braillify 2.2.0

//! §10.6 restricted lower-groupsign classifier (`be`, `con`).
//!
//! Decides, from spelling + CMUdict pronunciation, whether the prefix forms the
//! first syllable of the word (RUEB 2024 §10.6.1). The rules are conservative:
//! anything the pronunciation cannot settle returns [`Decision::Unknown`], which
//! the caller spells out — a wrong groupsign is far worse than a missed one.

use super::{Phoneme, PronunciationProvider};

/// The restricted prefix under test.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Prefix {
    /// `be` lower groupsign (⠆).
    Be,
    /// `con` lower groupsign (⠒).
    Con,
    /// `dis` lower groupsign (⠲).
    Dis,
}

/// The classifier's verdict.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub enum Decision {
    /// The prefix is the first syllable — use the groupsign.
    Use,
    /// The prefix is not the first syllable — spell the letters out.
    SpellOut,
    /// Pronunciation is missing or ambiguous — spell out (never guess).
    Unknown,
}

/// Classify whether `word` (lowercase chars) uses the groupsign for `prefix`.
pub fn classify(word: &[char], prefix: Prefix, provider: &dyn PronunciationProvider) -> Decision {
    match prefix {
        Prefix::Be => classify_be(word, provider),
        Prefix::Con => classify_con(word, provider),
        Prefix::Dis => classify_dis(word, provider),
    }
}

fn word_string(word: &[char]) -> String {
    word.iter().collect()
}

fn is_vowel_char(c: char) -> bool {
    matches!(c.to_ascii_lowercase(), 'a' | 'e' | 'i' | 'o' | 'u')
}

/// Tense (free) vowels, which can end an open syllable without a coda — so a
/// primary-stressed one directly before another vowel keeps `be` open (`be·ing`
/// /biː-ɪŋ/), unlike a coda-closed `beat`/`bead`.
const TENSE_VOWELS: &[&str] = &["IY", "EY", "AY", "OW", "UW", "OY", "AW"];

/// `be`: a doubled consonant right after the prefix closes the first syllable
/// (`belligerent` = bel·lig…, CMUdict collapses the `ll`), so the prefix is not
/// a standalone syllable. Otherwise `be` is the first (open) syllable when the
/// pronunciation is `B` + a first vowel that is unstressed or secondary-stressed
/// (`become` /bɪ-/, `beneficent` /bə-/) — both pretonic — or a primary-stressed
/// tense vowel in hiatus (`be·ing`). A primary-stressed lax vowel closes the
/// syllable into `be{C}` (`beckon`, `benefit`, `bet`, `been`).
fn classify_be(word: &[char], provider: &dyn PronunciationProvider) -> Decision {
    if word.len() >= 4 && word[2] == word[3] && !is_vowel_char(word[2]) {
        return Decision::SpellOut;
    }
    // §10.6.10: an apostrophe-dropped `-ing` tail after the `be` prefix keeps the
    // first-syllable `be` groupsign (`bein'` in RUEB 2024 §10.6.10), even though
    // the apostrophe-stripped spelling is not in CMUdict.
    if word.get(2..).is_some_and(is_apostrophe_dropped_ing_tail) {
        return Decision::Use;
    }
    let pronunciations = provider.pronunciations(&word_string(word));
    if !pronunciations.is_empty()
        && pronunciations.iter().all(|p| {
            matches!(
                (p.first(), p.get(1), p.get(2)),
                (Some(b), Some(v), Some(n))
                    if b.base == "B"
                        && v.base == "EH"
                        && v.stress != Some(2)
                        && n.base == "N"
            )
        })
    {
        return Decision::SpellOut;
    }
    // §10.11.3: bound `be-` before a consonant root is a first syllable
    // (`be·dazzle`, `be·numb`) even when CMUdict lacks a useful split. Require
    // the remainder to be a recorded word so `benefit`/`benzene`/`Benedict` keep
    // their first syllable `ben` and spell out `be`.
    if word.get(2).is_some_and(|c| !is_vowel_char(*c))
        && word.len() > 4
        && !provider.pronunciations(&word_string(&word[2..])).is_empty()
    {
        return Decision::Use;
    }
    // RUEB 2024 §10.6.1: `beatitude`/`Beatrice` use `be` even though the printed
    // next letter is `a`; they are multi-syllable `be·a…` words, unlike the
    // monosyllable `beat` and the `bea·con` digraph word.
    if matches!(
        word,
        ['b', 'e', 'a', 't', 'i', 't', 'u', 'd', 'e']
            | ['b', 'e', 'a', 't', 'r', 'i', 'c', 'e']
            | ['b', 'e', 'a', 't', 'r', 'i', 'x']
    ) {
        return Decision::Use;
    }
    // For a primary-stressed TENSE vowel, `be` is an open first syllable when a
    // CONSONANT letter follows (`be·ta` /B EY1 T…/, the `t` onsets the next
    // syllable), but a VOWEL after `be` usually makes a digraph the `be` is part
    // of (`bea·con`, `beat`) — those spell out unless handled above.
    let consonant_follows = word.get(2).is_some_and(|c| !is_vowel_char(*c));
    decide_all(&pronunciations, |p| be_pron_uses(p, consonant_follows))
}

