Skip to content

kgram/equation-resolver

Repository files navigation

equation-resolver – Resolve equations ASTs

Resolves an AST from equation-parser into a number or matrix, optionally with a unit.

Installation

npm install -S equation-resolver

or

yarn add equation-resolver

API

resolve(node: EquationNode | EquationParserError, options: ResolveOptions): ResultNode | ResultResolveError

Resolve an EquationNode (or EquationParserError) to a ResultNode (or ResultResolveError).

Options functions and variables that should be available should be added as part of options.

Example:

resolve(parse('1+2*3+2/4'))
// returns {
//   type: 'number',
//   value: 7.5
// }

format(equation: EquationNode | EquationParserError, unit: EquationNode | EquationParserError | null, options: FormatOptions): EquationNode | EquationParserError | EquationResolveError

Resolve an EquationNode (or EquationParserError), wraps it in an equals-node and adds the result as an EquationNode.

Options functions and variables that should be available should be added as part of options.

If a unit is passed, the result will be in the unit, with any unrepresented unit multiplied, e.g. if unit is m^2 and result is m^2, format uses m^2. If result is m^2*kg/s^2, format uses m^2 * (kg/s^2).

Option simplifiableUnits should be an array of variable-names that units can be represented as instead, e.g. showing N instead of m*kg/s^2. Only used when not passing a unit. Defaults to these:

  • N (Newton, m * kg / s^2)
  • J (Joule, m^2 * kg / s^2)
  • W (Watt, m^2 * kg / s^3)
  • Pa (Pascal, kg / m / s^2)
  • Hz (Hertz, 1 / s)
  • lx (lux, cd / m^2)
  • C (coulomb, A * s)
  • V (volt, m^2 * kg / A / s^3)
  • F (farad, s^4 * A^2 / m^2 / kg)
  • Ω (ohm, m^2 * kg / A^2 / s^3)
  • S (siemens, s^3 * A^2 / m^2 / kg)
  • Wb (weber, m^2 * kg / A / s^2)
  • T (tesla, kg / A / s^2)
  • H (henry, m^2 * kg / A^2 / s^2)
  • Gy (gray, m^2 / s^2)

Errors are automatically wrapped in equals with a placeholder-result and the appropriate EquationParserError or EquationResolveError, which allows highlighting the problematic part of the equation.

Example:

format(parse('1+2*3+2/4'))
// returns {
//     type: 'equals',
//     a: {...},
//     b: { type: 'number', value: '7.5' },
// }

format(parse('2m * 3m'), null, { variables: defaultVariables })
// returns {
//     type: 'equals',
//     a: {...},
//     b: {
//         type: 'multiply-dot',
//         a: { type: 'number', value: '6' },
//         b: {
//             type: 'power',
//             a: { type: 'variable', name: 'm' },
//             b: { type: 'number', value: '2' },
//         },
//     },
// }

defaultFunctions

A lookup of common mathmatical functions to be passed as part of options.

  • sin
  • cos
  • tan
  • asin
  • acos
  • atan
  • atan2
  • abs
  • ceil
  • floor
  • round (first argument is number to round, second is amount of decimals defaulting to 0)
  • max (arbitrary amount of arguments)
  • min (arbitrary amount of arguments)
  • pow
  • sqrt
  • root (first argument is radicand)
  • ln
  • log (second argument is base, defaulting to 10)
  • sum (first argument is variable name, second is start, third is end, fourth is expression) Example: sum(n, 1, 5, n^2) is 55

The functions can be freely renamed by simply assigning them to a new name, as such:

const customFunctions = {
    squareroot: defaultFunctions.sqrt,
}

The new name will be passed along in errors.

defaultVariables

A lookup of common units and numbers to be referenced in equations.

Currently, all units and significant numbers have been attempted to be included as part of defaultVariables. This does however mean a lot of probably useless units are included (petabecquerel are rarely used by most people). In the future, this can hopefully be split into some useful sets.

