Rudimentary symbolic math library for Ruby. This gem is mainly intended as a coding exercise. The operations have not been optimized for speed. The current state of the project is 'under construction'.
Supported features:
- Composing formulas from +, -, *, / and other standard algebraic operations
- Reduction rules
- Derivation
- Simple heuristic integration (not cas level)
- Simple polynomial factorization with one variable
- Complex numbers and quaternions
- Exterior algebra and exterior derivative (limited and possibly faulty)
- Operator composition algebra (limited)
- Vectors and linear forms
- Braket notation for vectors, forms and linear operators
- Matrices
Add this line to your application's Gemfile:
gem 'symath'
Then execute:
$ bundle
Or install it yourself as:
$ gem install symath
Using the library:
> require 'SyMath'
A convenient way to explore the SyMath library is using the interactive Ruby interpreter, irb:
> # Load the symath library > require 'symath' => false > # Add the symbols module to your environment > extend SyMath::Definitions => main
You can now say, for example:
> # Simplify an expression > sin(:x) + 2*sin(:x) => 3*sin(:x) > # Derivative of tan(2*y + 3) > (d(tan(2*:y + 3))/d(:y)).evaluate => 2*(tan(2*y + 3)**2 + 1)
Ruby symbols, :x and :y in the above example, are converted into symbolic math variables and Ruby numbers are converted into symbolic math numbers. Functions, operators and constants (e, pi, i, etc.) are available as methods through the SyMath::Definitions module. In some cases it is necessary to tell Ruby that your number or symbol is to be understood as a symbolic object, and not just a Ruby number or symbol. Use the to_m method to explicitly convert them to symbolic bjects:
> # Ruby integer math > 3**4 => 81 > # SyMath symbolic math > 3.to_m**4 => 3**4 > (3.to_m**4).normalize => 81
An complete expression can also be converted from a string, using the same to_m method:
> 'ln(e) + sin(pi/2)'.to_m => ln(e) + sin(pi/2) > 'ln(e) + sin(pi/2)'.to_m.normalize => 2
The module SyMath::Definitions is available to be included or extended to your class or code block. It gives a Ruby method for each operator, function and constant that exists, so they can be referred to by their name, as in the code examples above. If you don't want to use the module, functions, operators and constants must be referred to by the fn, op and definition methods:
> # Using the SyMath::Definitions methods > sin(:x) => sin(x) > int(:x) => int(x) > e => e > sin => sin(...) > # Using the generic creator functions > fn(:sin, :x) => sin(x) > op(:int, :x) => int(x) > definition(:e) => e > definition(:sin) => sin(...)
The SyMath::Definitions module is updated dynamically after the user has defined new functions, operators and constants.
Symbolic math objects, inheriting from SyMath::Value, all have a to_s method which returns a string representation of the object. The string representation is compatible with the String.to_m method which converts a string representation into a symbolic object:
> (ln(e) + sin(pi/2)).to_s => "ln(e) + sin(pi/2)" > 'ln(e) + sin(pi/2)'.to_m => ln(e) + sin(pi/2)
SyMath::Value overrides the Object.inspect method, returning the to_s representation rather than the more verbose and less readable Object.inspect output. This behaviour can be disabled with the setting 'inspect_to_s':
> SyMath.setting(:inspect_to_s, false) => false > ln(e) + sin(pi/2) => "#<SyMath::Sum:0x000055e8a1d93b38 @definition=..."
Simple reduction rules are automatically applied when composing an expression. These can be disabled with the setting 'compose_with_simplify'. More thorough reductions are done by the use of the normalize method.