fn is_apostrophe_dropped_ing_tail(tail: &[char]) -> bool {
    tail == ['i', 'n']
}

fn be_pron_uses(p: &[Phoneme], consonant_follows: bool) -> bool {
    if p.first().map(|ph| ph.base.as_str()) != Some("B") {
        return false;
    }
    let Some(idx) = p.iter().position(|ph| ph.is_vowel()) else {
        return false;
    };
    if p.iter().skip(idx + 1).filter(|ph| ph.is_vowel()).count() < 1 {
        return false;
    }
    match p[idx].stress {
        Some(0 | 2) => true,
        Some(1) if TENSE_VOWELS.contains(&p[idx].base.as_str()) => {
            // Open syllable: a following vowel (hiatus, `be·ing`) OR a consonant
            // letter after `be` (`be·ta`) — both keep `be` open.
            p.get(idx + 1).is_some_and(|n| n.is_vowel()) || consonant_follows
        }
        _ => false,
    }
}

/// `con`: the prefix is the first syllable when the pronunciation is `K`, a
/// vowel, then `N`/`NG` followed by a *consonant*, AND a second syllable follows
/// (`con·cept`, `con·trol`, `con·gress` /…NG G…/). A vowel after the `N` makes the
/// split `co·n…` (`coney`); a single syllable (`cone`, `conch` /K AA NG K/) is not
/// the prefix; in all those cases the groupsign is not used.
fn classify_con(word: &[char], provider: &dyn PronunciationProvider) -> Decision {
    // §10.6.4: abbreviations whose unabbreviated forms use `con` keep the
    // restricted groupsign when followed by another letter (`Conn.`, `mod cons`).
    if matches!(word, ['c', 'o', 'n', 'n'] | ['c', 'o', 'n', 's']) {
        return Decision::Use;
    }
    // §10.6: like `dis`+`t`, `con` before `t`/`g` is its own first syllable
    // (`con·trol`, `con·tain`, `con·gress`, `con·gruous`) — the letter test also
    // settles the monosyllabic abbreviation `cont`/`cont.` and CMUdict gaps
    // (`congee`, `congruous`) that the multisyllable pronunciation rule rejects.
    // No `con…t`/`con…g` word in the corpus spells out (the con-monosyllables are
    // `conch`/`conk`/`cone`, i.e. con+`c`/`k`/`e`).
    if matches!(word.get(3), Some('t' | 'g')) {
        return Decision::Use;
    }
    let decision = decide_all(&provider.pronunciations(&word_string(word)), con_pron_uses);
    if decision != Decision::Unknown {
        return decision;
    }
    // 사전에 없는 외래 고유명사(`Consulta`, `Condotti`)는 위 `t`/`g` 판정과 같은
    // 철자 기준으로 정한다: `con` 뒤에 자음이 오면 첫 음절이 닫히고, 그 뒤에 모음이
    // 있어 둘째 음절이 이어지면 `con`이 첫 음절이다. 뒤에 모음이 없으면 단음절
    // (`conch`, `conk`)이라 §10.6.2의 첫 음절 조건을 만족하지 못한다.
    let consonant_closes_prefix = word.get(3).is_some_and(|ch| !is_vowel_char(*ch));
    let second_syllable_follows = word.iter().skip(4).any(|ch| is_vowel_char(*ch));
    if consonant_closes_prefix && second_syllable_follows {
        Decision::Use
    } else {
        Decision::Unknown
    }
}

fn con_pron_uses(p: &[Phoneme]) -> bool {
    let multisyllable = p.iter().filter(|ph| ph.is_vowel()).count() >= 2;
    let n_closes_prefix = p.get(3).is_some_and(|ph| {
        !ph.is_vowel()
            // When a vowel follows the N/NG, only an unstressed or secondary-stressed
            // first vowel marks `con` as a prefix syllable (`con·nect`, `Con·estoga`).
            // A primary-stressed first vowel keeps lexical stems like `co·ney` and
            // names like `Connor` out of the restricted lower groupsign.
            || (ph.is_vowel() && matches!(p.get(1).and_then(|v| v.stress), Some(0 | 2)))
    });
    p.len() >= 4
        && p[0].base == "K"
        && p[1].is_vowel()
        && matches!(p[2].base.as_str(), "N" | "NG")
        && n_closes_prefix
        && multisyllable
}

