Mathematical equation manipulation in Python

Question

I want to develop a GUI application which displays a given mathematical equation. When you click upon a particular variable in the equation to signify that it is the unknown

Answer 1

Things have sure changed since 2009. I don't know how your GUI application is going, but this is now possible directly in IPython qtconsole (which one could embed inside a custom PyQt/PySide application, and keep track of all the defined symbols, to allow GUI interaction in a separate listbox, etc.)

(Uses the sympyprt extension for IPython)

Answer 2

Sage has support for symbolic math. You could just use some of the equation manipulating functions built-in:

http://sagemath.org/

Answer 3

If you want to do this out of the box, without relying on librairies, I think that the problems you will find are not Python related. If you want to find such equations, you have to describe the heuristics necessary to solve these equations.

First, you have to represent your equation. What about separating:

• operands:
  • symbolic operands (a,b)
  • numeric operands (1,2)
• operators:
  • unary operators (-, trig functions)
  • binary operators (+,-,*,/)

Unary operators will obviously enclose one operand, binary ops will enclose two.

What about types?

I think that all of these components should derivate from a single common expression type. And this class would have a getsymbols method to locate quickly symbols in your expressions.

And then distinguish between unary and binary operators, add a few basic complement/reorder primitives...

Something like:

class expression(object):
    def symbols(self):
        if not hasattr(self, '_symbols'):
            self._symbols = self._getsymbols()
        return self._symbols
    def _getsymbols(self):
        """
        return type: list of strings
        """
        raise NotImplementedError

class operand(expression): pass

class symbolicoperand(operand):
    def __init__(self, name):
        self.name = name
    def _getsymbols(self):
        return [self.name]
    def __str__(self):
        return self.name

class numericoperand(operand):
    def __init__(self, value):
        self.value = value
    def _getsymbols(self):
        return []
    def __str__(self):
        return str(self.value)

class operator(expression): pass

class binaryoperator(operator):
    def __init__(self, lop, rop):
        """
        @type lop, rop: expression
        """
        self.lop = lop
        self.rop = rop
    def _getsymbols(self):
        return self.lop._getsymbols() + self.rop._getsymbols()
    @staticmethod
    def complementop():
        """
        Return complement operator:
         op.complementop()(op(a,b), b) = a
        """
        raise NotImplementedError
    def reorder():
        """
        for op1(a,b) return op2(f(b),g(a)) such as op1(a,b) = op2(f(a),g(b))
        """
        raise NotImplementedError
    def _getstr(self):
        """
        string representing the operator alone
        """
        raise NotImplementedError
    def __str__(self):
        lop = str(self.lop)
        if isinstance(self.lop, operator):
            lop = '(%s)' % lop
        rop = str(self.rop)
        if isinstance(self.rop, operator):
            rop = '(%s)' % rop
        return '%s%s%s' % (lop, self._getstr(), rop)


class symetricoperator(binaryoperator): 
    def reorder(self):
        return self.__class__(self.rop, self.lop)

class asymetricoperator(binaryoperator):
    @staticmethod
    def _invert(operand):
        """
        div._invert(a) -> 1/a
        sub._invert(a) -> -a
        """
        raise NotImplementedError

    def reorder(self):
        return self.complementop()(self._invert(self.rop), self.lop)


class div(asymetricoperator):
    @staticmethod
    def _invert(operand):
        if isinstance(operand, div):
            return div(self.rop, self.lop)
        else:
            return div(numericoperand(1), operand)
    @staticmethod
    def complementop():
        return mul
    def _getstr(self):
        return '/'

class mul(symetricoperator):
    @staticmethod
    def complementop():
        return div
    def _getstr(self):
        return '*'

class add(symetricoperator):
    @staticmethod
    def complementop():
        return sub
    def _getstr(self):
        return '+'

class sub(asymetricoperator):
    @staticmethod
    def _invert(operand):
        if isinstance(operand, min):
            return operand.op
        else:
            return min(operand)
    @staticmethod
    def complementop():
        return add
    def _getstr(self):
        return '-'

class unaryoperator(operator):
    def __init__(self, op):
        """
        @type op: expression
        """
        self.op = op
    @staticmethod
    def complement(expression):
        raise NotImplementedError

    def _getsymbols(self):
        return self.op._getsymbols()

class min(unaryoperator):
    @staticmethod
    def complement(expression):
        if isinstance(expression, min):
            return expression.op
        else:
            return min(expression) 
    def __str__(self):
        return '-' + str(self.op)

With this basic structure set up, you should be able to describe a simple heuristic to solve very simple equations. Just think of the simple rules you learned to solve equations, and write them down. That should work :)

And then a very naive solver:

def solve(left, right, symbol):
    """
    @type left, right: expression
    @type symbol: string
    """
    if symbol not in left.symbols():
        if symbol not in right.symbols():
            raise ValueError('%s not in expressions' % symbol)
        left, right = right, left

    solved = False
    while not solved:
        if isinstance(left, operator):
            if isinstance(left, unaryoperator):
                complementor = left.complement
                right = complementor(right)
                left = complementor(left)
            elif isinstance(left, binaryoperator):
                if symbol in left.rop.symbols():
                    left = left.reorder()
                else:
                    right = left.complementop()(right, left.rop)
                    left = left.lop
        elif isinstance(left, operand): 
            assert isinstance(left, symbolicoperand)
            assert symbol==left.name
            solved = True

    print symbol,'=',right

a,b,c,d,e = map(symbolicoperand, 'abcde')

solve(a, div(add(b,mul(c,d)),e), 'd') # d = ((a*e)-b)/c
solve(numericoperand(1), min(min(a)), 'a') # a = 1

Answer 4

Using SymPy, your example would go something like this:

>>> import sympy
>>> a,b,c,d,e = sympy.symbols('abcde')
>>> r = (b+c*d)/e
>>> l = a
>>> r = sympy.solve(l-r,d)
>>> l = d
>>> r
[(-b + a*e)/c]
>>> 

It seems to work for trigonometric functions too:

>>> l = a
>>> r = b*sympy.sin(c)
>>> sympy.solve(l-r,c)
[asin(a/b)]
>>> 

And since you are working with a GUI, you'll (probably) want to convert back and forth from strings to expressions:

>>> r = '(b+c*d)/e'
>>> sympy.sympify(r)
(b + c*d)/e
>>> sympy.sstr(_)
'(b + c*d)/e'
>>> 

or you may prefer to display them as rendered LaTeX or MathML.

Answer 5

What you want to do isn't easy. Some equations are quite straight forward to rearrange (like make b the subject of a = b*c+d, which is b = (a-d)/c), while others are not so obvious (like make x the subject of y = x*x + 4*x + 4), while others are not possible (especially when you trigonometric functions and other complications).

As other people have said, check out Sage. It does what you want:

You can solve equations for one variable in terms of others:

sage: x, b, c = var('x b c')
sage: solve([x^2 + b*x + c == 0],x)
[x == -1/2*b - 1/2*sqrt(b^2 - 4*c), x == -1/2*b + 1/2*sqrt(b^2 - 4*c)]