> e*e*e*e => e**4 > SyMath.setting(:compose_with_simplify, false) => false > e*e*e*e => e*e*e*e > sin(pi/2).normalize => 1
The library comes with a number of built-in function, which the system knows how to derivate and integrate over. The built-in functions also have a number of reduction rules which are applied by the reduce method and also as part of the 'normalize' method. A list of the defined functions is returned by the functions method. The description method gives a small description of the function:
> SyMath::Definition::Function.functions
=> [sqrt, sin, cos, tan, sec, csc, cot, arcsin, arccos, arctan,
arcsec, arccsc, arccot, ln, exp, abs, fact, sinh, cosh, tanh,
coth, sech, csch, arsinh, arcosh, artanh, arcoth, arsech,
arcsch]
> sin.description
=> "sin(x) - trigonometric sine"
User-defined functions can be added by the method define_fn:
> define_fn('poly', [:x, :y], :x**3 + :y**2 + 1)
=> poly
The user-defined function will now be available as a method in the SyMath::Definitions module and can be used in expressions, just as the built in functions. Functions defined by an expression can be evaluated by the evaluate method, which returns the expression with each free variable replaced with the input arguments to the function:
> poly(2, 3).evaluate => 2**3 + 3**2 + 1 > poly(3).evaluate.normalize => 18
A nameless user-defined function can be created using the lmd method. The method returns a function object which does not have a name, but otherwise works as a function. The lambda function has important usages in operators. Since they eturn a function as the result, it will typically be a lambda function. Also, the lambda function can be used for wrapping an expression into a function before doing an integral or derivative, in this way telling which variables the operator should work on. The lambda function can be called using the call method or the Ruby 'call' operator '()':
> l = lmd(:x**3 + :y**2 + 1, :x, :y) > l.(2, 3) => (x**3 + y**2 + 1).(2,3) > l.(2, 3).evaluate => 2**3 + 3**2 + 1 > l.(2, 3).evaluate.normalize => 18
The library has some built-in operators, i.e. functions which take functions as arguments and return functions. A list of the defined operators is returned by the operators method. The description method gives a small description of the operator:
> SyMath::Definition::Operator.operators
=> [d(...), xd(...), int(...), [f](b,), #(), b(), hodge(...), grad(f),
curl(f), div(f), laplacian(f), codiff(f), laplace(f), fourier(f),
invfourier(f), dpart(f,t)]
> codiff.description
=> "codiff(f) - codifferential of function f"
User-defined operators can be added by the method define_op:
> define_op('d2', [:f, :x], d(d(:f)/d(:x))/d(:x))
=> d2
> d2(:x**3 + 2, :x).evaluate
=> 6*x
The user-defined function will now be available as a method in the SyMath::Definitions module and can be used in expressions.
Evaluating a functions or operators which is defined by an expression returns the expression with each free variable replaced with input arguments. Functions which do not have an expression will typically evaluate to itself (no reduction). Most operators which do not have an expression has a built in evaluation, and returns a function or expression according to the operator.
The d-operator returns the differential of a function or expresson. If a function is given, the differential is made over all the free variables of the function. If an expression is given, the operator differentiates over the first free variable found in the expression. Wrapping the expression into a lambda function makes it possible to say which variables to differentiate over. Note that the differential is an operator, so it returns the result in a a lambda function, and not just the expression.
> d(sin(:x)).evaluate => cos(x)*dx.(x) > d(:x**2 + :y**3 + 1).evaluate.normalize => (2*x*dx).(x) > d(lmd(:x**2 + :y**3 + 1, :y)).evaluate.normalize => (3*y**2*dy).(y) > d(lmd(:x**2 + :y**3 + 1, :x, :y)).evaluate.normalize => (3*y**2*dy + 2*x*dx).(x,y)
As a special case, the notatonal form d(f)/d(x) is recognized as the derivative of f with regards to x. This is calculated as d(lmd(f, x))/d(x). This evaluates to the derivative expression:
> (d(:y**2 + :x**3 + 1)/d(:x)).evaluate => 3*x**2
The partial derivative is available as well as 'syntactic sugar':
> dpart(:x**2 + :y**3 + 1, :x).evaluate.normalize => 2*x > dpart(:x**2 + :y**3 + 1, :y).evaluate.normalize => 3*y**2
Integration is available as the int-operator. The algorithm is only a very simple one, imitating the most basic techniques of finding the anti-derivative, combined with a few well known equation patterns. The operation also has a limitation when it comes to non-commutable terms (matrices, quaternions, etc.). In that case, the result is not reliable.