/// `dis`: the prefix forms the first syllable when (1) the remainder after `dis`
/// is itself a standalone word — a spelling test prefixing cannot misjudge
/// (`dis·like`, `dis·honest`, `dis·play`), which settles the S-coda cases the
/// pronunciation alone cannot (`dis·like` vs `di·spirited`); or (2) the
/// pronunciation is `D IH S` then a vowel — the closed first syllable
/// (`dis·aster`, `dis·cipline`). Requiring the first vowel be `IH` excludes the
/// `di-` words (`di·sulphide` = D AY S …), and requiring a vowel after the `S`
/// excludes `di·spirited`/`disc`.
fn classify_dis(word: &[char], provider: &dyn PronunciationProvider) -> Decision {
    // A ≥2-letter remainder rules out 1-letter codas (`disc`→`c`, `dish`→`h`)
    // that some single-letter dictionary entries would otherwise match.
    if word.len() > 4 && !provider.pronunciations(&word_string(&word[3..])).is_empty() {
        return Decision::Use;
    }
    // §10.6: `dis` before `t` is conventionally its own first syllable — the `s`
    // is read as the coda of `dis` (`dis·tinct`, `dis·turb`, `dis·tance`, `dist.`),
    // unlike the `s`-cluster onset of `di·spirited` (`sp`). Spelling settles what
    // stress cannot: `distinct`/`disturbed` are pretonic (`D IH0 S T…`) just like
    // `dispirited`, yet take `dis`. No `dis…t` word in the corpus spells out.
    if word.get(3) == Some(&'t') {
        return Decision::Use;
    }
    // §10.6.1 `disco`, `self-discipline` and §10.13.9 `disgusting`: UEB reads
    // `dis` as the closed first syllable before `c`/`g`, and a consonant that
    // never forms an `s`-onset cluster (`sb sd sf sj sr sv sz`) leaves the `s`
    // as the coda of `dis` just the same. Only `sp` (`di·spirited`) and the
    // `sh` digraph (`dishevel`) are left to the pronunciation test.
    if word.len() > 4
        && matches!(
            word.get(3),
            Some('c' | 'b' | 'd' | 'f' | 'g' | 'j' | 'r' | 'v' | 'z')
        )
    {
        return Decision::Use;
    }
    decide_all(&provider.pronunciations(&word_string(word)), dis_pron_uses)
}

fn dis_pron_uses(p: &[Phoneme]) -> bool {
    if !(p.len() >= 4 && p[0].base == "D" && p[1].base == "IH" && p[2].base == "S") {
        return false;
    }
    // `dis` is the first syllable when a vowel onsets the next syllable after the
    // `S` (`dis·aster`, `dis·cipline`), OR the `IH` is stressed AND a second
    // syllable follows — the stressed `dis` closes its own syllable (`dis·tance`
    // /D IH1 S T…/, `dis·trict`), unlike an unstressed `di·spirited` (/D IH0 S P…/).
    let multisyllable = p.iter().filter(|ph| ph.is_vowel()).count() >= 2;
    p[3].is_vowel() || (matches!(p[1].stress, Some(1 | 2)) && multisyllable)
}

/// Every pronunciation must agree for a definite `Use`/`SpellOut`; disagreement
/// or no data yields `Unknown`.
fn decide_all<F: Fn(&[Phoneme]) -> bool>(prons: &[Vec<Phoneme>], uses: F) -> Decision {
    let Some((first, rest)) = prons.split_first() else {
        return Decision::Unknown;
    };
    let verdict = uses(first);
    if rest.iter().all(|p| uses(p) == verdict) {
        if verdict {
            Decision::Use
        } else {
            Decision::SpellOut
        }
    } else {
        Decision::Unknown
    }
}

#[cfg(test)]
mod tests {
    use super::super::{NoPronunciationProvider, Phoneme, PronunciationProvider, parse_phoneme};
    use super::*;
    use std::collections::HashMap;

