {
            continue;
        }
        let before = chars[..i].iter().rev().find(|c| !is_bidi_transparent(**c));
        let after = chars[i + 1..].iter().find(|c| !is_bidi_transparent(**c));
        match (before, after) {
            (Some(&b), Some(&a)) if is_joining_letter(b) && is_joining_letter(a) => {}
            _ => return true,
        }
    }
    false
}

#[inline]
fn label_may_contain_zwj_or_zwnj(label: &str) -> bool {
    let bytes = label.as_bytes();
    let mut i = 0;
    while i + 2 < bytes.len() {
        if bytes[i] == 0xE2 && bytes[i + 1] == 0x80 && matches!(bytes[i + 2], 0x8C | 0x8D) {
            return true;
        }
        i += 1;
    }
    false
}

#[inline]
fn is_joining_letter(c: char) -> bool {
    c.is_alphabetic() && is_joining_letter_cp(c)
}

#[inline]
fn is_ltr_letter(c: char) -> bool {
    // ASCII / Latin / Greek / Cyrillic *letters* only (not ÷ U+00F7 etc.).
    c.is_alphabetic() && (c.is_ascii_alphabetic() || is_ltr_letter_block(c))
}

/// Fast path for [`Scan::SimpleIdentity`]: Punycode-encode non-ASCII labels only.
fn encode_domain_labels_simple(mapped: &str, out: &mut impl AppendBuf) -> Result<(), ()> {
    let mut first = true;
    for label in mapped.split('.') {
        if !first {
            out.push('.');
        }
        first = false;
        if label.is_empty() {
            continue;
        }
        if label.is_ascii() {
            out.push_str(label);
        } else {
            out.push_str("xn--");
            encode_punycode(label, out)?;
        }
    }
    Ok(())
}

fn encode_domain_labels(mapped: &str, out: &mut impl AppendBuf) -> Result<(), ()> {
    let mut first = true;
    for label in mapped.split('.') {
        if !first {
            out.push('.');
        }
        first = false;
        if label.is_empty() {
            continue;
        }
        // UTS #46: NFC before validity checks + Punycode. Composition can move a
        // mark onto a prior starter (e.g. n+virama+cedilla → ņ+virama), which
        // CheckJoiners must observe so ZWJ still sees CCC=Virama immediately before.
        let nfc_owned;
        let label = if unicode_normalization::is_nfc(label) {
            label
        } else {
            nfc_owned = label.nfc().collect::<String>();
            nfc_owned.as_str()
        };
        // Leading combining mark (no base character) — reject (ada/rust-url).
        if label.chars().next().is_some_and(is_bidi_ignored_mark) {
            return Err(());
        }
        if label_fails_check_bidi(label) {
            return Err(());
        }
        if label_fails_arabic_ext_b_mix(label) {
            return Err(());
        }
        if label_fails_ext_b_era_mark_mix(label) {
            return Err(());
        }
        if label_fails_check_joiners(label) {
            return Err(());
        }
        if label.is_ascii() {
            out.push_str(label);
        } else {
            out.push_str("xn--");
            encode_punycode(label, out)?;
        }
    }
    Ok(())
}

/// Map + reject disallowed code points that appear only after mapping.
fn uts46_map_checked(input: &str) -> Result<String, ()> {
    let mut out = String::with_capacity(input.len());
    for c in input.chars() {
        map_char(c, &mut out);
    }
    if out.is_empty()
        || out.chars().any(is_disallowed_idna)
        || contains_forbidden_domain_code_point(&out)
    {
        return Err(());
    }
    Ok(out)
}

fn map_char(c: char, out: &mut String) {
    // Ignored (UTS #46): soft hyphen, ZWSP, WJ, BOM, CGJ, Hangul fillers,
    // variation selectors (incl. supplementary) — membership from generated table.
    if is_uts46_ignored_cp(c) {
        return;
    }

    match c {
        // U+2000..=U+200A space-ish → ASCII space (then forbidden-host rejects).
        '\u{2000}'..='\u{200A}' => out.push(' '),

        // Non-breaking hyphen → NFKC hyphen (U+2010); stays non-ASCII for Punycode.
        '\u{2011}' => out.push('\u{2010}'),

        // Ideographic full stop / halfwidth variant → ASCII dot.
        '\u{3002}' | '\u{FF61}' => out.push('.'),

        // Cyrillic Extended-C (UTS #46 mapped; NFKC is identity).
        '\u{1C80}' => out.push('\u{0432}'), // rounded ve → в
        '\u{1C81}' => out.push('\u{0434}'), // long-legged de → д
        '\u{1C82}' => out.push('\u{043E}'), // narrow o → о
        '\u{1C83}' => out.push('\u{0441}'), // wide es → с
        '\u{1C84}' | '\u{1C85}' => out.push('\u{0442}'), // tall/three-legged te → т
        '\u{1C86}' => out.push('\u{044A}'), // tall hard sign → ъ
        '\u{1C87}' => out.push('\u{0463}'), // tall yat → ѣ
        '\u{1C88}' => out.push('\u{A64B}'), // unblended uk → ꙋ
        '\u{1C89}' => out.push('\u{1C8A}'), // capital tje → small tje

        // Cherokee small letters U+13F8..=U+13FD → capitals U+13F0..=U+13F5
        // (UTS #46; must not case-fold to U+AB70 phonetic extensions).
        '\u{13F8}' => out.push('\u{13F0}'),
        '\u{13F9}' => out.push('\u{13F1}'),
        '\u{13FA}' => out.push('\u{13F2}'),
        '\u{13FB}' => out.push('\u{13F3}'),
        '\u{13FC}' => out.push('\u{13F4}'),
        '\u{13FD}' => out.push('\u{13F5}'),

        // Cherokee Supplement small letters → capital Cherokee (UTS #46 mapped).
        // Rust `to_lowercase` is a no-op here; case fold is AB70+n → 13A0+n.
        c @ '\u{AB70}'..='\u{ABBF}' => {
            out.push(char::from_u32(0x13A0 + (c as u32 - 0xAB70)).unwrap_or(c));
        }

        // U+203E OVERLINE → NFKC `<space><combining overline>`; space then fails
        // the forbidden-host check (matches ada / rust-url rejection).
        '\u{203E}' => {
            out.push(' ');
            out.push('\u{0305}');
        }

        // Compatibility chars whose NFKC introduces SPACE (then forbidden-host rejects).
        c if maps_nfkc_via_space(c) => {
            for ch in c.nfkc() {
                out.push(ch);
            }
        }

        // Vulgar fractions / Number Forms → NFKC (e.g. ¼ → 1⁄4, ⅛ → 1⁄8).
        // Case-fold too: U+2183 ROMAN NUMERAL REVERSED ONE HUNDRED → U+2184.
        '\u{00BC}'..='\u{00BE}' | '\u{2150}'..='\u{2189}' => {
            for ch in c.nfkc() {
                for folded in ch.to_lowercase() {
                    if folded.is_ascii() {
                        out.push(folded.to_ascii_lowercase());
                    } else {
                        out.push(folded);
                    }
                }
            }
        }

