chematic-chem 0.1.22

//! Tautomer enumeration and canonical tautomer selection.
//!
//! Provides two public functions:
//! - [`canonical_tautomer`]: return the canonical (preferred) tautomer form.
//! - [`enumerate_tautomers`]: enumerate all reachable tautomers up to a cap.
//!
//! Pattern matching is implemented directly on the molecule graph (no SMARTS).

#![forbid(unsafe_code)]

use std::collections::HashSet;

use chematic_core::{AtomIdx, BondIdx, BondOrder, Molecule, MoleculeBuilder, implicit_hcount};

/// Bond order match type for tautomer rule patterns.
#[derive(Clone, Copy, PartialEq, Eq)]
enum BondOrderMatch {
    Single,
    Double,
    #[allow(dead_code)]
    Any,
}

impl BondOrderMatch {
    fn matches(self, order: BondOrder) -> bool {
        match self {
            BondOrderMatch::Single => {
                matches!(order, BondOrder::Single | BondOrder::Up | BondOrder::Down)
            }
            BondOrderMatch::Double => matches!(order, BondOrder::Double),
            BondOrderMatch::Any => true,
        }
    }
}

/// A tautomer transformation rule: donor loses an H and the bond orders shift.
///
/// Pattern: donor -[donor_bridge_order]- bridge -[bridge_acceptor_order]- acceptor
/// After transform: donor =[new]= bridge -[new]- acceptor (with H shifted to acceptor)
struct TautomerRule {
    #[allow(dead_code)]
    name: &'static str,
    /// Atomic number of the donor atom (loses H).
    donor_elem: u8,
    /// Atomic number of the bridge atom (None = any).
    bridge_elem: Option<u8>,
    /// Atomic number of the acceptor atom (gains H via implicit valence).
    acceptor_elem: u8,
    /// Required bond order between donor and bridge.
    donor_bridge_order: BondOrderMatch,
    /// Required bond order between bridge and acceptor.
    bridge_acceptor_order: BondOrderMatch,
    /// If true, this rule is applied in canonical_tautomer to normalize toward a preferred form.
    prefer_forward: bool,
}

