dimvexa_core/

unit.rs

// SPDX-License-Identifier: MIT OR Apache-2.0
//! Unit definitions (static data and owned form) and resolved unit expressions.

use core::fmt;

use crate::dim::Dim;
use crate::error::{Error, Result};
use crate::expr::{Expr, is_valid_symbol};
use crate::kind::{Exactness, Kind, Shape};
use crate::prefix::PrefixMode;
use crate::ratio::Ratio;

/// A unit as it appears in a compiled data table (plain data, `const`-constructible).
///
/// Fractions are `(numerator, denominator)` pairs; the registry validates and reduces them
/// when loading.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct UnitDef {
    /// Primary symbol, for example `psi`.
    pub symbol: &'static str,
    /// Alternative symbols, for example `["\u{b0}C"]` for `degC`.
    pub aliases: &'static [&'static str],
    /// Dimension exponents of `[kg, m, s, A, K, mol, cd]`.
    pub dim: [i8; 7],
    /// Quantity kind.
    pub kind: Kind,
    /// Conversion shape to the coherent SI unit.
    pub shape: Shape,
    /// Factor `a` in `si = a * x + b`, as `(numerator, denominator)`.
    pub factor: (i128, i128),
    /// Offset `b` in `si = a * x + b`, as `(numerator, denominator)`.
    pub offset: (i128, i128),
    /// Defined or measured.
    pub exactness: Exactness,
    /// Accepted prefixes.
    pub prefixes: PrefixMode,
    /// Namespace, `si` or `std` for the core table.
    pub namespace: &'static str,
    /// Citation of the primary definition.
    pub source: &'static str,
    /// Free-text note (derivation, caveats).
    pub note: &'static str,
}

/// A bare symbol that is deliberately rejected, with the explicit symbols to use instead.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub struct AmbiguousDef {
    /// The ambiguous bare symbol, for example `ton`.
    pub symbol: &'static str,
    /// Explicit unit symbols present in the table.
    pub candidates: &'static [&'static str],
    /// Explicit symbols that are planned but not yet in the table.
    pub planned: &'static [&'static str],
    /// Why the symbol is ambiguous.
    pub note: &'static str,
}

/// An owned unit definition held by a [`crate::Registry`].
///
/// `si_value = factor * x + offset`, where `x` is the value in this unit.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct Unit {
    symbol: String,
    aliases: Vec<String>,
    dim: Dim,
    kind: Kind,
    factor: Ratio,
    offset: Ratio,
    exactness: Exactness,
    prefixes: PrefixMode,
    namespace: String,
    source: String,
    note: String,
}

impl Unit {
    /// A linear unit: `si = factor * x`, defined exactly, no prefixes, namespace `std`.
    ///
    /// # Errors
    ///
    /// [`Error::InvalidDefinition`] for a malformed symbol or a non-positive factor.
    ///
    /// # Examples
    ///
    /// ```
    /// use dimvexa_core::{Dim, Ratio, Unit};
    /// let inch = Unit::linear("in", Dim::M, Ratio::new(127, 5000)?)?;
    /// assert_eq!(inch.symbol(), "in");
    /// # Ok::<(), dimvexa_core::Error>(())
    /// ```
    pub fn linear(symbol: &str, dim: Dim, factor: Ratio) -> Result<Self> {
        Self::new(symbol, dim, Kind::Plain, factor, Ratio::ZERO)
    }

    /// A general unit with an offset (affine when `offset` is non-zero).
    ///
    /// # Errors
    ///
    /// [`Error::InvalidDefinition`] for a malformed symbol, a non-positive factor, or an
    /// offset on a unit that is not a temperature.
    pub fn new(symbol: &str, dim: Dim, kind: Kind, factor: Ratio, offset: Ratio) -> Result<Self> {
        let bad = |message: &str| Error::InvalidDefinition {
            symbol: symbol.to_string(),
            message: message.to_string(),
        };
        if !is_valid_symbol(symbol) {
            return Err(bad(
                "symbol must be non-empty and contain only letters, '_' or the degree sign",
            ));
        }
        if factor.signum() <= 0 {
            return Err(bad("factor must be positive"));
        }
        if !offset.is_zero() && kind != Kind::AbsoluteTemperature {
            return Err(bad("only absolute temperature units may have an offset"));
        }
        Ok(Self {
            symbol: symbol.to_string(),
            aliases: Vec::new(),
            dim,
            kind,
            factor,
            offset,
            exactness: Exactness::Defined,
            prefixes: PrefixMode::None,
            namespace: "std".to_string(),
            source: String::new(),
            note: String::new(),
        })
    }