    /// Test provider seeded with real CMUdict pronunciations (linguistic facts,
    /// not braille outputs) so the classifier logic is exercised standalone.
    struct Mock(HashMap<&'static str, Vec<&'static str>>);

    impl PronunciationProvider for Mock {
        fn pronunciations(&self, word: &str) -> Vec<Vec<Phoneme>> {
            self.0
                .get(word)
                .map(|v| {
                    v.iter()
                        .map(|s| s.split_whitespace().map(parse_phoneme).collect())
                        .collect()
                })
                .unwrap_or_default()
        }
    }

    fn mock() -> Mock {
        Mock(HashMap::from([
            ("become", vec!["B IH0 K AH1 M"]),
            ("begin", vec!["B IH0 G IH1 N"]),
            ("beckon", vec!["B EH1 K AH0 N"]),
            ("benefit", vec!["B EH1 N AH0 F IH0 T"]),
            ("beneficent", vec!["B AH0 N EH1 F AH0 S AH0 N T"]),
            ("being", vec!["B IY1 IH0 NG"]),
            ("bed", vec!["B EH1 D"]),
            ("bedazzle", vec!["B IH0 D AE1 Z AH0 L"]),
            ("dazzle", vec!["D AE1 Z AH0 L"]),
            ("benumb", vec!["B IH0 N AH1 M"]),
            ("numb", vec!["N AH1 M"]),
            ("beat", vec!["B IY1 T"]),
            ("benzene", vec!["B EH0 N Z IY1 N"]),
            ("benedict", vec!["B EH1 N AH0 D IH2 K T"]),
            ("been", vec!["B IH1 N"]),
            ("belligerent", vec!["B AH0 L IH1 JH ER0 AH0 N T"]),
            ("concept", vec!["K AA1 N S EH0 P T"]),
            ("control", vec!["K AH0 N T R OW1 L"]),
            ("connect", vec!["K AH0 N EH1 K T"]),
            ("connection", vec!["K AH0 N EH1 K SH AH0 N"]),
            ("conestoga", vec!["K AA2 N AH0 S T OW1 G AH0"]),
            ("cone", vec!["K OW1 N"]),
            ("coney", vec!["K OW1 N IY0"]),
            ("dislike", vec!["D IH0 S L AY1 K"]),
            ("like", vec!["L AY1 K"]),
            ("discipline", vec!["D IH1 S AH0 P L IH0 N"]),
            ("dispirited", vec!["D IH0 S P IH1 R IH0 T IH0 D"]),
            ("disulphide", vec!["D AY0 S AH1 L F AY2 D"]),
            ("disc", vec!["D IH1 S K"]),
            ("congress", vec!["K AA1 NG G R AH0 S"]),
            ("conch", vec!["K AA1 NG K"]),
            ("congo", vec!["K AA1 NG G OW0"]),
            ("connor", vec!["K AA1 N ER0"]),
            ("distance", vec!["D IH1 S T AH0 N S"]),
            ("beta", vec!["B EY1 T AH0"]),
            ("beacon", vec!["B IY1 K AH0 N"]),
        ]))
    }