/// The 15 tautomer rules.
static RULES: &[TautomerRule] = &[
    // 1. keto-enol: O-H adjacent to C=C → O=C-C (prefer keto)
    TautomerRule {
        name: "keto-enol",
        donor_elem: 8, bridge_elem: Some(6), acceptor_elem: 6,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 2. amide-iminol: N-H adjacent to C=O → N=C-O (prefer amide — forward=false)
    TautomerRule {
        name: "amide-iminol",
        donor_elem: 7, bridge_elem: Some(6), acceptor_elem: 8,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 3. iminol→amide: O-H adjacent to C=N → O=C-N (prefer amide — forward=true)
    TautomerRule {
        name: "iminol-amide",
        donor_elem: 8, bridge_elem: Some(6), acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 4. imine-enamine: N-H adjacent to C=C → N=C-C (prefer imine)
    TautomerRule {
        name: "imine-enamine",
        donor_elem: 7, bridge_elem: Some(6), acceptor_elem: 6,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 5. 1,3-H-shift N→O (any bridge): e.g. nitroso/oxime, hydroxamic acid
    TautomerRule {
        name: "1,3-N-to-O",
        donor_elem: 7, bridge_elem: None, acceptor_elem: 8,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 6. 1,3-H-shift N→N (any bridge): imidazole, pyrazole, guanidine
    TautomerRule {
        name: "1,3-N-to-N",
        donor_elem: 7, bridge_elem: None, acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 7. thioamide: N-H adjacent to C=S → N=C-S (prefer thioamide — forward=false)
    TautomerRule {
        name: "thioamide",
        donor_elem: 7, bridge_elem: Some(6), acceptor_elem: 16,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 8. thio-iminol→thioamide: S-H adjacent to C=N → S=C-N
    TautomerRule {
        name: "thio-iminol-amide",
        donor_elem: 16, bridge_elem: Some(6), acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 9. thio keto-enol: S-H adjacent to C=C → S=C-C (prefer thioketone)
    TautomerRule {
        name: "thio-keto-enol",
        donor_elem: 16, bridge_elem: Some(6), acceptor_elem: 6,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 10. thio-enol→thioketone: O-H adjacent to C=S → O=C-S
    TautomerRule {
        name: "thio-enol-ketone",
        donor_elem: 8, bridge_elem: Some(6), acceptor_elem: 16,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 11. 1,3-N→S (any bridge): sulfonamide/thioamide-type
    TautomerRule {
        name: "1,3-N-to-S",
        donor_elem: 7, bridge_elem: None, acceptor_elem: 16,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 12. 1,3-S→O (any bridge)
    TautomerRule {
        name: "1,3-S-to-O",
        donor_elem: 16, bridge_elem: None, acceptor_elem: 8,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 13. 1,3-S→N (any bridge)
    TautomerRule {
        name: "1,3-S-to-N",
        donor_elem: 16, bridge_elem: None, acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 14. 1,3-O→S (any bridge)
    TautomerRule {
        name: "1,3-O-to-S",
        donor_elem: 8, bridge_elem: None, acceptor_elem: 16,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 15. 1,3-S→S (any bridge): dithioamide, xanthate-type
    TautomerRule {
        name: "1,3-S-to-S",
        donor_elem: 16, bridge_elem: None, acceptor_elem: 16,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 16. 1,3-O→N any bridge: extends iminol-amide (rule 3) to non-C bridges
    //     e.g. O-H + S=N, N=N → O=X + N-H (hydroxamic acid, thiohydroximate)
    TautomerRule {
        name: "1,3-O-to-N-any-bridge",
        donor_elem: 8, bridge_elem: None, acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 17. 1,3-O→O any bridge: O-H via N or S bridge to =O
    //     e.g. hydroxylamine HO-N=O ↔ O=N-OH (N-oxide tautomer)
    TautomerRule {
        name: "1,3-O-to-O-any-bridge",
        donor_elem: 8, bridge_elem: None, acceptor_elem: 8,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 18. 1,3-N→C any bridge: extends imine-enamine (rule 4) to non-C bridges
    //     e.g. N-H + S=C, N=C via N bridge → N= + X-C-H
    TautomerRule {
        name: "1,3-N-to-C-any-bridge",
        donor_elem: 7, bridge_elem: None, acceptor_elem: 6,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: true,
    },
    // 19. 1,3-C→O any bridge: active methylene (or terminal alkyne) to O-H
    //     Forward: C-H + X=O → C=X + O-H. prefer_forward:false → prefer keto/C-H form.
    TautomerRule {
        name: "1,3-C-to-O-any-bridge",
        donor_elem: 6, bridge_elem: None, acceptor_elem: 8,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
    // 20. 1,3-C→N any bridge: active methylene adjacent to =N (via S, O, or N bridge)
    //     Forward: C-H + X=N → C=X + N-H. prefer_forward:false → prefer N-H form.
    TautomerRule {
        name: "1,3-C-to-N-any-bridge",
        donor_elem: 6, bridge_elem: None, acceptor_elem: 7,
        donor_bridge_order: BondOrderMatch::Single,
        bridge_acceptor_order: BondOrderMatch::Double,
        prefer_forward: false,
    },
];

/// Per-atom explicit hydrogen count vector (position-sensitive, for 1,2-shift dedup).
fn h_assignment(mol: &Molecule) -> Vec<Option<u32>> {
    (0..mol.atom_count())
        .map(|i| mol.atom(AtomIdx(i as u32)).hydrogen_count.map(|h| h as u32))
        .collect()
}