        // Superscripts / subscripts → NFKC (e.g. ⁵ → 5) so IPv4 ends-in-a-number works.
        '\u{00B2}' | '\u{00B3}' | '\u{00B9}' | '\u{2070}'..='\u{209F}' => {
            for ch in c.nfkc() {
                out.push(ch);
            }
        }

        // U+0345 COMBINING GREEK YPOGEGRAMMENI → ι (Unicode case fold).
        '\u{0345}' => out.push('\u{03B9}'),
        // U+03F2 LUNATE SIGMA → U+03C3 (browsers keep U+03C2 final sigma as-is).
        '\u{03F2}' => out.push('\u{03C3}'),
        // Mathematical *final* sigma variants: UTS #46 maps to U+03C3 (not NFKC's U+03C2).
        '\u{1D6D3}' | '\u{1D70D}' | '\u{1D747}' | '\u{1D781}' | '\u{1D7BB}' => out.push('\u{03C3}'),
        '\u{00AA}'
        | '\u{00BA}'
        | '\u{01C4}'..='\u{01CC}'
        | '\u{03D0}'..='\u{03D7}'
        | '\u{03F0}'..='\u{03F1}'
        | '\u{03F4}'..='\u{03F5}' => {
            for ch in c.nfkc() {
                for folded in ch.to_lowercase() {
                    if folded.is_ascii() {
                        out.push(folded.to_ascii_lowercase());
                    } else {
                        out.push(folded);
                    }
                }
            }
        }

        // Fullwidth ASCII (! through ~): U+FF01..=U+FF5E → U+0021..=U+007E
        '\u{FF01}'..='\u{FF5E}' => {
            let mapped = char::from_u32(u32::from(c) - 0xFEE0).unwrap_or(c);
            out.push(mapped.to_ascii_lowercase());
        }

        // Mathematical Bold Capitals A–Z (U+1D400..=U+1D419)
        c if (0x1D400..=0x1D419).contains(&(c as u32)) => {
            out.push(char::from(b'a' + (c as u32 - 0x1D400) as u8));
        }
        // Mathematical Bold Small a–z (U+1D41A..=U+1D433)
        c if (0x1D41A..=0x1D433).contains(&(c as u32)) => {
            out.push(char::from(b'a' + (c as u32 - 0x1D41A) as u8));
        }

        // Letterlike Symbols → ASCII NFKC (then lowercased).
        c if letterlike_nfkc_ascii(c).is_some() => {
            let mapped = letterlike_nfkc_ascii(c).unwrap();
            for ch in mapped.chars() {
                out.push(ch.to_ascii_lowercase());
            }
        }

        c if c.is_ascii() => out.push(c.to_ascii_lowercase()),

        // Latin-1 already-lowercase / ß: push without std Unicode tables.
        c if {
            let u = c as u32;
            (0x00E0..=0x00F6).contains(&u) || (0x00F8..=0x00FF).contains(&u) || u == 0x00DF
        } =>
        {
            out.push(c);
        }

        // Latin-1 uppercase → lowercase without iterator.
        c if {
            let u = c as u32;
            (0x00C0..=0x00D6).contains(&u) || (0x00D8..=0x00DE).contains(&u)
        } =>
        {
            out.push(char::from_u32((c as u32) + 0x20).unwrap_or(c));
        }

        c => {
            // UTS #46 compatibility mapping via NFKC (ligatures, Hangul compat,
            // math symbols like U+2230 → 3×U+222E, …), then case fold.
            // Cherokee U+13A0..=U+13F5: ada/ICU keep uppercase code points (do not
            // map to U+AB70 small letters) — matches Node domainToASCII.
            for ch in c.nfkc() {
                if matches!(ch, '\u{13A0}'..='\u{13F5}') {
                    out.push(ch);
                    continue;
                }
                if ch == '\u{3002}' {
                    out.push('.');
                    continue;
                }
                // Greek ypogegrammeni forms (e.g. U+1FC4): NFD then U+0345 → ι
                // so NFC later yields ήι like ICU/Node (Rust to_lowercase keeps 1FC4).
                let nfd: String = ch.nfd().collect();
                if nfd.contains('\u{0345}') {
                    for d in nfd.chars() {
                        if d == '\u{0345}' {
                            out.push('\u{03B9}');
                        } else {
                            for folded in d.to_lowercase() {
                                if folded.is_ascii() {
                                    out.push(folded.to_ascii_lowercase());
                                } else {
                                    out.push(folded);
                                }
                            }
                        }
                    }
                    continue;
                }
                for folded in ch.to_lowercase() {
                    if folded.is_ascii() {
                        out.push(folded.to_ascii_lowercase());
                    } else {
                        out.push(folded);
                    }
                }
            }
        }
    }
}

/// Compatibility characters whose NFKC form starts with SPACE (rejected as host).
#[inline]
fn maps_nfkc_via_space(c: char) -> bool {
    maps_nfkc_via_space_cp(c)
}

/// Compact UTS #46 / NFKC ASCII mappings for Letterlike Symbols (U+2100..=U+214F).
#[inline]
fn letterlike_nfkc_ascii(c: char) -> Option<&'static str> {
    Some(match c {
        '\u{2100}' => "a/c",
        '\u{2101}' => "a/s",
        '\u{2102}' => "C",
        '\u{2105}' => "c/o",
        '\u{2106}' => "c/u",
        '\u{210A}' => "g",
        '\u{210B}' => "H",
        '\u{210C}' => "H",
        '\u{210D}' => "H",
        '\u{210E}' => "h",
        '\u{2110}' => "I",
        '\u{2111}' => "I",
        '\u{2112}' => "L",
        '\u{2113}' => "l",
        '\u{2115}' => "N",
        '\u{2116}' => "No",
        '\u{2119}' => "P",
        '\u{211A}' => "Q",
        '\u{211B}' => "R",
        '\u{211C}' => "R",
        '\u{211D}' => "R",
        '\u{2120}' => "SM",
        '\u{2121}' => "TEL",
        '\u{2122}' => "TM",
        '\u{2124}' => "Z",
        '\u{2128}' => "Z",
        '\u{212A}' => "K",
        '\u{212C}' => "B",
        '\u{212D}' => "C",
        '\u{212F}' => "e",
        '\u{2130}' => "E",
        '\u{2131}' => "F",
        '\u{2133}' => "M",
        '\u{2134}' => "o",
        '\u{2139}' => "i",
        '\u{213B}' => "FAX",
        '\u{2145}' => "D",
        '\u{2146}' => "d",
        '\u{2147}' => "e",
        '\u{2148}' => "i",
        '\u{2149}' => "j",
        _ => return None,
    })
}

// ---------------------------------------------------------------------------
// RFC 3492 Punycode encode
// ---------------------------------------------------------------------------

const BASE: u32 = 36;
const TMIN: u32 = 1;
const TMAX: u32 = 26;
const SKEW: u32 = 38;
const DAMP: u32 = 700;
const INITIAL_BIAS: u32 = 72;
const INITIAL_N: u32 = 128;

