Equato - An Equation Solver

Equato is a relatively simple equation solver that works using constraint propagation and Newton Raphson iteration. It takes a set of simple algebraic equations and a list of input variables and produces a list of output variables. You can invoke an iterative solution by taking a guess for an unknown value. The program is flakey and incomplete, and relies on a solution technique of limited reliability. On the other hand, it is free, it runs on the Macintosh and possible on Windows, and it sometimes produces useful answers.

DOWNLOAD

• for Mac OS X (Universal Binary)
• for Mac OS X and possibly Mac OS 9 (Carbon)
• possibly even for Windows
• maybe for Linux

The Equation Pane

There are two panes in Equato, one for equations and one for variables. You click and edit an equation as you might enter values in a single column spreadsheet. The equations may contain variables (with any reasonable number of letters or numbers in their name) and may use the operators +, -, *, /, ^ for exponentiation, sin, cos, atan of two arguments, exp and ln.

You can enter comments alone or after an equation line by preceding them with an apostrophe.

Equations and Solutions

Equato works by constraint propagation. Each operator, for example, a plus operator, works several ways. If there is an equation a=b+c, and b and c are known, it will do the addition to find a. If a and c are known, it will subtract to find b. Some operators are irreversible. For example, if the divisor would be zero, Equato just ignores the constraint. Some operators, particularly the trigonometric functions are only partially reversible since there are a lot of values of x for which sin(x) = 0. Generally, Equato will pick one of the smaller values so that typical geometeric computations will reverse solve correctly.

If there is an error in the equation there will be a cryptic, unreadable message in the status column, otherwise that column will be either blank or contain an indication that the equation cannot be balanced because the problem is overdetermined.

Integrals and Derivatives

integral(expr, vbl, a, b) - The definite integral function has four arguments: an expression to integrate, an integration variable, a low bound and a high bound. It integrates using Simpson's method (see the Preferences below), and within the expression, any references to the integration variable are local.

derivative(expr, vbl, a) - The definite derivative function has three arguments: an expression to compute the derivative of, the variable with respect to which the derivative is to be computed, and a value for that variable at which the derivative is to be computed. The derivative is computed iteratively by decreasing the interval of approximation.

Neither of these functions backsolves. They are only computed forwards, but the Newton-Raphson iterative solver can be used to solve for integral bounds or derivative values. For example, solving for the zero of a derivative function can be used to find maxima or minima.

The Variable Pane

The variable pane shows the variables, their status, their input value or guess value, their computed output value and, if you care to enter one, a descriptive comment. The ??? in the status column means that the values have not been solved for. The ?1 specifies a guess value of 1 for the variable and indicates that an iterative solution should be used if necessary.

The solver first tries simple constraint relaxation. This works surprisingly often. If it fails, it shows a lot of question marks, unless the problem is overdetermined in which case see above. If there are any guess values it formulates the problem as having N equations and N unknowns, or at least tries to, and then uses a gradient following algorithm, Newton Raphson, to iterate towards an answer. If this doesn't work, it tries working with some subset of the guesses specified in hopes of finding a stable approach to an answer.

As you can see in this example, it has computed the square root of minus one to be i.

Preferences

If Equato is still causing trouble, or you'd like to try your luck at getting some more accuracy out of it, try tweaking the preferences. You can try more iterations, a tighter iteration tolerance, or you can try relaxing the constraint tolerance and adding some more equations, or even dink around with the step size used for approximating the derivatives. Newton-Raphson is actually a pretty stable algorithm if you restrict the functions you use.

The maximum guesses to ignore is useful for cases which are sometimes unstable, but could be stable if you weren't using so many iteration variables. For simple problems, Equato can be pretty fast, but if you are clever with the preferences you can probably get it to explode combinatorically.

The integration steps are the number of steps to use for Simpson's rule when evaluating expressions containing integral(f(x), x, a, b). This should be even, but Equato will add one if it is odd.

The text font and size stuff might be useful, but it is poorly implemented. The default font is the system font, and the default size should work with the tables. Changing either of these might make things look "pretty".