    /// Validates and converts a static table row.
    ///
    /// # Errors
    ///
    /// [`Error::InvalidDefinition`] when the row is inconsistent (shape/offset mismatch,
    /// bad fraction, bad symbol or alias), or [`Error::DivisionByZero`] for a zero denominator.
    pub fn from_def(def: &UnitDef) -> Result<Self> {
        let factor = Ratio::new(def.factor.0, def.factor.1)?;
        let offset = Ratio::new(def.offset.0, def.offset.1)?;
        let mut unit = Self::new(def.symbol, Dim::new(def.dim), def.kind, factor, offset)?;
        let bad = |message: &str| Error::InvalidDefinition {
            symbol: def.symbol.to_string(),
            message: message.to_string(),
        };
        match (def.shape, offset.is_zero()) {
            (Shape::Linear, false) => return Err(bad("linear shape requires a zero offset")),
            (Shape::Affine, true) => return Err(bad("affine shape requires a non-zero offset")),
            _ => {}
        }
        if !offset.is_zero() && def.prefixes != PrefixMode::None {
            return Err(bad("affine units cannot take prefixes"));
        }
        for alias in def.aliases {
            if !is_valid_symbol(alias) {
                return Err(bad(
                    "alias contains characters that cannot appear in a symbol",
                ));
            }
            unit.aliases.push((*alias).to_string());
        }
        unit.exactness = def.exactness;
        unit.prefixes = def.prefixes;
        unit.namespace = def.namespace.to_string();
        unit.source = def.source.to_string();
        unit.note = def.note.to_string();
        Ok(unit)
    }

    /// Adds aliases (alternative symbols).
    ///
    /// # Errors
    ///
    /// [`Error::InvalidDefinition`] if an alias is not a well-formed symbol.
    pub fn with_aliases(mut self, aliases: &[&str]) -> Result<Self> {
        for a in aliases {
            if !is_valid_symbol(a) {
                return Err(Error::InvalidDefinition {
                    symbol: self.symbol,
                    message: format!("alias '{a}' is not a well-formed symbol"),
                });
            }
            self.aliases.push((*a).to_string());
        }
        Ok(self)
    }

    /// Sets the accepted prefixes (ignored for units with an offset, which never take any).
    #[must_use]
    pub const fn with_prefixes(mut self, prefixes: PrefixMode) -> Self {
        if self.offset.is_zero() {
            self.prefixes = prefixes;
        }
        self
    }

    /// Sets exactness, source and note.
    #[must_use]
    pub fn with_provenance(mut self, exactness: Exactness, source: &str, note: &str) -> Self {
        self.exactness = exactness;
        self.source = source.to_string();
        self.note = note.to_string();
        self
    }

    /// Sets the namespace.
    #[must_use]
    pub fn with_namespace(mut self, namespace: &str) -> Self {
        self.namespace = namespace.to_string();
        self
    }

    /// Primary symbol.
    #[must_use]
    pub fn symbol(&self) -> &str {
        &self.symbol
    }

    /// Alternative symbols.
    #[must_use]
    pub fn aliases(&self) -> &[String] {
        &self.aliases
    }

    /// Dimension.
    #[must_use]
    pub const fn dim(&self) -> Dim {
        self.dim
    }

    /// Quantity kind.
    #[must_use]
    pub const fn kind(&self) -> Kind {
        self.kind
    }

    /// Conversion shape.
    #[must_use]
    pub const fn shape(&self) -> Shape {
        if self.offset.is_zero() {
            Shape::Linear
        } else {
            Shape::Affine
        }
    }

    /// Factor to the coherent SI unit.
    #[must_use]
    pub const fn factor(&self) -> Ratio {
        self.factor
    }

    /// Offset to the coherent SI unit.
    #[must_use]
    pub const fn offset(&self) -> Ratio {
        self.offset
    }

    /// Defined or measured.
    #[must_use]
    pub const fn exactness(&self) -> Exactness {
        self.exactness
    }

    /// Accepted prefixes.
    #[must_use]
    pub const fn prefixes(&self) -> PrefixMode {
        self.prefixes
    }

    /// Namespace.
    #[must_use]
    pub fn namespace(&self) -> &str {
        &self.namespace
    }

    /// Citation of the definition.
    #[must_use]
    pub fn source(&self) -> &str {
        &self.source
    }

    /// Note.
    #[must_use]
    pub fn note(&self) -> &str {
        &self.note
    }
}