/// Stack buffer for typical IDN labels (DNS label ≤ 63 octets after ACE).
const STACK_LABEL_CHARS: usize = 64;

fn encode_punycode(input: &str, out: &mut impl AppendBuf) -> Result<(), ()> {
    #![allow(clippy::many_single_char_names)] // RFC 3492 names: n, h, b, m, q, k
    // Stack ACE buffer (DNS label ≤ 63; headroom for edge cases before reject).
    let mut ace = [0u8; 128];
    let mut ace_len = 0usize;
    let push_ascii = |ace: &mut [u8; 128], ace_len: &mut usize, b: u8| -> Result<(), ()> {
        if *ace_len >= ace.len() {
            return Err(());
        }
        ace[*ace_len] = b;
        *ace_len += 1;
        Ok(())
    };

    let mut n = INITIAL_N;
    let mut delta: u32 = 0;
    let mut bias = INITIAL_BIAS;

    let mut stack = [0u32; STACK_LABEL_CHARS];
    let mut heap: Vec<u32> = Vec::new();
    let mut len = 0usize;
    for c in input.chars() {
        let cp = c as u32;
        if len < STACK_LABEL_CHARS && heap.is_empty() {
            stack[len] = cp;
            len += 1;
        } else {
            if heap.is_empty() {
                heap.reserve(len + 8);
                heap.extend_from_slice(&stack[..len]);
            }
            heap.push(cp);
            len += 1;
        }
    }
    // Oversized inputs rejected early (DNS-scale labels stay well below this).
    if len > 256 {
        return Err(());
    }
    let chars: &[u32] = if heap.is_empty() {
        &stack[..len]
    } else {
        &heap
    };

    let mut basic_count = 0u32;
    for &cp in chars {
        if cp < 0x80 {
            push_ascii(&mut ace, &mut ace_len, (cp as u8).to_ascii_lowercase())?;
            basic_count += 1;
        }
    }

    let mut h = basic_count;
    let b = basic_count;
    if b > 0 {
        push_ascii(&mut ace, &mut ace_len, b'-')?;
    }

    while (h as usize) < chars.len() {
        let mut m = u32::MAX;
        for &cp in chars {
            if cp >= n && cp < m {
                m = cp;
            }
        }
        if m == u32::MAX {
            return Err(());
        }

        let advance = m.checked_sub(n).ok_or(())?;
        delta = delta
            .checked_add(advance.checked_mul(h.checked_add(1).ok_or(())?).ok_or(())?)
            .ok_or(())?;
        n = m;

        for &cp in chars {
            if cp < n {
                delta = delta.checked_add(1).ok_or(())?;
            } else if cp == n {
                let mut q = delta;
                let mut k = BASE;
                loop {
                    let t = if k <= bias {
                        TMIN
                    } else if k >= bias.saturating_add(TMAX) {
                        TMAX
                    } else {
                        k - bias
                    };
                    if q < t {
                        break;
                    }
                    let code = t + ((q - t) % (BASE - t));
                    push_ascii(&mut ace, &mut ace_len, digit_to_byte(code)?)?;
                    q = (q - t) / (BASE - t);
                    k = k.checked_add(BASE).ok_or(())?;
                }
                push_ascii(&mut ace, &mut ace_len, digit_to_byte(q)?)?;
                bias = adapt(delta, h + 1, h == b);
                delta = 0;
                h += 1;
            }
        }
        delta = delta.checked_add(1).ok_or(())?;
        n = n.checked_add(1).ok_or(())?;
    }
    let ace_str = core::str::from_utf8(&ace[..ace_len]).map_err(|_| ())?;
    out.push_str(ace_str);
    Ok(())
}

fn adapt(mut delta: u32, num_points: u32, first_time: bool) -> u32 {
    delta = if first_time { delta / DAMP } else { delta / 2 };
    delta += delta / num_points;
    let mut k = 0u32;
    while delta > ((BASE - TMIN) * TMAX) / 2 {
        delta /= BASE - TMIN;
        k += BASE;
    }
    k + (((BASE - TMIN + 1) * delta) / (delta + SKEW))
}

fn digit_to_byte(d: u32) -> Result<u8, ()> {
    match d {
        0..=25 => Ok(b'a' + d as u8),
        26..=35 => Ok(b'0' + (d as u8 - 26)),
        _ => Err(()),
    }
}

#[cfg(test)]
mod tests {
    use super::*;

    #[test]
    fn ascii_passthrough_invalid_punycode() {
        assert_eq!(
            to_ascii("a.b.c.xn--pokxncvks").unwrap(),
            "a.b.c.xn--pokxncvks"
        );
        assert_eq!(to_ascii("XN--").unwrap(), "xn--");
    }

    #[test]
    fn punycode_examples() {
        assert_eq!(to_ascii("é").unwrap(), "xn--9ca");
        assert_eq!(to_ascii("你好你好").unwrap(), "xn--6qqa088eba");
        assert_eq!(to_ascii("faß.ExAmPlE").unwrap(), "xn--fa-hia.example");
        assert_eq!(to_ascii("☃").unwrap(), "xn--n3h");
        assert!(to_ascii("\u{fffd}").is_err());
        assert!(to_ascii("GOO\u{a0}goo.com").is_err());
    }

    #[test]
    fn mapping_examples() {
        assert_eq!(to_ascii("Ｇｏ.com").unwrap(), "go.com");
        assert_eq!(to_ascii("www.foo。bar.com").unwrap(), "www.foo.bar.com");
        assert_eq!(
            to_ascii("GOO\u{200b}\u{2060}\u{feff}goo.com").unwrap(),
            "googoo.com"
        );
        assert_eq!(to_ascii("a\u{ad}b").unwrap(), "ab");
        assert_eq!(to_ascii("loC𝐀𝐋𝐇𝐨𝐬𝐭").unwrap(), "localhost");
        // U+210B SCRIPT CAPITAL H → h (letterlike NFKC)
        assert_eq!(to_ascii("\u{210B}").unwrap(), "h");
        // CheckBidi: Arabic + Cyrillic in one label
        assert!(to_ascii("ےў").is_err());
        // CheckBidi: RTL label must not start with hyphen
        assert!(to_ascii("-\u{068B}").is_err());
        // Unassigned plane
        assert!(to_ascii("\u{d71cb}").is_err());
        // OVERLINE maps to space+mark → forbidden
        assert!(to_ascii("\u{203E}").is_err());
        // Combining mark + Latin can fail joining/bidi depending on input; `*` alone is allowed
        // (UseSTD3ASCIIRules=false in WHATWG URL).
        assert_eq!(to_ascii("sw*").unwrap(), "sw*");
    }

    #[test]
    fn identity_turkish_bench_label() {
        assert_eq!(to_ascii("türkçe.com").unwrap(), "xn--trke-2oa7j.com");
        assert_eq!(scan_non_ascii("türkçe.com"), Ok(Scan::SimpleIdentity));
        assert_eq!(scan_non_ascii("你好你好.com"), Ok(Scan::SimpleIdentity));
        // Combining mark → full Identity / NeedsMap path (NFC required).
        assert_ne!(scan_non_ascii("a\u{0301}.com"), Ok(Scan::SimpleIdentity));
    }