With one argument, the operator evaluates to the antiderivative of the expression:
> int(2**:x).evaluate => 2**x/ln(2) + C
The variable C is used by convention to represent the free constant factor of the antiderivative.
With three arguments, the int-operator evaluates to the definite integral from a to b:
> int(2**:x, 3, 4).evaluate.normalize => 8/ln(2)
The imaginary unit, i, is available as a constant, and can be used for composing expressions with complex numbers. Simple reduction rules are built in which reduces i*i to -1, and so on.
The basic quaternions, i, j, k are also available as constants. The quaternion i is identical to the complex imaginary unit. Some simple reduction rules are available for the quaternions as well.
Variables can be defined as vectors, oneforms, multilinear forms and other vector-like objects by specifying a type when created. Vector-like objects are associated with a vector space. The 'vector space' is matematically speaking an abuse of the term since vectors, n-forms and other linear operators live in separate vector spaces (typically denoted as V, V^V and VxVx...xV). They are, however, closely related, and share many properties. So we let all these vector-like object share their vector space. This may be changed in the future.
The vector space may be defined with a set of basis vectors and a metric. This is optional, but a requirement if the vectors are to be used with the exteriar algebra operations (see next section).
A set of built-in vector spaces are created at startup, and a default space, the euclidean_3d, is chosen. Vector-like objects in an expression can be put on matrix representation by the method to_matrix (only objects which has an obvious representation, like basis vectors and other known objects, are converted).
Vector-vector and oneform-oneform multiplications are automatically composed into outer products. Oneform-vector products are, on the other hand, composed into inner products. Vectors and oneforms multiplied with a linear operator are composed into a normal product, which is interpreted as an operator composition operation.
> # Create vector objects within the default vector space
> a1 = :a1.to_m('vector')
> a2 = :a2.to_m('form')
> # Create objects within a given vector space
> b = :b.to_m('vector', 'minkowski_4d')
> c = :c.to_m('form', 'minkowski_4d')
> m4 = SyMath.get_vector_space('minkowski_4d')
> d = m4.vector(:d)
> e = m4.oneform(:e)
> # Set the default vector space
> SyMath.list_vector_spaces
=> ["euclidean_3d", "minkowski_4d", "quantum_logic"]
> SyMath.set_default_vector_space('minkowski_4d')
=> minkowski_4d ([x0, x1, x2, x3])
> # Create a new vector space, and create a vector in it
> myspace = SyMath::VectorSpace.new('myspace', dimension: 5,
basis: [:v, :w, :x, :y, :z].to_m)
> SyMath.list_vector_spaces
=> ["euclidean_3d", "minkowski_4d", "quantum_logic", "myspace"]
> v1 = myspace.vector(:v1)
A vector space can be defined as 'normalized'. In this case, all vectors, co-vectors and linear operators are assumed to be unitary. Vectors in normalized vector spaces are stringified in bra-ket/dirac notation form since such vector spaces are often used in quantum mechanical formula. The bra-ket notation is available for all vectors in the String.to_m parser.
> norm = SyMath::VectorSpace.new('norm', dimension: 3,
basis: [:a, :b, :c].to_m, normalized: true )
> SyMath.set_default_vector_space('norm')
=> norm ([a, b, c])
> 'A B|c>'.to_m
=> A B|c>
> ''.to_m.normalize
=> 1
> '|a>|b>'.to_m
=> |a,b>
A built-in normalized vector space, 'quantum_logic', is available. It has the basis vectors '|0>' and '|1>', and various linear operators corresponding to quantum logical gates like, qX, qY, qZ, qH, qS. Simple reduction rules exist between these objects. They can also be converted to matrix form for further calculations.
> SyMath.set_default_vector_space('quantum_logic')
=> quantum_logic ([q0, q1])
> 'qX|0>'.to_m
=> qX|0>
> 'qX|0>'.to_m.normalize
=> |1>
> '<1|qX|0>'.to_m.normalize
=> 1
> 'qH|0>'.to_m.normalize
=> |+>
Caveat: Exterior algebra and differential forms are not well understood by the author of this code. The following has not been reviewed by anyone who understand the subject well. It may very well contain a lot of errors and misunderstandings.