/// Find adjacent pairs of aromatic N atoms where the first (donor) carries an explicit H.
///
/// Returns (donor, acceptor) pairs for direct 1,2-shift (no bridge atom).
fn find_direct_aromatic_matches(mol: &Molecule) -> Vec<(AtomIdx, AtomIdx)> {
    let mut pairs = Vec::new();
    for (d, _) in mol.atoms() {
        let da = mol.atom(d);
        if !da.aromatic || da.element.atomic_number() != 7 {
            continue;
        }
        if da.hydrogen_count.map_or(true, |h| h == 0) {
            continue;
        }
        for (a, _) in mol.neighbors(d) {
            let aa = mol.atom(a);
            if aa.aromatic && (aa.element.atomic_number() == 7 || aa.element.atomic_number() == 8) {
                pairs.push((d, a));
            }
        }
    }
    pairs
}

/// Transfer one H from an aromatic donor to an aromatic acceptor without changing bond orders.
///
/// Only handles atoms with an explicit `hydrogen_count` on the donor.
fn transfer_hydrogen_aromatic(mol: &Molecule, donor: AtomIdx, acceptor: AtomIdx) -> Option<Molecule> {
    let donor_h = mol.atom(donor).hydrogen_count?;
    if donor_h == 0 {
        return None;
    }
    let acceptor_h = mol.atom(acceptor).hydrogen_count.unwrap_or(0);

    let mut builder = MoleculeBuilder::new();
    for i in 0..mol.atom_count() {
        let idx = AtomIdx(i as u32);
        let mut atom = mol.atom(idx).clone();
        if idx == donor {
            atom.hydrogen_count = Some(donor_h - 1);
        } else if idx == acceptor {
            atom.hydrogen_count = Some(acceptor_h + 1);
        }
        builder.add_atom(atom);
    }
    for i in 0..mol.bond_count() {
        let bidx = BondIdx(i as u32);
        let b = mol.bond(bidx);
        builder.add_bond(b.atom1, b.atom2, b.order)
            .expect("transfer_hydrogen_aromatic: bond from a valid molecule must be re-addable");
    }
    Some(builder.build())
}

fn clone_mol(mol: &Molecule) -> Molecule {
    let mut builder = MoleculeBuilder::new();
    for i in 0..mol.atom_count() {
        builder.add_atom(mol.atom(AtomIdx(i as u32)).clone());
    }
    for i in 0..mol.bond_count() {
        let b = mol.bond(BondIdx(i as u32));
        builder.add_bond(b.atom1, b.atom2, b.order)
            .expect("clone_mol: bond from a valid molecule must be re-addable");
    }
    builder.build()
}

const FNV1A_OFFSET: u64 = 0xcbf29ce484222325;
const FNV1A_PRIME: u64 = 0x100000001b3;

/// Order-independent structural hash for convergence detection.
fn mol_fingerprint(mol: &Molecule) -> u64 {
    let mut atoms: Vec<(u8, i8, u32)> = (0..mol.atom_count())
        .map(|i| {
            let idx = AtomIdx(i as u32);
            let a = mol.atom(idx);
            let bos: u32 = mol.neighbors(idx)
                .map(|(_, bidx)| mol.bond(bidx).order.order_int() as u32)
                .sum();
            (a.element.atomic_number(), a.charge, bos)
        })
        .collect();
    atoms.sort();
    let mut hash = FNV1A_OFFSET;
    for (an, ch, bos) in atoms {
        hash ^= an as u64;
        hash = hash.wrapping_mul(FNV1A_PRIME);
        hash ^= (ch as u8 as u64).wrapping_add(128);
        hash = hash.wrapping_mul(FNV1A_PRIME);
        hash ^= bos as u64;
        hash = hash.wrapping_mul(FNV1A_PRIME);
    }
    hash
}