    #[test]
    fn to_ascii_into_matches_to_ascii() {
        let domain = "türkçe.com";
        let mut buf = String::new();
        to_ascii_into(domain, &mut buf).unwrap();
        assert_eq!(buf, to_ascii(domain).unwrap());
    }
}

#[cfg(test)]
mod triage_nsm {
    use super::*;

    #[test]
    fn arabic_nsm_with_latin_ok() {
        let domain = "ex\u{613}";
        assert!(!domain.is_ascii());
        let scan = scan_non_ascii(domain);
        assert!(scan.is_ok(), "scan={scan:?}");
        println!("scan={:?}", scan.unwrap());

        // CheckBidi on label
        assert!(!label_fails_check_bidi(domain), "bidi failed");

        let mut buf = String::new();
        let enc = encode_domain_labels(domain, &mut buf);
        assert!(enc.is_ok(), "encode={enc:?} buf={buf}");
        println!("encoded={buf}");

        let r = to_ascii(domain);
        assert!(r.is_ok(), "to_ascii={r:?}");
    }
}

#[cfg(test)]
mod tick_fixes {
    use super::*;
    #[test]
    fn nfc_cedilla() {
        assert_eq!(to_ascii("--tabout\u{327}").unwrap(), "xn----tabou-drb");
    }
    #[test]
    fn leading_mark_rejected() {
        assert!(to_ascii("\u{34b}g").is_err());
        assert!(to_ascii("\u{5c7}e").is_err());
        // U+05CC is unassigned (not a mark) — rejected via disallowed set.
        assert!(to_ascii("\u{5cc}sw*").is_err());
    }
    #[test]
    fn nfkc_fractions_and_space_compat() {
        assert_eq!(to_ascii("sst\u{00bc}r").unwrap(), "xn--sst14r-sq0c");
        assert_eq!(to_ascii("h\u{215b}1").unwrap(), "xn--h181-1g7a");
        assert!(to_ascii("h\u{00a8}J").is_err());
        assert!(to_ascii("z\u{0385}").is_err());
        assert!(to_ascii("0w\u{02d9}\u{043f}.").is_err());
        assert!(to_ascii("8888888888888\u{202b}out8B88HTTP").is_err());
        assert!(to_ascii("+t\u{00fc}M\u{008e}=1\u{00e7}.com").is_err());
        assert_eq!(to_ascii("\u{01ca}s").unwrap(), "njs");
        assert_eq!(to_ascii("h\u{2011}").unwrap(), "xn--h-0gn");
        assert_eq!(to_ascii("\u{03d2}").unwrap(), "xn--6xa");
        assert_eq!(to_ascii("2e2\u{00aa}dta0").unwrap(), "2e2adta0");
        assert!(to_ascii("c\u{e691}").is_err());
        assert!(to_ascii("\u{200d}b").is_err());
        assert_eq!(
            to_ascii("255.255.255.2\u{2075}85").unwrap(),
            "255.255.255.2585"
        );
        assert_eq!(to_ascii("\u{587}7\u{187}7f").unwrap(), "xn--77f-h4a750c8e");
        assert_eq!(to_ascii("7\u{2230},").unwrap(), "xn--7,-cevaa");
        assert_eq!(to_ascii("\u{314f}ile").unwrap(), "xn--ile-3do");
        assert_eq!(to_ascii("\u{6de}e").unwrap(), "xn--e-fxc");
        assert!(to_ascii("\u{6de}").is_ok(), "{:?}", to_ascii("\u{6de}"));
        // Thaana letter + trailing vowel sign (NSM): CheckBidi end skips NSM.
        assert_eq!(to_ascii("\u{7a3}\u{7aa}").unwrap(), "xn--hrbo");
        assert_eq!(to_ascii("\u{7ff}\u{5e8}").unwrap(), "xn--teb38h");
        assert_eq!(to_ascii("\u{5c0}").unwrap(), "xn--odb");
        assert!(to_ascii("0x=7f00\u{248b}ac").is_err());
        assert_eq!(to_ascii("s\u{3f2}x").unwrap(), "xn--sx-vbc");
        assert_eq!(to_ascii("c\u{3c2}").unwrap(), "xn--c-ymb"); // keep final sigma
        assert_eq!(to_ascii("D\u{7fd}").unwrap(), "xn--d-ued");
        assert_eq!(to_ascii("\u{7ff}\u{7c8}").unwrap(), "xn--jsb3i");
        assert_eq!(to_ascii("a\u{1160}b").unwrap(), "ab");
        assert_eq!(to_ascii("5x\u{345}fl").unwrap(), "xn--5xfl-6md");
    }
    #[test]
    fn unassigned_greek() {
        assert!(to_ascii("s\u{38d}w").is_err());
    }
}

#[cfg(test)]
mod tick18 {
    use super::*;
    #[test]
    fn cases() {
        assert_eq!(
            to_ascii("\u{5c6}\u{7fd}\u{5c7}\u{6fd}").unwrap(),
            "xn--udbc44gi3a"
        );
        assert_eq!(
            to_ascii("87\u{116e}\u{13a5}\u{116b}B").unwrap(),
            "xn--87b-bfoo615a"
        );
    }
}

#[cfg(test)]
mod tick19 {
    use super::*;
    #[test]
    fn check_bidi_skips_cyrillic_marks() {
        assert_eq!(to_ascii("\u{5c0}\u{484}\u{5c5}").unwrap(), "xn--n3a54dpa");
    }
    #[test]
    fn greek_ypogegrammeni_expands() {
        assert_eq!(
            to_ascii("\u{1940}DXXE\u{1c80}").unwrap(),
            "xn--dxxe-j4d9443a"
        );
        assert_eq!(to_ascii("\u{13fa}").unwrap(), "xn--ice");
        assert_eq!(to_ascii("\u{1c89}").unwrap(), "xn--d4f");
        assert_eq!(to_ascii("\u{1fc4}\u{1c89}").unwrap(), "xn--jxaw371q");
        assert_eq!(to_ascii("\u{fd9e}\u{ff9e}").unwrap(), "xn--ngbm8fw67r");
        assert_eq!(to_ascii("\u{ffa0}.").unwrap(), ".");
        assert!(to_ascii("\u{ffa0}").is_err()); // ignored → empty host
        assert_eq!(
            to_ascii("\u{700}\u{1dc0}\u{1dc0}\u{1dc0}").unwrap(),
            "xn--tmb267iaa"
        );
        assert_eq!(to_ascii("\u{069c}\u{1032}").unwrap(), "xn--yjb220c");
        assert_eq!(
            to_ascii("\u{079c}\u{065c}\u{0658}\u{0a4b}").unwrap(),
            "xn--0hbh37gu8e"
        );
        assert_eq!(
            to_ascii("\u{079c}\u{065e}\u{065c}\u{0658}\u{0902}").unwrap(),
            "xn--0hbhg48jizc"
        );
        assert_eq!(
            to_ascii("vvvvv\u{fb1e}\u{fb1e}").unwrap(),
            "xn--vvvvv-xi82aa"
        );
        assert_eq!(to_ascii("f\u{fb1e}\u{ff9e}").unwrap(), "xn--f-wdu9988e");
        assert_eq!(
            to_ascii("\u{079c}\u{065c}\u{1886}").unwrap(),
            "xn--4hb35dg58a"
        );
        assert_eq!(to_ascii("\u{080f}2\u{09c1}").unwrap(), "xn--2-ufd71k");
        assert_eq!(to_ascii("\u{079c}\u{0861}").unwrap(), "xn--9qb80b");
        assert_eq!(
            to_ascii("\u{0642}\u{065c}\u{200c}\u{0642}\u{065c}").unwrap(),
            "xn--ehba1fb2993a"
        );
        assert!(to_ascii("a\u{200c}b").is_err());
        assert_eq!(to_ascii("\u{0629}\u{0c81}").unwrap(), "xn--ogb263a");
        assert_eq!(to_ascii("\u{0869}\u{0c81}").unwrap(), "xn--4wb01s");
        assert_eq!(to_ascii("\u{088f}").unwrap(), "xn--7xb");
        assert!(to_ascii("\u{088f}\u{084f}").is_err()); // CheckBidi mix
        assert!(to_ascii("\u{0890}").is_err());
        assert!(to_ascii("\u{1c8b}").is_err());
        assert!(to_ascii("\u{079c}\u{08d2}").is_err()); // Node rejects Ext-A mark mix
        assert!(to_ascii("AGAAA.\u{33c2}\u{3002}DFx").is_err());
        assert_eq!(to_ascii("\u{1940}DE\u{1fc4}").unwrap(), "xn--de-38b8b108q");
    }
}