To see the values included, check the ./src/defaultVariables-file. Additions are welcome.

createResolverFunction(argNames: string[], expression: EquationNode, options: ResolveOptions)

Create a function for future resolving from an EquationNode.

Options functions and variables that should be available should be added as part of options. Note that these may be different from the variables and functions available when the function is invoked.

This is primarily a tool to let users define their own functions.

Example:

// Parse function expression
const node = parse('f(a, b) = a^2 + b')

// Check that expression can be made into a resolver function
if (node.type !== 'equals' || node.a.type !== 'function' || node.a.args.every((arg) => arg.type === 'variable')) {
    throw new Error('Expected equals with a function with only variable-args as left-hand side')
}

// Create the function
const func = createResolverFunction(node.a.args.map((arg) => arg.name), node.b)

resolve(parse('f(2,3)'), {
    functions: {
        // Assign the function as the correct name, `f`
        [node.a.name]: func
    }
})
// returns {
//     type: 'number',
//     value: 7,
// }

AST – ResultNode

After resolving, a ResultNode is returned. The structure is similar to the EquationNode from equation-parser, plain objects with a type property to distinguish them.

The TS typing enforces the correct ordering of the nodes.

'number'ResultNodeNumber

Represents a plain number

Additional values:

  • value: number

'matrix'ResultNodeMatrix

Represents a matrix result. The values can only be numbers.

Additional values:

  • n: number
  • m: number
  • values: ResultNodeNumber[][]

'unit'ResultNodeUnit

Represents a matrix or number with a unit. The units are represented as an object with units as keys and numbers as values, e.g. acceleration is { m: 2, s: -1 }.

Additional values:

  • units: UnitLookup
  • value: ResultNodeMatrix | ResultNodeNumber

'resolver-error'ResultResolveError

Represents an error during resolution. Not technically a ResultNode.

The type of error is represented by the errorType-value, taking one of the following:

  • functionUnknown: The name of the function is not included in the function-lookup.

  • functionArgLength: Wrong number of arguments.

  • functionNumberOnly: Argument was a matrix or has a unit, where only plain numbers are supported.

  • functionSqrt1Positive: Function sqrt must have a positive number as first argument.

  • functionRoot1PositiveInteger: Function root must have a positive integer as first argument.

  • functionRoot2Positive: Function root must have a positive number as second arguement.

  • functionSum1Variable: Function sum must have a variable name as first arguement.

  • functionSum2Integer: Function sum must have an integer as second argument.

  • functionSum3Integer: Function sum must have an integer as third argument.

  • variableUnknown: The name of the variable is not included in the variable lookup.

  • plusDifferentUnits: Cannot add numbers with different units.

  • plusMatrixMismatch: Cannot add matrices of wrong dimensions.

  • plusminusUnhandled: No implementation of plus/minus is included.

  • scalarProductUnbalanced: Scalar (dot) product requires equal-sized vectors.

  • vectorProduct3VectorOnly: Vector (cross) product requires vectors of size 3.

  • matrixProductMatrixMismatch: Cannot multiply matrices of wrong dimensions.

  • multiplyImplicitNoVectors: Cannot implicitly multiply two vectors (wouldn't know if scalar or vector product is intended)

  • divideNotZero: Cannot divide by zero.

  • divideMatrixMatrix: Cannot divide two matrices.

  • powerUnitlessNumberExponent: Exponent can only be a plain number.

  • operatorInvalidArguments: Operator has invalid arguments. This should generally be handled by specific errors, such as divideMatrixMatrix, ut is include for safety.

  • noComparison: Cannot resolve a comparison.

  • matrixDifferentUnits: All cells in a matrix must have the same unit.

  • matrixNoNesting: A matrix-cell cannot be another matrix.

  • invalidEquation: Tried to resolve a parser-error.

  • placeholder: Cannot resolve anything including a placeholder.

  • invalidUnit: Only used by the format-function. The provided unit argument was not valid as a unit.

Known limitations

Plus-minus operators don't work, since they would require the addition of a ResultNodeSet-type.

Default math-functions (except for sum, which is a special case) are only implemented for unitless numbers.