Dual numbers represented as a one-dimensional degenerate geometric algebra.
This module implements the dual number algebra:
D = R[ε] / (ε²)using the metric:
{0}The basis element e1 represents the infinitesimal unit:
ε² = 0A dual number:
a + bεconsists of:
a- the real componentb- the infinitesimal component
Dual numbers are useful for automatic differentiation because the infinitesimal coefficient propagates derivatives through arithmetic operations and elementary functions.
Summary
Functions
Adds two multivectors component-wise.
Checks whether a multivector is a blade.
Returns the number of coefficients stored by the algebra.
Returns the mapping between blade names and storage indices.
Returns the canonical sign of a multivector.
Returns the coefficient of a basis blade.
Computes the dual conjugate.
Computes the cosine function.
Extracts the derivative component.
Returns the dimension of the algebra.
Constructs a dual number.
Returns the infinitesimal unit.
Computes the exponential function.
Formats this multivector as geometric algebra notation.
Computes the geometric product of two multivectors.
Returns the grade of a homogeneous multivector.
Extracts the grade-g component of a multivector.
Checks whether a multivector contains components of the given grade only.
Returns the grades present in a multivector.
Computes the Hodge dual of a multivector.
Computes the inverse Hodge dual of a multivector.
Extracts the infinitesimal component.
Computes the inner product of two multivectors.
Computes the multiplicative inverse.
Computes the inverse of a multivector.
Computes the inverse of a multivector.
A guard that checks if a floating point number is below the epsilon threshold.
Checks whether a multivector contains components of the given grade only.
A guard that matches multi vectors that are all zero.
Computes the left contraction of two multivectors.
Computes the natural logarithm.
Returns the maximum absolute coefficient of a multivector.
Returns the metric of the algebra.
Returns the norm of a multivector.
Returns the normalized (real) value of a dual number.
Normalizes a multivector.
Returns the multiplicative identity
Returns the pseudoscalar
Extracts the real component.
Applies the reverse operation to a multivector.
Computes the right contraction of two multivectors.
Computes the rotor that maps one frame to another.
Turns a Multivector with only a scalar part into an elixir number.
Checks whether a multivector contains only a scalar component.
Returns the scalar coefficient of a multivector.
Computes the scalar product of two dual multivectors.
Creates a multivector from a string representation.
Computes the sine function.
Computes the square root.
Returns the squared norm of a multivector.
Subtracts two multivectors component-wise.
Returns the multiplication table for the algebra.
Turns a Multivector with only a scalar part into an elixir number.
Computes the outer product (wedge product) of two multivectors.
Computes the outer product of a list of multivectors.
Returns the zero dual number
Checks whether all coefficients of a multivector are zero.
Functions
Adds two multivectors component-wise.
Examples
iex> a = new(scalar: 2)
iex> b = new(scalar: 3)
iex> add(a, b)
new(scalar: 5)
Checks whether a multivector is a blade.
A blade is a multivector containing components from at most one grade.
Scalars are considered blades.
Examples
iex> blade?(new(e1: 2))
true
iex> blade?(new(scalar: 2))
true
iex> blade?(new(scalar: 2, e1: 1))
false
Returns the number of coefficients stored by the algebra.
A dimension n algebra contains 2^n basis blades.
Returns the mapping between blade names and storage indices.
Blade coefficients are stored in a fixed-blade_count tuple. This map translates canonical blade names into their corresponding tuple index.
Example
iex> blade_indices()[:e1]
1
Returns the canonical sign of a multivector.
The canonical sign is determined by the first non-zero coefficient in storage order.
Returns:
1if the first non-zero coefficient is positive-1if the first non-zero coefficient is negative0if all coefficients are zero
Examples
iex> canonical_sign(new(e1: 2))
1
iex> canonical_sign(new(e1: -2))
-1
iex> canonical_sign(new())
0
Returns the coefficient of a basis blade.
The requested blade can be given in canonical form or as any registered blade alias. Aliases are automatically converted to the canonical blade and the appropriate sign is applied.
Examples
iex> coefficient(
...> new(e1: 3),
...> :e1
...> )
3.0
Computes the dual conjugate.
For:
x = a + bεreturns:
x* = a - bε
Computes the cosine function.