Forms can be defined in several ways. The following are equal. All of them will create a oneform in the default vector space:
> # Using the to_d method on a scalar variable
> :x.to_m.to_d
=> dx
> # Differentiating a scalar variable
> d(:x)
=> dx
> # Creating a variable, and specifying the form type
> :dx.to_m('form')
=> dx
N-forms can be created by specifying the type explicitly:
> :da.to_m('form'.to_t(indexes: 'll'))
Forms can be wedged together, forming n-forms (note that the ^ operator has lower preceedence in Ruby than in math, so parantheses must be used, e.g. when adding):
> d(:x)^d(:y)^d(:z) => dx^dy^dz > (d(:x)^d(:x)^d(:z)).normalize => 0
The exterior derivative and related operators all work in a local coordinate system defined by the basic vectors of the current vector space.
The rest of this section assumes that the following scalars, vectors and forms are defined:
> SyMath.assign_variable('basis', [:x1, :x2, :x3])
=> {dx1=>x1', dx2=>x2', dx3=>x3'}
> x1 = :x1.to_m
> x2 = :x2.to_m
> x3 = :x3.to_m
> x1v = :x1.to_m('vector')
> x2v = :x2.to_m('vector')
> x3v = :x3.to_m('vector')
> dx1 = x1.to_d
> dx2 = x2.to_d
> dx3 = x3.to_d
The exterior derivative is available as the xd-operator:
> xd(:x1 - :x1*:x2 + :x3**2).evaluate => dx1 - (dx1*x2 + x1*dx2) + 2*x3*dx3
The musical isomorphisms are available as the flat and sharp operators:
> flat(x1v^x2v).evaluate => dx1^dx2 > sharp(dx1^dx2).evaluate => x1'^x2'
The flat and sharp operators use the metric tensor in their calculations and thus require that the current vector space has a metric.
The hodge star operator is available as well:
> hodge(dx1^dx2).evaluate => dx3 > hodge(3).evaluate => 3*dx1^dx2^dx3
Gradient, curl, divergence, laplacian and co-differential are defined from the above operators in the usual way:
> grad(x1 - x1*x2 + x3**2).evaluate => 2*x3*x3' - x2*x1' - x1*x2' + x1' > curl(-x2*x1v + x1*x2*x2v + x3*x3v).evaluate => x2*x3' + x3' > div(-x2*x1v + x1*x2*x2v + x3*x3v).evaluate => x1 + 1 > laplacian(x1**2 + x2**2 + x3**2).evaluate => 6 > codiff(x1**2*(dx1^dx3) + x2**2*(dx3^dx1) + x3**2*(dx1^dx2)).evaluate => 2*x1*dx3
Row matrices can be defined by converting an array to a math object, using the to_m method. Column matrices and two dimensional matrices can similarily be created from two dimensional arrays:
> a = [[1, 2, 3], [4, 5, 6]].to_m > b = [-3, 4, 1].to_m
Various operations are available on matrices such as:
> m = [1, 2, 3].to_m > m.transpose => [1; 2; 3] > n = [[4], [5], [5]].to_m > m.kroenecker(n) => [4, 2 4, 3 4; 5, 2 5, 3 5; 5, 2 5, 3 5] > m.kroenecker(n).normalize => [4, 8, 12; 5, 10, 15; 5, 10, 15] > s = [[4, 7], [2, 6]].to_m > s.determinant => 10 > s.inverse => [6/10, - 7/10; - 2/10, 4/10] > s.trace => 10 > s.adjugate => [6, - 7; - 2, 4]
The vector and matrix cells can of course contain symbolic expressions instead of just numbers.
Row- and column vectors can be translated into sums of basis forms and basis vectors, respectively, as long as the length of the row or column is equal to the vector space.
> m.to_vector => dx1 + 2 dx2 + 3 dx3 > n.to_vector => 4 x1' + 5 x2' + 5 x3'
The library contains a few more complex expression manipulation methods which are available to all math expression objects inheriting from the SyMath::Value class (the root class of the expression components).
The normalization method tries to put an expression on a normal form, based on some heuristics.