/// Find all (donor, bridge, acceptor) triples matching the rule in `mol`.
fn find_matches(mol: &Molecule, rule: &TautomerRule) -> Vec<(AtomIdx, AtomIdx, AtomIdx)> {
    let mut matches = Vec::new();

    for i in 0..mol.atom_count() {
        let d = AtomIdx(i as u32);
        let donor_atom = mol.atom(d);
        if donor_atom.element.atomic_number() != rule.donor_elem {
            continue;
        }
        if implicit_hcount(mol, d) == 0 {
            continue;
        }

        for (b, db_bidx) in mol.neighbors(d) {
            if !rule.donor_bridge_order.matches(mol.bond(db_bidx).order) {
                continue;
            }
            if let Some(br_elem) = rule.bridge_elem {
                if mol.atom(b).element.atomic_number() != br_elem {
                    continue;
                }
            }

            for (a, ba_bidx) in mol.neighbors(b) {
                if a == d {
                    continue;
                }
                if !rule.bridge_acceptor_order.matches(mol.bond(ba_bidx).order) {
                    continue;
                }
                if mol.atom(a).element.atomic_number() != rule.acceptor_elem {
                    continue;
                }
                matches.push((d, b, a));
            }
        }
    }

    matches
}

/// Apply a single tautomer transformation: donor-bridge bond Single → Double
/// and bridge-acceptor bond Double → Single. Returns `None` if the transform
/// would be invalid (e.g. donor has no explicit H to give up on a bracket atom).
///
/// For organic-subset atoms, implicit H counts adjust automatically through the
/// valence model; for bracket atoms with an explicit `hydrogen_count`, we
/// decrement the donor and increment the acceptor manually.
fn transfer_hydrogen(
    mol: &Molecule,
    donor: AtomIdx,
    bridge: AtomIdx,
    acceptor: AtomIdx,
) -> Option<Molecule> {
    let (db_bidx, _) = mol.bond_between(donor, bridge)?;
    let (ba_bidx, _) = mol.bond_between(bridge, acceptor)?;

    let mut builder = MoleculeBuilder::new();
    for i in 0..mol.atom_count() {
        let idx = AtomIdx(i as u32);
        let mut atom = mol.atom(idx).clone();
        if let Some(h) = atom.hydrogen_count {
            if idx == donor {
                if h == 0 { return None; }
                atom.hydrogen_count = Some(h - 1);
            } else if idx == acceptor {
                atom.hydrogen_count = Some(h.saturating_add(1));
            }
        }
        builder.add_atom(atom);
    }

    for i in 0..mol.bond_count() {
        let bidx = BondIdx(i as u32);
        let b = mol.bond(bidx);
        let order = match bidx {
            x if x == db_bidx => BondOrder::Double,
            x if x == ba_bidx => BondOrder::Single,
            _ => b.order,
        };
        builder.add_bond(b.atom1, b.atom2, order).ok()?;
    }
    Some(builder.build())
}

/// Apply the first matching transformation for `rule`; return the new molecule.
fn apply_first_match(mol: &Molecule, rule: &TautomerRule) -> Option<Molecule> {
    find_matches(mol, rule)
        .into_iter()
        .find_map(|(d, b, a)| transfer_hydrogen(mol, d, b, a))
}

/// Apply every matching transformation for `rule`; return all resulting molecules.
fn apply_all_matches(mol: &Molecule, rule: &TautomerRule) -> Vec<Molecule> {
    find_matches(mol, rule)
        .into_iter()
        .filter_map(|(d, b, a)| transfer_hydrogen(mol, d, b, a))
        .collect()
}