#[cfg(test)]
mod tick20 {
    use super::*;
    #[test]
    fn division_not_ltr() {
        assert_eq!(to_ascii("\u{624}\u{f7}0001").unwrap(), "xn--0001-9qa360f");
        assert!(to_ascii("2a\u{2026}7g").is_err());
    }
}

#[cfg(test)]
mod tick21 {
    use super::*;
    #[test]
    fn check_bidi_tibetan_and_ext_marks() {
        // Tibetan Mn U+0F8F with Thaana (Node: xn--9qb481b).
        assert_eq!(to_ascii("\u{079c}\u{0f8f}").unwrap(), "xn--9qb481b");
        // Combining Ext U+1ABA with Syriac punctuation (Node: xn--tmb412h).
        assert_eq!(to_ascii("\u{0700}\u{1aba}").unwrap(), "xn--tmb412h");
        // U+1ACF unassigned / Node CheckBidi fail with RTL; alone is fine.
        assert!(to_ascii("\u{0798}\u{1acf}").is_err());
        assert_eq!(to_ascii("\u{1acf}").unwrap(), "xn--prf");
        assert!(to_ascii("\u{0798}\u{1ace}").is_err()); // Node rejects Ext Mn ≥1AC1
    }
}

#[cfg(test)]
mod tick22 {
    use super::*;
    #[test]
    fn vedic_thai_rejang_and_syriac_zwnj() {
        assert_eq!(to_ascii("\u{0704}\u{1ce7}").unwrap(), "xn--xmb023i");
        assert_eq!(to_ascii("\u{069c}\u{0e49}").unwrap(), "xn--yjb440b");
        assert_eq!(
            to_ascii("\u{0642}\u{065c}\u{200c}\u{0722}").unwrap(),
            "xn--ehb6cs6ah11f"
        );
        assert_eq!(
            to_ascii("\u{079c}\u{065e}\u{a94f}").unwrap(),
            "xn--6hb94dg266a"
        );
        assert!(to_ascii("\u{079c}\u{1ce1}").is_err()); // Vedic Mc
        assert!(to_ascii("\u{05d0}\u{200c}\u{0642}").is_err()); // Hebrew↔Arabic ZWNJ
    }
}