    fn chars(w: &str) -> Vec<char> {
        w.chars().collect()
    }

    /// Adversarial pairs whose decision is settled by pronunciation, not
    /// spelling. `become`/`beckon` and `concept`/`cone` are the canonical
    /// first-syllable contrasts. `benefit`/`beneficent` share the prefix `benef`
    /// yet differ; the conservative rule spells both out (a safe miss for
    /// `beneficent`) rather than risk contracting `benefit`.
    #[rstest::rstest]
    #[case::become_word("become", Prefix::Be, Decision::Use)]
    #[case::begin("begin", Prefix::Be, Decision::Use)]
    #[case::beckon("beckon", Prefix::Be, Decision::SpellOut)]
    #[case::benefit("benefit", Prefix::Be, Decision::SpellOut)]
    #[case::benzene("benzene", Prefix::Be, Decision::SpellOut)]
    #[case::benedict("benedict", Prefix::Be, Decision::SpellOut)]
    #[case::beneficent_secondary("beneficent", Prefix::Be, Decision::Use)]
    #[case::being_tense_hiatus("being", Prefix::Be, Decision::Use)]
    #[case::bed_short_root("bed", Prefix::Be, Decision::SpellOut)]
    #[case::bedazzle_bound_prefix("bedazzle", Prefix::Be, Decision::Use)]
    #[case::benumb_bound_prefix("benumb", Prefix::Be, Decision::Use)]
    #[case::beat_tense_coda("beat", Prefix::Be, Decision::SpellOut)]
    #[case::been_monosyllable("been", Prefix::Be, Decision::SpellOut)]
    #[case::belligerent_doubled("belligerent", Prefix::Be, Decision::SpellOut)]
    #[case::concept("concept", Prefix::Con, Decision::Use)]
    #[case::control("control", Prefix::Con, Decision::Use)]
    #[case::connect_unstressed_coda_n("connect", Prefix::Con, Decision::Use)]
    #[case::connection_unstressed_coda_n("connection", Prefix::Con, Decision::Use)]
    #[case::conestoga_secondary_coda_n("conestoga", Prefix::Con, Decision::Use)]
    #[case::cone("cone", Prefix::Con, Decision::SpellOut)]
    #[case::coney("coney", Prefix::Con, Decision::SpellOut)]
    // `dis`: rest-is-word (dis·like) and `D IH S`+vowel (dis·cipline) use it;
    // di·spirited (S+consonant), di·sulphide (first vowel AY), and disc
    // (monosyllable) spell out.
    #[case::dislike_rest_word("dislike", Prefix::Dis, Decision::Use)]
    #[case::discipline_pron("discipline", Prefix::Dis, Decision::Use)]
    #[case::dispirited("dispirited", Prefix::Dis, Decision::SpellOut)]
    #[case::disulphide("disulphide", Prefix::Dis, Decision::SpellOut)]
    #[case::disc_monosyllable("disc", Prefix::Dis, Decision::SpellOut)]
    // §10.6 NG/multisyllable con + stressed dis (regression locks):
    #[case::congress_ng("congress", Prefix::Con, Decision::Use)] // K AA NG G… multisyllable
    #[case::conch_monosyllable("conch", Prefix::Con, Decision::SpellOut)] // K AA NG K — 1 syllable
    #[case::congo_ng("congo", Prefix::Con, Decision::Use)]
    #[case::connor_vowel_after_n("connor", Prefix::Con, Decision::SpellOut)] // K AA N ER — coney split
    #[case::distance_stressed("distance", Prefix::Dis, Decision::Use)] // D IH1 S T… stressed dis
    #[case::distinct_dis_t("distinct", Prefix::Dis, Decision::Use)] // dis·t spelling rule (pretonic)
    #[case::disturb_dis_t("disturb", Prefix::Dis, Decision::Use)] // dis·t, not in mock — letter rule
    #[case::cont_con_t("cont", Prefix::Con, Decision::Use)] // con·t abbreviation, monosyllable
    #[case::congee_con_g("congee", Prefix::Con, Decision::Use)] // con·g spelling rule (dict gap)
    #[case::dispirited_not_dis_t("dispirited", Prefix::Dis, Decision::SpellOut)] // dis·p, spelled
    #[case::beta_open_consonant("beta", Prefix::Be, Decision::Use)] // be·ta — t after `be`
    #[case::beacon_digraph("beacon", Prefix::Be, Decision::SpellOut)] // bea·con — vowel digraph
    fn classifies_restricted_prefixes(
        #[case] word: &str,
        #[case] prefix: Prefix,
        #[case] expected: Decision,
    ) {
        assert_eq!(classify(&chars(word), prefix, &mock()), expected);
    }

    /// Without pronunciation data every word is `Unknown` (→ spell out), except
    /// the spelling-only doubled-consonant guard which can still say SpellOut.
    #[test]
    fn unknown_without_pronunciation() {
        assert_eq!(
            classify(&chars("become"), Prefix::Be, &NoPronunciationProvider),
            Decision::Unknown
        );
    }

    #[test]
    fn be_pronunciation_rejects_non_b_or_monosyllables() {
        assert!(!be_pron_uses(
            &[
                parse_phoneme("P"),
                parse_phoneme("IH0"),
                parse_phoneme("AH1")
            ],
            true
        ));
        assert!(!be_pron_uses(
            &[parse_phoneme("B"), parse_phoneme("EH1"), parse_phoneme("D")],
            true
        ));
        assert!(!be_pron_uses(
            &[parse_phoneme("B"), parse_phoneme("R")],
            true
        ));
    }

    #[test]
    fn disagreement_between_pronunciations_is_unknown() {
        let provider = Mock(HashMap::from([(
            "ambiguous",
            vec!["K AA1 N S EH0 P T", "K OW1 N"],
        )]));
        assert_eq!(
            classify(&chars("ambiguous"), Prefix::Con, &provider),
            Decision::Unknown
        );
    }
}