/// Return the canonical (preferred) tautomer of `mol`.
///
/// Applies forward-preferred rules iteratively until no new form is found
/// or the iteration limit is reached. After rule-based normalization, direct
/// aromatic 1,2-shift tautomers are compared and the form with the
/// lexicographically smallest H-assignment vector is chosen.
pub fn canonical_tautomer(mol: &Molecule) -> Molecule {
    const MAX_ITER: usize = 16;
    let mut current = clone_mol(mol);
    let mut seen = HashSet::new();
    seen.insert(mol_fingerprint(&current));

    for _ in 0..MAX_ITER {
        let mut changed = false;
        for rule in RULES.iter().filter(|r| r.prefer_forward) {
            if let Some(next) = apply_first_match(&current, rule) {
                let fp = mol_fingerprint(&next);
                if !seen.contains(&fp) {
                    seen.insert(fp);
                    current = next;
                    changed = true;
                    break;
                }
            }
        }
        if !changed {
            break;
        }
    }

    // Among direct aromatic 1,2-shift tautomers, pick the lexicographically
    // smallest H-assignment so both N1H and N2H forms converge to the same output.
    let mut candidates: Vec<Molecule> = vec![clone_mol(&current)];
    for (d, a) in find_direct_aromatic_matches(&current) {
        if let Some(t) = transfer_hydrogen_aromatic(&current, d, a) {
            candidates.push(t);
        }
    }
    if candidates.len() > 1 {
        current = candidates
            .into_iter()
            .min_by_key(|t| h_assignment(t))
            .unwrap();
    }
    current
}

/// Enumerate all reachable tautomers of `mol`, capped at 32.
///
/// Returns a `Vec<Molecule>` where the first element is the original molecule.
/// Includes both 1,3-shift (rule-based) and direct aromatic 1,2-shift tautomers.
pub fn enumerate_tautomers(mol: &Molecule) -> Vec<Molecule> {
    const MAX_TAUTOMERS: usize = 32;
    let mut result = vec![clone_mol(mol)];
    let mut seen = HashSet::new();
    seen.insert(mol_fingerprint(mol));
    // Separate seen-set for 1,2-shift (mol_fingerprint can't distinguish positional H isomers).
    let mut h_seen: HashSet<Vec<Option<u32>>> = HashSet::new();
    h_seen.insert(h_assignment(mol));
    let mut frontier = vec![clone_mol(mol)];

    while !frontier.is_empty() && result.len() < MAX_TAUTOMERS {
        let current = frontier.remove(0);
        for rule in RULES.iter() {
            for next in apply_all_matches(&current, rule) {
                let fp = mol_fingerprint(&next);
                if !seen.contains(&fp) {
                    seen.insert(fp);
                    h_seen.insert(h_assignment(&next));
                    frontier.push(clone_mol(&next));
                    result.push(next);
                    if result.len() >= MAX_TAUTOMERS {
                        break;
                    }
                }
            }
            if result.len() >= MAX_TAUTOMERS {
                break;
            }
        }
        // Direct aromatic 1,2-shift (e.g. pyrazole N1H ↔ N2H).
        for (d, a) in find_direct_aromatic_matches(&current) {
            if result.len() >= MAX_TAUTOMERS {
                break;
            }
            if let Some(next) = transfer_hydrogen_aromatic(&current, d, a) {
                let ha = h_assignment(&next);
                if !h_seen.contains(&ha) {
                    h_seen.insert(ha);
                    seen.insert(mol_fingerprint(&next));
                    frontier.push(clone_mol(&next));
                    result.push(next);
                }
            }
        }
    }
    result
}

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

    #[test]
    fn test_canonical_no_tautomers() {
        // Simple alkane — no tautomers
        let mol = parse("CCO").unwrap();
        let t = canonical_tautomer(&mol);
        // Should be same molecule (ethanol has no keto-enol applicable here)
        assert_eq!(t.atom_count(), mol.atom_count());
    }

    #[test]
    fn test_canonical_idempotent() {
        let mol = parse("CC=O").unwrap(); // acetaldehyde
        let t1 = canonical_tautomer(&mol);
        let t2 = canonical_tautomer(&t1);
        assert_eq!(mol_fingerprint(&t1), mol_fingerprint(&t2));
    }