#[cfg(test)]
mod tick23 {
    use super::*;
    #[test]
    fn vedic_nihshvasa_and_cherokee_supplement() {
        // U+1CD3 Po is UTS #46 valid (NV8); not a leading Mn.
        assert_eq!(to_ascii("\u{1cd3}").unwrap(), "xn--g6f");
        assert!(to_ascii("\u{1cd0}").is_err()); // Mn alone
        // Cherokee Supplement → capital Cherokee (Node: xn--78d).
        assert_eq!(to_ascii("\u{ab72}").unwrap(), "xn--78d");
        assert_eq!(to_ascii("\u{13a2}").unwrap(), "xn--78d");
        assert_eq!(to_ascii("\u{ab70}").unwrap(), "xn--58d");
        assert_eq!(
            to_ascii("\u{069c}\u{0e49}\u{0ec9}").unwrap(),
            "xn--yjb440b8k"
        ); // Thai + Lao Mn
        assert_eq!(to_ascii("\u{069c}\u{0d81}").unwrap(), "xn--yjb446a");
        assert_eq!(to_ascii("\u{069c}\u{0dd4}").unwrap(), "xn--yjb018a");
        assert_eq!(to_ascii("\u{0704}\u{1a77}").unwrap(), "xn--xmb270h"); // Tai Tham Mn
        assert_eq!(
            to_ascii("\u{0704}\u{0704}\u{1ce7}\u{2de7}").unwrap(),
            "xn--xmba329nnrp"
        );
        assert_eq!(to_ascii("\u{0704}\u{1b37}").unwrap(), "xn--xmb654h");
        assert_eq!(to_ascii("\u{0700}\u{1b3a}").unwrap(), "xn--tmb074h");
        assert_eq!(to_ascii("\u{0851}=\u{083c}").unwrap(), "xn--=-gid5d");
        assert_eq!(
            to_ascii("\u{062c}\u{1c33}\u{061a}\u{1b73}").unwrap(),
            "xn--8fb9a697i5va"
        ); // Lepcha Mn
        assert!(to_ascii("\u{088f}\u{fb9e}").is_err()); // Node rejects Ext-B + Arabic
        assert_eq!(to_ascii("\u{069c}\u{1bef}").unwrap(), "xn--yjb230i"); // Batak Mn
        assert_eq!(to_ascii("\u{069c}\u{1bed}").unwrap(), "xn--yjb820i");
        assert_eq!(to_ascii("\u{0834}\u{1b80}").unwrap(), "xn--mvb499g"); // Sundanese Mn
        assert_eq!(
            to_ascii("\u{08b5}\u{aac1}\u{0ac1}").unwrap(),
            "xn--bzb26h302y"
        ); // Tai Viet Mn
        assert!(to_ascii("\u{0868}\u{08d1}\u{0a51}").is_err()); // Node rejects Ext-A mark mix
        assert!(to_ascii("\u{17b4}").is_err()); // Khmer inherent — ignored → empty
        assert_eq!(to_ascii("\u{0624}\u{17b4}").unwrap(), "xn--jgb");
        assert_eq!(to_ascii("\u{0851}\u{a9bc}").unwrap(), "xn--gwb0673f"); // Javanese Mn
        assert_eq!(
            to_ascii("\u{0704}\u{1773}\u{1cdd}").unwrap(),
            "xn--xmb825f25d"
        ); // Tagbanwa Mn
        assert_eq!(to_ascii("\u{0851}\u{aaf6}").unwrap(), "xn--gwb8834f");
        assert_eq!(
            to_ascii("\u{0624}\u{abed}\u{1bed}").unwrap(),
            "xn--jgb862i025h"
        );
        assert_eq!(to_ascii("\u{0851}\u{aa36}").unwrap(), "xn--gwb4004f"); // Cham Mn
        assert_eq!(to_ascii("\u{0851}\u{a8f0}").unwrap(), "xn--gwb2533f"); // Combining Devanagari
        // Old Hungarian (Bidi=R) + Hebrew NSM/letter — CheckBidi must treat OH as RTL
        assert_eq!(
            to_ascii("\u{10c9d}\u{05c7}\u{05dd}\u{10cb2}").unwrap(),
            "xn--vdb8b4637jqca"
        );
        // Avestan (Bidi=R) + Old South Arabian
        assert_eq!(to_ascii("\u{10b1d}\u{10a72}").unwrap(), "xn--hu9cll");
        // Syriac punct + Tagalog Mn
        assert_eq!(to_ascii("\u{0704}\u{1713}").unwrap(), "xn--xmb633f");
        assert_eq!(to_ascii("\u{10c32}\u{11c32}").unwrap(), "xn--969c66y"); // Bhaiksuki Mn
        assert_eq!(to_ascii("\u{0851}\u{a80b}").unwrap(), "xn--gwb4982f"); // Syloti Mn
        assert_eq!(to_ascii("\u{10e9d}\u{10e72}").unwrap(), "xn--qo0dzc"); // Rumi AN trailer
        assert_eq!(to_ascii("\u{10810}&0\u{11c32}").unwrap(), "xn--&0-li9nk85d");
        // Meroitic Cursive number (Bidi=R, No) + letter
        assert_eq!(to_ascii("\u{109f2}\u{109b2}").unwrap(), "xn--zo9che");
        // Kharoshthi + Nabataean (both Bidi=R)
        assert_eq!(to_ascii("\u{10a1d}\u{1089d}").unwrap(), "xn--2g9crz");
        // Sidetic (U+10945) + Old North Arabian — Node rejects (CheckBidi); we match Node
        assert!(to_ascii("\u{10a85}\u{10a85}\u{10945}").is_err());
        assert_eq!(to_ascii("\u{10b0b}\u{10f1d}").unwrap(), "xn--uy9co3e"); // Old Sogdian number
        assert_eq!(
            to_ascii("\u{0624}\u{1bed}\u{a92d}").unwrap(),
            "xn--jgb862ip43h"
        ); // Kayah Li Mn
        assert_eq!(to_ascii("\u{10b1d}\u{10a9d}").unwrap(), "xn--pv9cpi"); // ONA number
        assert_eq!(
            to_ascii("\u{10a9d}\u{10e9d}.\u{10e9d}\u{10e9d}.\u{10cb2}").unwrap(),
            "xn--pv9ci2e.xn--yp0da.xn--rd0d"
        );
        assert_eq!(to_ascii("\u{10a85}\u{11a05}").unwrap(), "xn--0u9cq8x"); // Zanabazar Mn
        assert_eq!(
            to_ascii("\u{10810}0\u{1143a}\u{11c3a}").unwrap(),
            "xn--0-tf4im51ac5h"
        ); // Newa Mn
        assert_eq!(to_ascii("\u{10a72}\u{10af2}").unwrap(), "xn--hu9cqi"); // Manichaean punct
        assert!(to_ascii("\u{10a85}\u{10ec5}").is_err()); // Arabic Ext-C mix; Node rejects
        assert_eq!(to_ascii("\u{10877}\u{10872}").unwrap(), "xn--uf9cja"); // Palmyrene fleuron
        assert_eq!(to_ascii("\u{10b72}\u{10b79}").unwrap(), "xn--s19coa"); // Pahlavi number
        assert_eq!(to_ascii("\u{108af}\u{108f2}").unwrap(), "xn--lh9coe"); // Nabataean number
        assert_eq!(to_ascii("\u{10c32}\u{116b2}").unwrap(), "xn--969cg3p"); // Takri Mn
        assert_eq!(
            to_ascii("\u{0704}\u{135d}\u{1cdd}").unwrap(),
            "xn--xmb634dpmh"
        ); // Ethiopic Mn
        assert_eq!(to_ascii("\u{0624}\u{a9e5}").unwrap(), "xn--jgb6594f"); // Myanmar Shan Mn
        assert_eq!(
            to_ascii("\u{0624}\u{a9e5}\u{abe5}").unwrap(),
            "xn--jgb6594f5ib"
        );
        assert!(to_ascii("\u{10a85}\u{10d85}").is_err()); // Garay+ONA; Node rejects
        assert_eq!(to_ascii("\u{10810}0\u{11832}").unwrap(), "xn--0-tf4iw31b"); // Dogra Mn
        assert_eq!(to_ascii("\u{10add}\u{102e6}").unwrap(), "xn--987cr4l"); // Coptic Epact EN
        assert_eq!(to_ascii("\u{109bf}\u{1093f}").unwrap(), "xn--pl9cpi"); // Lydian mark
        assert_eq!(to_ascii("\u{10810}\u{11d3a}").unwrap(), "xn--1c9cg57a"); // Masaram Gondi Mn
        assert_eq!(to_ascii("\u{10a85}\u{10fc5}").unwrap(), "xn--0u9cs3g"); // Chorasmian number
        assert_eq!(
            to_ascii("\u{10810}\u{1183a}\u{1143a}\u{1103a}").unwrap(),
            "xn--1c9ck3l0md96e"
        ); // Brahmi Mn
        assert_eq!(to_ascii("\u{10878}\u{10cfc}").unwrap(), "xn--0f9c60f"); // Old Hungarian number
        assert_eq!(
            to_ascii("\u{10c9d}\u{10c9d}\u{11cb2}.\u{10cb2}").unwrap(),
            "xn--5c0da372j.xn--rd0d"
        ); // Marchen Mn
        assert_eq!(
            to_ascii("\u{0851}\u{aab8}\u{a678}").unwrap(),
            "xn--gwb8802fixc"
        ); // Cyr Ext Mn
        assert_eq!(
            to_ascii("\u{10810}\u{11831}\u{1163a}\u{1143a}").unwrap(),
            "xn--1c9cs1s4pbq1c"
        ); // Modi Mn
        assert_eq!(to_ascii("\u{10910}\u{1091a}").unwrap(), "xn--dk9cua"); // Phoenician number
        assert_eq!(
            to_ascii("\u{0642}\u{20dc}\u{065c}\u{065c}\u{065c}\u{065c}").unwrap(),
            "xn--ehb6caaa8181b"
        ); // Comb marks for symbols
        assert_eq!(to_ascii("\u{10a90}\u{1da9d}").unwrap(), "xn--cv9cw466a"); // SignWriting Mn
        assert_eq!(
            to_ascii("\u{10e9d}\u{10e9d}\u{11236}").unwrap(),
            "xn--yp0da77y"
        ); // Khojki Mn
        assert_eq!(
            to_ascii("\u{10810}\u{11d32}\u{10810}\u{10d32}").unwrap(),
            "xn--1c9ca650bc5t"
        ); // Hanifi Rohingya digit
        assert_eq!(
            to_ascii("\u{0851}\u{a672}\u{aa36}").unwrap(),
            "xn--gwb6702fwmc"
        ); // Cyr Me
        assert_eq!(
            to_ascii("\u{10810}\u{1183a}\u{1143a}\u{114ba}").unwrap(),
            "xn--1c9cs1s8kar9i"
        ); // Tirhuta Mn
