Symbolic Computing

The CortexJS Compute Engine essentially applies transformations to a MathJSON expression by applying rewriting rules.

There are several kind of transformations, depending on the desired result:

Format Put an expression in canonical ("standard") form, for easier sorting, comparing and computing. Typically limited to accounting for the flags associative idempotent involution and for sorting the arguments if commutative. Independent of the assumptions.
Simplify Apply rewriting rules specific to each function, eliminating constants and common sub-expressions. Use available assumptions to determine which rules are applicable. Limit calculations to exact results using integers. The result is in canonical format.
Evaluate Calculate the value of an expression and all its terms. Replace symbols with their value. Can perform approximate calculations using floating point numbers. The result is simplified and canonical.

Example, given f is \( 2 + (\sqrt{x^2 \times 4} + 1) \) and x is 3:

ce.format(f) 1 + 2 + \sqrt{4x^2} Arguments sorted, distributed
ce.simplify(f) 2 + 2x Exact calculations of numeric constants, simplification
ce.evaluate(f) 8 Evaluation of symbols

Other operations can be performed on an expression: comparing it to a pattern, replacing part of it, and applying conditional rewrite rules.

Functions such as ce.simplify(), ce.evaluate(),, ce.N(), ce.canonical(), etc… require a ComputeEngine instance which is denoted by the ce. prefix.

import { ComputeEngine, parse } from '@cortex-js/compute-engine';
const ce = new ComputeEngine();
ce.simplify(ce.parse('3x^2 + 2x^2 + x + 5'));

Format with a Canonical Form

The canonical form of an expression is obtained by rewriting an expression without making assumptions about any variables in the expression.

For example:

  • terms can be sorted in a specific order
  • some operations may be substituted with others, for example substraction replaced by addition. \(1 + 2 - 3 \longrightarrow Add(1, 2, -3)\)

Canonical forms are somewhat arbitrary, and not necessarily "the simplest" way to represent an expression. But just like the order of the letters of the alphabet is arbitrary, the canonical forms are nonetheless convenient to sort, search and compare expressions.

For example \( 1 + x\) and \(x + 1\) are two expressions with the same canonical form.

To obtain the canonical form of an expression, use the ce.canonical() function.

console.log(ce.canonical(["Add", 2, "x", 3]);
// ➔ ["Add", 2, 3, "x"]


To obtain a simpler expression of a symbolic expression, use the ce.simplify() function.

The ce.simplify() function makes use of available assumptions about symbols and return an exact result: there are no numerical evaluation done that could result in a loss of precision.


To apply a sequence of definitions to an expression in order to reduce, simplify and calculate its value, use the ce.evaluate() function.

An expression to be evaluated will also be simplified, evaluated numerically and put into canonical form.

When a function is evaluated, its arguments are first evaluated left to right, the the function is applied to the arguments.

However, a function definition can specify that some or all of its arguments should not be evaluated. This can be useful for a function that needs to perform symbolic manipulation of an expression: otherwise, the expression would be evaluated without giving a chance to the function to access the symbolic expression.

While a function definition will usually indicate which arguments should be evaluated or not, it is possible to override this.

To prevent an argument from being evaluated, use the Hold function.

To force an argument to be evaluated, use the Evaluate function.

Other Symbolic Manipulation

You can compare two expressions, check if an expression match a pattern, apply a substitution to some elements in an expression or apply a conditional rewriting rule to an expression.