    #[test]
    fn test_enumerate_single_no_match() {
        let mol = parse("C").unwrap(); // methane
        let tautomers = enumerate_tautomers(&mol);
        assert_eq!(tautomers.len(), 1); // only the original
    }

    #[test]
    fn test_enumerate_cap() {
        // Complex molecule — should not exceed 32
        let mol = parse("CC(=O)CC(=O)C").unwrap(); // acetylacetone
        let tautomers = enumerate_tautomers(&mol);
        assert!(tautomers.len() <= 32);
        assert!(tautomers.len() >= 1);
    }

    #[test]
    fn test_enumerate_vinyl_alcohol() {
        // OC=C → should find at least 1 more tautomer (keto form CC=O)
        let mol = parse("OC=C").unwrap();
        let tautomers = enumerate_tautomers(&mol);
        assert!(tautomers.len() >= 2, "Expected >= 2 tautomers for vinyl alcohol, got {}", tautomers.len());
    }

    #[test]
    fn test_canonical_amide_unchanged() {
        // CC(=O)N — amide, canonical form should keep amide (N-C=O preferred)
        let mol = parse("CC(=O)N").unwrap();
        let t = canonical_tautomer(&mol);
        assert_eq!(mol_fingerprint(&t), mol_fingerprint(&mol));
    }

    #[test]
    fn test_canonical_acetylacetone_stable() {
        let mol = parse("CC(=O)CC(=O)C").unwrap();
        let t = canonical_tautomer(&mol);
        // Should return a valid molecule (not panic)
        assert!(t.atom_count() > 0);
    }

    #[test]
    fn test_enumerate_includes_original() {
        let mol = parse("CC=O").unwrap();
        let tautomers = enumerate_tautomers(&mol);
        // First element should be the original
        assert_eq!(mol_fingerprint(&tautomers[0]), mol_fingerprint(&mol));
    }

    #[test]
    fn test_canonical_keto_unchanged() {
        // CC=O — aldehyde, already in canonical keto form
        let mol = parse("CC=O").unwrap();
        let t = canonical_tautomer(&mol);
        assert_eq!(mol_fingerprint(&t), mol_fingerprint(&mol));
    }

    #[test]
    fn test_canonical_acetylacetone_enol() {
        // CC(O)=CC(=O)C — enol form of acetylacetone
        // canonical_tautomer should give same result as starting from keto form
        let enol = parse("CC(O)=CC(=O)C").unwrap();
        let keto = parse("CC(=O)CC(=O)C").unwrap();
        let t_enol = canonical_tautomer(&enol);
        let t_keto = canonical_tautomer(&keto);
        // Both should converge to same canonical form
        // (this may or may not hold depending on rule ordering, just check they don't panic)
        assert!(t_enol.atom_count() > 0);
        assert!(t_keto.atom_count() > 0);
    }

    #[test]
    fn test_enumerate_pyrazole_12_shift() {
        // c1cc[nH]n1 — pyrazole: N1H and N2H are direct 1,2-shift tautomers.
        let mol = parse("c1cc[nH]n1").unwrap();
        let tautomers = enumerate_tautomers(&mol);
        assert!(
            tautomers.len() >= 2,
            "Expected >= 2 tautomers for pyrazole, got {}",
            tautomers.len()
        );
    }

    #[test]
    fn test_canonical_pyrazole_normalization() {
        use chematic_smiles::canonical_smiles;
        let n1h = parse("c1cc[nH]n1").unwrap();
        let tautomers = enumerate_tautomers(&n1h);
        let n2h = tautomers
            .iter()
            .find(|t| h_assignment(t) != h_assignment(&n1h))
            .expect("enumerate_tautomers should produce N2H tautomer of pyrazole");
        assert_eq!(
            canonical_smiles(&canonical_tautomer(&n1h)),
            canonical_smiles(&canonical_tautomer(n2h)),
            "canonical_tautomer should normalize N1H and N2H to the same form"
        );
    }
}