        assert!(to_ascii("\u{10810}\u{11f3a}").is_err()); // Kawi+Cypriot; Node rejects
        assert_eq!(to_ascii("\u{1e810}\u{1e010}").unwrap(), "xn--xh4h51k"); // Comb Glagolitic
        assert_eq!(to_ascii("\u{10810}\u{115b3}").unwrap(), "xn--1c9cw6u"); // Siddham Mn
        assert_eq!(to_ascii("\u{fe73}\u{fe2f}").unwrap(), "xn--s96cte"); // Half Marks
        assert_eq!(to_ascii("\u{10e85}\u{11943}").unwrap(), "xn--9o0dk7p"); // Dives Akuru Mn
        assert!(to_ascii("\u{10810}\u{13440}").is_err()); // Egyptian hieroglyph + Cypriot; Node rejects
        assert_eq!(to_ascii("\u{10f52}\u{10f50}").unwrap(), "xn--2u0dd"); // Sogdian Mn
        assert_eq!(to_ascii("\u{0798}\u{1a18}").unwrap(), "xn--5qb685g"); // Buginese Mn
        assert_eq!(
            to_ascii("\u{0704}\u{1922}\u{1ba2}\u{1ba2}\u{1ba2}").unwrap(),
            "xn--xmb093g4ybaa"
        ); // Limbu Mn
        assert_eq!(to_ascii("\u{1e950}\u{1e810}").unwrap(), "xn--g55h5t"); // Adlam digit
        assert_eq!(to_ascii("\u{10a52}\u{11a52}").unwrap(), "xn--kt9c66y"); // Soyombo Mn
        assert_eq!(to_ascii("\u{10810}\u{11128}").unwrap(), "xn--1c9cg9m"); // Chakma Mn
        assert_eq!(
            to_ascii("\u{10810}\u{11d32}\u{10810}\u{11372}").unwrap(),
            "xn--1c9ca658fhyj"
        ); // Grantha Mn
        assert_eq!(to_ascii("\u{10e9d}\u{110b6}").unwrap(), "xn--yp0dy9a"); // Kaithi Mn
        assert!(to_ascii("\u{fe12}").is_err()); // Vertical ideographic full stop; Node rejects
        assert!(to_ascii("\u{fe52}").is_err()); // Small full stop; Node rejects
        assert_eq!(to_ascii("\u{10817}\u{111b8}").unwrap(), "xn--8c9ck8n"); // Sharada Mn
        assert_eq!(
            to_ascii("\u{10e9d}\u{114b6}\u{10e9d}\u{114b6}\u{10ead}\u{114b6}").unwrap(),
            "xn--yp0da2c889abab"
        ); // Yezidi hyphenation mark RTL
        assert_eq!(to_ascii("-\u{10d25}").unwrap(), "xn----6j6i"); // Hanifi Rohingya Mn
        assert_eq!(to_ascii("\u{10810}\u{11173}").unwrap(), "xn--1c9cg4n"); // Mahajani Mn
        assert!(to_ascii("\u{10a85}\u{10f85}").is_err()); // Old Uyghur+ONA; Node rejects
        assert_eq!(
            to_ascii("..\u{6fd}\u{d4d}\u{d4d}\u{16b35}").unwrap(),
            "..xn--qmb643aa81496a"
        ); // Pahawh Hmong Mn
        assert_eq!(to_ascii("\u{10cb2}\u{119d6}").unwrap(), "xn--rd0d43t"); // Nandinagari Mn
        assert_eq!(to_ascii("\u{1e950}\u{1e946}").unwrap(), "xn--bf6hta"); // Adlam Mn
        assert_eq!(
            to_ascii("\u{1e810}\u{1e020}\u{1ec90}\u{1e020}").unwrap(),
            "xn--ei4ha802dyxe"
        ); // Indic Siyaq number
        assert!(to_ascii("\u{10e9d}\u{11b63}").is_err()); // unassigned+Yezidi; Node rejects
        assert_eq!(
            to_ascii("\u{8b5}\u{9e2}\u{9fe}\u{9e2}\u{9e2}").unwrap(),
            "xn--bzb61daa2l"
        ); // Bengali sandhi mark
        assert_eq!(to_ascii("\u{7d1}\u{aa7c}").unwrap(), "xn--ssb0044f"); // Tai Laing tone
        assert_eq!(
            to_ascii("\u{10e9d}\u{1bc9d}\u{11c92}\u{11c92}\u{11c92}").unwrap(),
            "xn--yp0dg2vaa6695r"
        ); // Duployan Mn
        assert_eq!(to_ascii("\u{1e810}\u{1e8c7}").unwrap(), "xn--g55hol"); // Mende digit
        assert_eq!(
            to_ascii("\u{851}\u{a8e3}\u{a8e3}\u{a8c4}\u{a8e3}").unwrap(),
            "xn--gwb4623fxcaaa"
        ); // Saurashtra virama
        assert_eq!(to_ascii("\u{1ec90}\u{1e8d0}").unwrap(), "xn--xb6h53d"); // Mende combining teens
        assert_eq!(
            to_ascii("\u{1e810}\u{1e010}\u{1e8d0}\u{1e010}\u{1e010}").unwrap(),
            "xn--xh4haa303fp1a"
        );
        assert!(to_ascii("\u{1ed1e}&\u{088c}").is_err()); // Siyaq+Ext-B; Node rejects
        assert_eq!(to_ascii("\u{856}\u{a6f1}").unwrap(), "xn--lwb0232f"); // Bamum Mn
        assert_eq!(
            to_ascii("\u{856}\u{a6f1}\u{aa31}\u{aa31}\u{aa31}").unwrap(),
            "xn--lwb0232flbcaa"
        );
        assert!(to_ascii("\u{1bca0}").is_err()); // Shorthand alone → empty; Node rejects
        assert_eq!(to_ascii("\u{1bca2}tttt").unwrap(), "tttt"); // Shorthand format ignored
        assert_eq!(to_ascii("'\u{1bca2}").unwrap(), "'");
        assert_eq!(
            to_ascii("\u{10a94}\u{1d17d}\u{1da3d}").unwrap(),
            "xn--gv9cw805a9qf"
        ); // Musical + SignWriting Mn
        assert_eq!(
            to_ascii("\u{10cea}\u{1da84}\u{1d244}\u{1da84}\u{1da84}").unwrap(),
            "xn--jd0d8184a9ffbaa"
        ); // Greek musical Mn
        assert_eq!(to_ascii("\u{7dc}\u{302a}").unwrap(), "xn--3sb057r"); // Ideographic tone mark
        assert_eq!(to_ascii("-\u{1d17a}").unwrap(), "-"); // Musical format control ignored
        assert_eq!(
            to_ascii("nx\u{11a99}\u{200d}atac").unwrap(),
            "xn--nxatac-qf0ct836r"
        ); // ZWJ after virama (Soyombo)
        assert!(to_ascii("a\u{200d}b").is_err()); // ZWJ without virama
        assert!(to_ascii("\u{1ee10}1\u{088f}").is_err()); // Arabic math + Ext-B; Node rejects
        assert_eq!(to_ascii("\u{1ec90}\u{1e131}").unwrap(), "xn--6p4h35p"); // Nyiaeng Hmong Mn
        assert_eq!(to_ascii("\u{10af2}\u{16af2}").unwrap(), "xn--4x9cm25n"); // Bassa Vah Mn
        assert_eq!(to_ascii("\u{fdf8}\u{fbba}").unwrap(), "xn--ygb8adj78252a"); // Arabic symbol
        assert_eq!(to_ascii("\u{10810}\u{11ef4}").unwrap(), "xn--1c9cu40b"); // Makasar Mn
        assert_eq!(
            to_ascii("\u{7c1}\u{7c1}\u{1172b}\u{116ab}\u{116ab}\u{116ab}").unwrap(),
            "xn--csba6719raa97e"
        ); // Ahom Mn
        assert!(to_ascii("\u{10f32}\u{1cf32}").is_err()); // Znamenny+Sogdian; Node rejects
        assert_eq!(to_ascii("\u{10828}\u{10378}").unwrap(), "xn--ge8cv6f"); // Old Permic Mn
        assert_eq!(to_ascii("\u{108fd}\u{101fd}").unwrap(), "xn--m27cq7j"); // Phaistos Disc Mn
        assert!(to_ascii("\u{1e810}\u{1e6e6}").is_err()); // unassigned+Mende; Node rejects
        // Syriac + Coptic combining Ni above (Node: xn--xnb854q)
        assert_eq!(to_ascii("\u{0727}\u{2cef}").unwrap(), "xn--xnb854q");
        // Thaana + Coptic Epact thousands + Khudawadi Mn
        assert_eq!(
            to_ascii("\u{0798}\u{102e0}\u{112e4}").unwrap(),
            "xn--5qb0779jpbk"
        );
        // Manichaean + Khudawadi Mn
        assert_eq!(to_ascii("\u{10add}\u{112e6}").unwrap(), "xn--jx9c80l");
        // Kharoshthi + Khudawadi anusvara + Masaram Gondi Mn
        assert_eq!(
            to_ascii("\u{10a10}\u{112df}\u{11d3f}").unwrap(),
            "xn--or9ck4mx1i"
        );
        // NFC before CheckJoiners: n+Soyombo virama+cedilla → ņ+virama, then ZWJ OK
        assert_eq!(
            to_ascii("n\u{11a99}\u{0327}\u{200d}n").unwrap(),
            "xn--n-lla157uje6t"
        );
        assert!(to_ascii("n\u{11a99}\u{0308}\u{200d}n").is_err()); // no n+diaeresis composite
        // Math final sigma → σ (UTS #46), not NFKC's ς
        assert_eq!(to_ascii("\u{1d781}").unwrap(), "xn--4xa");
        assert_eq!(to_ascii("\u{1d6d3}").unwrap(), "xn--4xa");
        assert_eq!(to_ascii("\u{03c2}").unwrap(), "xn--3xa"); // bare final sigma kept
        // Arabic Extended-B + Ext-A (Node: xn--3xb6j)
        assert_eq!(to_ascii("\u{088b}\u{08c8}").unwrap(), "xn--3xb6j");
        // Ext-B letter + Ext-A Mn; Greek + Ext-B Mn (trailing NSM)
        assert_eq!(to_ascii("\u{088f}\u{08cf}").unwrap(), "xn--7xb2k");
        assert_eq!(to_ascii("\u{0398}\u{089f}").unwrap(), "xn--txa42w");
        assert_eq!(to_ascii("\u{088f}\u{0888}").unwrap(), "xn--0xbn"); // Ext-B Sk
        assert_eq!(to_ascii("\u{0888}\u{089f}").unwrap(), "xn--0xb0c"); // Sk + Ext-B Mn
        assert_eq!(to_ascii("\u{0888}\u{0898}").unwrap(), "xn--0xb6a");
        // Combining Cyrillic Ext letter as LTR with Mende — Node rejects; we match Node
        assert!(to_ascii("\u{1e810}\u{1e08f}\u{1e010}").is_err());
    }
}