- Expressions formed by natural numbers are calculated.
- Fractions of natural numbers are simplified as far as possible.
- Products of equal factors are collapsed to power expressions.
- Products of powers with equal base are collapsed.
- Sums of equal terms are simplified to integer products.
- Product factors are ordered if permitted by commutativity.
- Sum terms are ordered.
> # FIXME: Find some better examples > (:x*4*:x*3*:y*:y**10).normalize => 12*x**2*y**11
The replace method replaces takes a map of 'variable => expression' as argument. It looks up all instances of the variables in the original expression, and replaces them with the expressions given by the map:
> (:x**:y).replace({:x.to_m => :a + 2, :y.to_m => :b + 3})
=> (a + 2)**(b + 3)
The match method can be seen as a 'reverse' operation to replace-method covered in the last section. It compares an expression to a template expression containing some free variables. It returns an array of all possible maps for the free variables in the template so that it matches the original expression:
> (:x**2 + :y**2 + 3).match(:a + :b, [:a.to_m, :b.to_m])
=> [{a=>x**2, b=>y**2 + 3},
{a=>y**2, b=>x**2 + 3},
{a=>3, b=>x**2 + y**2},
{a=>x**2 + y**2, b=>3},
{a=>x**2 + 3, b=>y**2},
{a=>y**2 + 3, b=>x**2}]
The match_replace method tries to find an occurence of a pattern in the expression, and replaces it if it is found. The method can be used together with the iterate method. The latter repeats a method until there are no more changes:
> a = sin(sin(sin(:a + :b) - sin(:f)))
> a.match_replace(sin(:x), :e*:x, [:x])
=> e*sin(sin(a + b) - sin(f))
> a.iterate('match_replace', sin(:x), :e*:x, [:x])
=> e*e*(e*(a + b) - e*f)
The factorization method has been ripped from the python Py-library. It factorizes a polynomial of one variable:
> (6*:x**2 + 24*:x**3 - 27*:x**4 + 18*:x**5 + 72*:x**6 - 9*:x).factorize => 3*x*(2*x - 1)*(4*x + 3)*(3*x**3 + 1)
The expand method expands the polynomial:
> (3*:x*(2*:x - 1)*(4*:x + 3)*(3*:x**3 + 1)).expand.normalize => 72*x**6 + 18*x**5 - 27*x**4 + 24*x**3 + 6*x**2 - 9*x
The library has some global settings which change the behaviour of the system:
> # List all settings
> SyMath.settings
=> {
# Symbol used when a vector is stringified
:vector_symbol => "'",
# Show all parentheses when stringifying an expression
:expl_parentheses => false,
# Put square roots on exponent form
:sq_exponent_form => false,
# Put fractions on exponent form
:fraction_exponent_form => false,
# In latex strings, insert a product sign between the factors
:ltx_product_sign => false,
# Use simplification rules when expressions are composed
:compose_with_simplify => true,
# Use oo as complex infinity
:complex_arithmetic => true,
# Return to_s representation by the inspect method
:inspect_to_s => true,
# Maximum value of factorial which is normalized to a number
:max_calculated_factorial => 100
}
> # Show one setting
> SyMath.setting(:vector_symbol)
=> "'"
> # Change a setting
> SyMath.setting(:vector_symbol, '¤')
=> "¤"
After checking out the repo, run bin/setup to install
dependencies. Then, run rake spec to run the tests. You can also run
bin/console for an interactive prompt that will allow you to
experiment.
To install this gem onto your local machine, run bundle exec rake install. To release a new version, update the version number in
version.rb, and then run bundle exec rake release, which will
create a git tag for the version, push git commits and tags, and push
the .gem file to rubygems.org.
Bug reports and pull requests are welcome on GitHub at https://github.com/erikoest/symath. This project is intended to be a safe, welcoming space for collaboration, and contributors are expected to adhere to the Contributor Covenant code of conduct.
The gem is available as open source under the terms of the MIT License.
Everyone interacting in the SyMath project’s codebases, issue trackers, chat rooms and mailing lists is expected to follow the code of conduct.