/// A resolved unit expression such as `km/h` or `kW.h`: dimension, kind and the exact
/// relation to the coherent SI unit, composed once.
///
/// `si_value = factor * x + offset`. Only a single (possibly prefixed) unit can have a
/// non-zero offset; compound expressions are linear.
#[derive(Debug, Clone, PartialEq, Eq)]
pub struct UnitExpr {
    expr: Expr,
    dim: Dim,
    kind: Kind,
    factor: Ratio,
    offset: Ratio,
    exactness: Exactness,
    sources: Vec<String>,
    /// Symbol of the registered temperature-difference unit of the same scale (`delta_K`
    /// for `K`), only set for a plain absolute temperature unit.
    difference: Option<String>,
}

impl UnitExpr {
    /// The unit `1` (dimensionless, factor one).
    #[must_use]
    pub const fn unity() -> Self {
        Self {
            expr: Expr::One,
            dim: Dim::NONE,
            kind: Kind::Plain,
            factor: Ratio::ONE,
            offset: Ratio::ZERO,
            exactness: Exactness::Defined,
            sources: Vec::new(),
            difference: None,
        }
    }

    pub(crate) const fn from_parts(
        expr: Expr,
        dim: Dim,
        kind: Kind,
        factor: Ratio,
        offset: Ratio,
        exactness: Exactness,
        sources: Vec<String>,
    ) -> Self {
        Self {
            expr,
            dim,
            kind,
            factor,
            offset,
            exactness,
            sources,
            difference: None,
        }
    }

    pub(crate) fn with_difference(mut self, difference: Option<String>) -> Self {
        self.difference = difference;
        self
    }

    /// Symbol of the registered temperature-difference unit that goes with this absolute
    /// temperature unit (`delta_K` for `K`, `delta_degC` for `degC`). `None` for every other
    /// unit expression, including prefixed temperatures such as `mK`.
    #[must_use]
    pub fn difference_symbol(&self) -> Option<&str> {
        self.difference.as_deref()
    }

    /// True for the plain scalar `1`.
    #[must_use]
    pub fn is_unity(&self) -> bool {
        self.dim.is_dimensionless()
            && self.kind == Kind::Plain
            && self.factor == Ratio::ONE
            && self.offset.is_zero()
    }

    /// The syntax tree this expression was resolved from.
    #[must_use]
    pub const fn expr(&self) -> &Expr {
        &self.expr
    }

    /// Dimension.
    #[must_use]
    pub const fn dim(&self) -> Dim {
        self.dim
    }

    /// Kind.
    #[must_use]
    pub const fn kind(&self) -> Kind {
        self.kind
    }

    /// Exact factor to the coherent SI unit.
    #[must_use]
    pub const fn factor(&self) -> Ratio {
        self.factor
    }

    /// Exact offset to the coherent SI unit.
    #[must_use]
    pub const fn offset(&self) -> Ratio {
        self.offset
    }

    /// Conversion shape relative to SI.
    #[must_use]
    pub const fn shape(&self) -> Shape {
        if self.offset.is_zero() {
            Shape::Linear
        } else {
            Shape::Affine
        }
    }

    /// Defined, or measured if any component is measured.
    #[must_use]
    pub const fn exactness(&self) -> Exactness {
        self.exactness
    }

    /// Distinct source citations of all components, in order of first use.
    #[must_use]
    pub fn sources(&self) -> &[String] {
        &self.sources
    }

    fn merged_sources(&self, other: &Self) -> Vec<String> {
        let mut out = self.sources.clone();
        for s in &other.sources {
            if !out.contains(s) {
                out.push(s.clone());
            }
        }
        out
    }

    fn refuse_affine(&self, other: &Self) -> Result<()> {
        for u in [self, other] {
            if !u.offset.is_zero() {
                let suggestions: Vec<String> = u.difference.iter().cloned().collect();
                let advice = suggestions.first().map_or_else(
                    || {
                        String::from(
                            "use a difference unit (delta_...) for a temperature difference",
                        )
                    },
                    |d| format!("use the difference unit {d} for a temperature difference"),
                );
                return Err(Error::Refused {
                    reason: format!(
                        "'{}' has an offset (temperature scale) and cannot be part of a compound unit; {advice}",
                        u.expr
                    ),
                    suggestions,
                });
            }
        }
        Ok(())
    }

    /// Product of two unit expressions. Multiplying by the plain scalar `1` returns the
    /// other operand unchanged.
    ///
    /// # Errors
    ///
    /// [`Error::Refused`] if an operand has an offset, [`Error::Overflow`] on overflow.
    pub fn mul(&self, rhs: &Self) -> Result<Self> {
        if self.is_unity() {
            return Ok(rhs.clone());
        }
        if rhs.is_unity() {
            return Ok(self.clone());
        }
        self.refuse_affine(rhs)?;
        Ok(Self {
            expr: Expr::Mul(Box::new(self.expr.clone()), Box::new(rhs.expr.clone())),
            dim: self.dim.checked_add(rhs.dim)?,
            kind: Kind::Plain,
            factor: self.factor.checked_mul(rhs.factor)?,
            offset: Ratio::ZERO,
            exactness: self.exactness.combine(rhs.exactness),
            sources: self.merged_sources(rhs),
            difference: None,
        })
    }