#[cfg(test)]
mod security_fast_path {
    use super::*;

    #[test]
    fn leading_kana_voicing_mark_rejected() {
        // U+3099 sits inside the kana block but must not take SimpleIdentity
        // (would skip the leading-Mn reject).
        let domain = "\u{3099}a.com";
        assert_ne!(scan_non_ascii(domain), Ok(Scan::SimpleIdentity));
        assert!(to_ascii(domain).is_err(), "leading Mn must be rejected");
        assert!(to_ascii("\u{309A}").is_err());
    }

    #[test]
    fn kana_nfd_and_nfc_share_ace() {
        // Skipping NFC for U+3099 would encode か+゙ differently from が.
        let nfd = "\u{304b}\u{3099}"; // か + voiced mark
        let nfc = "\u{304c}"; // が
        assert_ne!(scan_non_ascii(nfd), Ok(Scan::SimpleIdentity));
        assert_eq!(to_ascii(nfd).unwrap(), to_ascii(nfc).unwrap());
    }

    #[test]
    fn turkish_and_cjk_still_simple_identity() {
        assert_eq!(scan_non_ascii("türkçe.com"), Ok(Scan::SimpleIdentity));
        assert_eq!(scan_non_ascii("你好你好.com"), Ok(Scan::SimpleIdentity));
        assert_eq!(to_ascii("türkçe.com").unwrap(), "xn--trke-2oa7j.com");
    }

    #[test]
    fn noncharacter_still_rejected() {
        assert!(to_ascii("\u{ffff}").is_err());
    }
}

#[cfg(test)]
mod audit_unassigned_simple_identity {
    use super::*;

    #[test]
    fn unassigned_in_former_simple_ranges_rejected() {
        let cps: &[u32] = &[
            0x3040, 0x3097, 0x3098, // unassigned kana
            0xD7A4, 0xD7AF, // unassigned hangul tail
        ];
        for &cp in cps {
            let ch = char::from_u32(cp).unwrap();
            let domain = format!("{ch}.com");
            assert!(
                to_ascii(&domain).is_err(),
                "U+{cp:04X} must be rejected (servo/UTS46 unassigned)"
            );
            assert_ne!(
                scan_non_ascii(&domain),
                Ok(Scan::SimpleIdentity),
                "U+{cp:04X} must not be SimpleIdentity"
            );
        }
    }
}