The infinitesimal component contains the derivative:
d/dx cos(x) = -sin(x)
Extracts the derivative component.
Equivalent to returning the infinitesimal coefficient.
Returns the dimension of the algebra.
This is the number of basis vectors defined by the metric.
Constructs a dual number.
Creates:
real + infinitesimal*ε
Returns the infinitesimal unit.
The infinitesimal satisfies:
ε² = 0
Computes the exponential function.
For:
x = a + bεreturns:
exp(x) = exp(a) + b exp(a) ε
Formats this multivector as geometric algebra notation.
This is the same representation used by Inspect.
Examples
iex> inspect(new())
"~G[0.0]"
iex> inspect(new(), custom_options: [all_blades: true])
"~G[0.0 + 0.0e1]"
iex> inspect(new(scalar: -2))
"~G[-2.0]"
iex> inspect(new(scalar: 2))
"~G[2.0]"
iex> inspect(new(scalar: -2))
"~G[-2.0]"
iex> inspect(new(e1: 1))
"~G[1.0e1]"
iex> inspect(new(scalar: 1, e1: 2))
"~G[1.0 + 2.0e1]"
iex> inspect(new(scalar: 1, e1: -2))
"~G[1.0 - 2.0e1]"
iex> inspect(new(scalar: -1, e1: -2))
"~G[-1.0 - 2.0e1]"
iex> inspect(new(scalar: -1, e1: 2))
"~G[-1.0 + 2.0e1]"
iex> inspect(new(e1: 2))
"~G[2.0e1]"
iex> inspect(new(e1: -2))
"~G[-2.0e1]"
iex> inspect(~G[ε + εe1])
"~G[ε + εe1]"
iex> inspect(~G[-ε - εe1])
"~G[-ε - εe1]"
Computes the geometric product of two multivectors.
The geometric product is the fundamental multiplication operation of geometric algebra. It combines the outer product and metric-dependent inner product into a single associative operation.
The result depends on the algebra's metric.
Examples
iex> gp(
...> new(e1: 1),
...> new(e1: 1)
...> )
new(scalar: 0)
Returns the grade of a homogeneous multivector.
The zero multivector is considered grade 0.
Returns nil for mixed-grade multivectors.
iex> grade( ...> new(scalar: 1) ...> ) 0
iex> grade( ...> new(e1: 2) ...> ) 1
iex> grade( ...> new(scalar: 1, e1: 2) ...> ) nil
iex> grade( ...> new() ...> ) nil
Extracts the grade-g component of a multivector.
All coefficients whose basis blades are not of grade g are set to zero.
Raises ArgumentError if g is outside the range 0..dimension().
Examples
iex> grade(
...> new(scalar: 1, e1: 2),
...> 1
...> )
new(e1: 2)
iex> grade(
...> new(scalar: 1, e1: 2),
...> 0
...> )
new(scalar: 1)
Checks whether a multivector contains components of the given grade only.
A multivector is considered to have a grade if all non-zero components belong to that grade. The zero multivector is considered to have grade 0.
Examples
iex> grade?(new(e1: 1), 1)
true
iex> grade?(new(scalar: 1, e1: 2), 2)
false
iex> grade?(new(scalar: 1), 0)
true
iex> grade?(new(), 0)
trueSee is_grade/2
Returns the grades present in a multivector.
The returned list contains every grade with at least one non-zero coefficient, ordered from lowest to highest.
Examples
iex> grades(
...> new(scalar: 1)
...> )
[0]
iex> grades(
...> new(e1: 2)
...> )
[1]
iex> grades(
...> new(scalar: 1, e1: 2)
...> )
[0, 1]
iex> grades(
...> new()
...> )
[]
Computes the Hodge dual of a multivector.
The Hodge dual maps each basis blade to its complementary blade with the appropriate orientation sign. The complementary blade is determined by the pseudoscalar of the algebra.
The operation is linear and is applied independently to each basis blade and its coefficient.
Unlike the pseudoscalar dual, the Hodge dual does not require the pseudoscalar to be invertible and is therefore also defined for degenerate algebras.
The operations are inverses:
hodge_undual(hodge_dual(a)) == a
hodge_dual(hodge_undual(a)) == aSee hodge_undual/1 for the inverse operation.