    /// Quotient of two unit expressions. Dividing by the plain scalar `1` returns `self`.
    ///
    /// # Errors
    ///
    /// [`Error::Refused`] if an operand has an offset, [`Error::Overflow`] on overflow.
    pub fn div(&self, rhs: &Self) -> Result<Self> {
        if rhs.is_unity() {
            return Ok(self.clone());
        }
        self.refuse_affine(rhs)?;
        Ok(Self {
            expr: Expr::Div(Box::new(self.expr.clone()), Box::new(rhs.expr.clone())),
            dim: self.dim.checked_sub(rhs.dim)?,
            kind: Kind::Plain,
            factor: self.factor.checked_div(rhs.factor)?,
            offset: Ratio::ZERO,
            exactness: self.exactness.combine(rhs.exactness),
            sources: self.merged_sources(rhs),
            difference: None,
        })
    }

    /// Integer power. The power 1 returns `self` unchanged (so `degC^1` stays valid), the
    /// power 0 returns the scalar `1`.
    ///
    /// # Errors
    ///
    /// [`Error::Refused`] for an offset unit with a power other than 1,
    /// [`Error::Overflow`] on overflow.
    pub fn pow(&self, exp: i32) -> Result<Self> {
        if exp == 1 {
            return Ok(self.clone());
        }
        self.refuse_affine(&Self::unity())?;
        Ok(Self {
            expr: Expr::Pow(Box::new(self.expr.clone()), exp),
            dim: self.dim.checked_scale(exp)?,
            kind: Kind::Plain,
            factor: self.factor.checked_pow(exp)?,
            offset: Ratio::ZERO,
            exactness: self.exactness,
            sources: self.sources.clone(),
            difference: None,
        })
    }
}

impl fmt::Display for UnitExpr {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        self.expr.fmt(f)
    }
}

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

    fn m() -> UnitExpr {
        UnitExpr::from_parts(
            Expr::Unit("m".into()),
            Dim::M,
            Kind::Plain,
            Ratio::ONE,
            Ratio::ZERO,
            Exactness::Defined,
            vec!["SI".into()],
        )
    }

    #[test]
    fn unity_is_identity_for_mul_and_div() {
        let u = UnitExpr::unity();
        assert_eq!(u.mul(&m()).unwrap(), m());
        assert_eq!(m().mul(&u).unwrap(), m());
        assert_eq!(m().div(&u).unwrap(), m());
        assert!(m().pow(0).unwrap().is_unity() || m().pow(0).unwrap().dim().is_dimensionless());
    }

    #[test]
    fn products_compose_exactly() {
        let sq = m().mul(&m()).unwrap();
        assert_eq!(sq.dim(), Dim::new([0, 2, 0, 0, 0, 0, 0]));
        assert_eq!(sq.to_string(), "m.m");
        let inv = UnitExpr::unity().div(&m()).unwrap();
        assert_eq!(inv.dim(), Dim::new([0, -1, 0, 0, 0, 0, 0]));
        assert_eq!(inv.sources(), ["SI"]);
    }

    #[test]
    fn rejects_bad_definitions() {
        assert!(Unit::linear("", Dim::M, Ratio::ONE).is_err());
        assert!(Unit::linear("m2", Dim::M, Ratio::ONE).is_err());
        assert!(Unit::linear("m", Dim::M, Ratio::ZERO).is_err());
        assert!(Unit::linear("m", Dim::M, Ratio::from_int(-1)).is_err());
        assert!(Unit::new("x", Dim::M, Kind::Plain, Ratio::ONE, Ratio::ONE).is_err());
        assert!(
            Unit::linear("m", Dim::M, Ratio::ONE)
                .unwrap()
                .with_aliases(&["a b"])
                .is_err()
        );
    }

    #[test]
    fn from_def_checks_shape() {
        let mut d = UnitDef {
            symbol: "x",
            aliases: &[],
            dim: [0, 1, 0, 0, 0, 0, 0],
            kind: Kind::Plain,
            shape: Shape::Linear,
            factor: (2, 4),
            offset: (0, 1),
            exactness: Exactness::Defined,
            prefixes: PrefixMode::Si,
            namespace: "std",
            source: "test",
            note: "",
        };
        assert_eq!(
            Unit::from_def(&d).unwrap().factor(),
            Ratio::new(1, 2).unwrap()
        );
        d.shape = Shape::Affine;
        assert!(Unit::from_def(&d).is_err());
        d.shape = Shape::Linear;
        d.offset = (1, 1);
        assert!(Unit::from_def(&d).is_err());
        d.factor = (1, 0);
        assert!(Unit::from_def(&d).is_err());
    }
}