Examples
iex> hodge_dual(new(e1: 1)) |> inspect
new(scalar: 1.0) |> inspect
Computes the inverse Hodge dual of a multivector.
hodge_undual/1 reverses the blade-complement operation performed by
hodge_dual/1, mapping each basis blade back to its complementary blade
with the appropriate orientation sign.
The operation is linear and is applied independently to each basis blade and its coefficient.
Unlike the pseudoscalar dual, the Hodge dual and its inverse are defined for degenerate algebras as well.
The operations are inverses:
hodge_undual(hodge_dual(a)) == a
hodge_dual(hodge_undual(a)) == aSee hodge_dual/1 for the inverse operation.
Examples
iex> hodge_undual(hodge_dual(new(e1: 2)))
new(e1: 2)
Extracts the infinitesimal component.
This corresponds to the derivative component when using dual numbers for automatic differentiation.
Computes the inner product of two multivectors.
The operation is generated from the geometric product and retains only terms satisfying the grade selection rule.
Example
iex> inner(
...> new(e1: 2),
...> new(e1: 3)
...> )
Computes the multiplicative inverse.
For:
x = a + bεwhere a != 0:
x⁻¹ = 1/a - b/a² ε
Computes the inverse of a multivector.
The inverse is computed using the reverse:
inverse!(a) = reverse(a) / scalar_part(a * reverse(a))
This formula is valid when a * reverse(a) is a non-zero scalar.
Returns either {:ok, result} if inverse can be compute or :error otherwise
## Examples
iex> inverse(
...> new(scalar: 2)
...> )
{:ok, new(scalar: 0.5)}
iex> inverse(new(e1: 1))
...> :error
Computes the inverse of a multivector.
The inverse is computed using the reverse:
inverse!(a) = reverse(a) / scalar_part(a * reverse(a))
This formula is valid when a * reverse(a) is a non-zero scalar.
Raises ArgumentError if the multivector is not invertible by this formula.
## Examples
iex> inverse!(
...> new(scalar: 2)
...> )
new(scalar: 0.5)
iex> assert_raise ArgumentError, fn ->
...> inverse!(new(e1: 1))
...> end
A guard that checks if a floating point number is below the epsilon threshold.
Can be used in function guards:
def foo(x) when is_near_zero(x), do: x
Checks whether a multivector contains components of the given grade only.
A multivector is considered to have a grade if all non-zero components belong to that grade. The zero multivector is considered to have grade 0.
The is_grade(vector, grade) guard can be used in function guards:
def foo(mv) when is_grade(mv, 1) do
mv
endExamples
iex> grade?(new(e1: 1), 1)
true
iex> grade?(new(scalar: 1, e1: 2), 2)
false
iex> grade?(new(scalar: 1), 0)
true
iex> grade?(new(), 0)
trueSee grade?/2
A guard that matches multi vectors that are all zero.
Can be used in function guards:
def foo(x) when is_zero(x), do: x
Computes the left contraction of two multivectors.
The operation is generated from the geometric product and retains only terms satisfying the grade selection rule.
Example
iex> left_contraction(
...> new(e1: 2),
...> new(e1: 3)
...> )
Computes the natural logarithm.
For:
x = a + bεreturns:
log(x) = log(a) + b/a ε
Returns the maximum absolute coefficient of a multivector.
Accepts either a multivector struct or the internal coefficient tuple.
Example
iex> max_abs_component(new(e1: 2, scalar: 5))
5.0
iex> max_abs_component(new(e1: 5, scalar: 2))
5.0
Returns the metric of the algebra.
Example:
{1, 1, 1, 0}represents a projective geometric algebra with three Euclidean basis vectors and one null basis vector.
Returns the norm of a multivector.
The norm is the square root of the absolute squared norm.
Example
iex> a = new(scalar: 3)
iex> norm(a)
3.0
Returns the normalized (real) value of a dual number.
This discards the infinitesimal component.
Normalizes a multivector.
The result has unit norm while preserving the direction of the multivector.
Raises ArgumentError when attempting to normalize a null
multivector.
Example
iex> a = new(scalar: 2)
iex> norm(normalize!(a))
1.0
iex> normalize!(new())
** (ArgumentError) cannot normalize given null multivector ...
iex> normalize!(new(scalar: epsilon() / 2.0))
** (ArgumentError) cannot normalize given null multivector ...
Returns the multiplicative identity:
1 + 0ε
Returns the pseudoscalar:
e1
iex> grades(pseudoscalar()) [1]
Extracts the real component.
Applies the reverse operation to a multivector.
Reverse (also called reversion) changes the sign of basis blades according to their grade:
grade 0: +
grade 1: +
grade 2: -
grade 3: -
grade 4: +
...For a blade with grade r, the sign is:
(-1)^(r(r-1)/2)Examples
iex> reverse(new(e1: 2))
new(e1: 2)
iex> reverse(new(scalar: 3))
new(scalar: 3)
Computes the right contraction of two multivectors.
The operation is generated from the geometric product and retains only terms satisfying the grade selection rule.
Example
iex> right_contraction(
...> new(e1: 2),
...> new(e1: 3)
...> )
Computes the rotor that maps one frame to another.
The function constructs blades from the source and target frames and computes the transformation rotor:
R = normalize(1 + T * S⁻¹)
where S is the source frame blade and T is the target frame blade.
The resulting rotor can be applied to multivectors to rotate the source frame into the target frame.
Turns a Multivector with only a scalar part into an elixir number.
iex> ~G[3.0] |> scalar()
{:ok, 3.0}
iex> ~G[3.0 + e1] |> scalar()
:error
Checks whether a multivector contains only a scalar component.
Components with an absolute value smaller than eps are considered
zero.
Examples
iex> scalar?(new(scalar: 3))
true
iex> scalar?(new(e1: 3))
false
iex> scalar?(new())
true
Returns the scalar coefficient of a multivector.
This is equivalent to retrieving the coefficient of the scalar blade.
Examples
iex> scalar_part(new(scalar: 5.0, e1: 2.0))
5.0
Computes the scalar product of two dual multivectors.
Creates a multivector from a string representation.
The ~G sigil provides a convenient syntax for constructing multivectors
using basis blades and coefficients.
Examples:
iex> ~G"e1"
new(e1: 1)
iex> ~G"3"
new(scalar: 3)The parsed expression is converted into the same representation accepted by
new/1, so blade ordering and signs are handled by the algebra implementation.
Computes the sine function.
The infinitesimal component contains the derivative:
d/dx sin(x) = cos(x)
Computes the square root.
For:
x = a + bεreturns:
sqrt(x) = sqrt(a) + b/(2sqrt(a)) ε
Returns the squared norm of a multivector.
The squared norm is computed as:
scalar_part(a * reverse(a))The result may be negative for algebras with indefinite metrics.
This is also called the Spinor Magnitude.
Example
iex> a = new(scalar: 3)
iex> squared_norm(a)
9.0
Subtracts two multivectors component-wise.
Examples
iex> a = new(scalar: 5)
iex> b = new(scalar: 2)
iex> sub(a, b)
new(scalar: 3)
Returns the multiplication table for the algebra.
The table contains precomputed geometric products between basis blades.
Each entry maps {left_blade, right_blade} to {coefficient, result_blade}.
The blades are represented internally as bitmasks.
Example
iex> table() |> Map.has_key?({1, 1})
false
Turns a Multivector with only a scalar part into an elixir number.
Raises if the given Multivector is not grade 0.
iex> ~G[3.0] |> to_scalar() 3.0
iex> ~G[3.0 + e1] |> to_scalar() ** (ArgumentError) cannot convert multivector to scalar: only grade-0 multivectors can be converted to a scalar, got ...
Computes the outer product (wedge product) of two multivectors.
The wedge product combines blades by joining their basis vectors. It is antisymmetric:
a ∧ b = -(b ∧ a)and vanishes when the operands share a basis vector.
Examples
iex> wedge(
...> new(e1: 1),
...> new(e1: 1)
...> )
new()
Computes the outer product of a list of multivectors.
The vectors are combined from left to right using the wedge product.
The result is a blade representing the subspace spanned by all input multivectors.
Returns the zero dual number:
0 + 0ε
Checks whether all coefficients of a multivector are zero.
Examples
iex> zero?(new())
true
iex> zero?(new(e1: 1))
false