Basic type checking

The double application function requires that the i parameter is an int, which is verified by a single predicate (the lambda expression):

@preconditions(lambda i: isinstance(i, int))
def double(i):
    return 2*i

Multiple predicates

Multiple predicates may be specified:

@preconditions(
    lambda i: isinstance(i, int),
    lambda i: i > 0,
    )
def double(i):
    return 2*i

Note that this is functionally equivalent to this single predicate version:

@preconditions(
    lambda i: isinstance(i, int) and i > 0,
    )
def double(i):
    return 2*i

The multi-predicate version should (eventually) have more specific error reporting for a failure, while the single predicate version may be more efficient.

Multiple arguments

Multiple predicates can express preconditions for multiple arguments:

@preconditions(
    lambda s: isinstance(s, unicode),
    lambda n: isinstance(n, int) and n >= 0,
    )
def repeat(s, n):
    return s*n

However, a single predicate can express preconditions for multiple arguments. This allows relational preconditions:

@preconditions(
    lambda a, b: a <= b
    )
def strict_range(a, b):
    return range(a, b)

Method preconditions

Predicates can be expressed for methods, including relations to self. For example, a Monotonic instance ensures that each call to .set must pass a value larger than any previous call:

class Monotonic (object):
    def __init__(self):
        self.v = 0

    @preconditions(lambda self, v: v > self.v)
    def set(self, v):
        self.v = v

Preconditions can be applied to special methods, such as __new__, __init__, __call__, etc…

class LinearRange (tuple):
    @preconditions(
           lambda a: isinstance(a, float),
           lambda b: isinstance(b, float),
           lambda a, b: a < b,
           )
    def __new__(cls, a, b):
        return super(OrderedTuple, cls).__new__(cls, (a, b))

    @preconditions(lambda w: 0 <= w < 1.0)
    def __call__(self, w):
        lo, hi = self
        return w * (hi - lo) + lo

    @preconditions(lambda x: self[0] <= x < self[1])
    def invert(self, x):
        lo, hi = self
        return (x - lo) / (hi - lo)

Parameter Name Equivalence

The parameter names in a predicate must match parameter names in the application function. This is known as parameter name equivalence [1].

One exception to this rule is for default parameters within predicates. Default parameters may be used to associate some state at predicate definition time. For example:

scores = {}

@preconditions(
    lambda color, _colors=['RED', 'GREEN', 'BLUE']: color in _colors
    )
def get_color_score(color):
    return scores[color]

This feature may be most convenient when there’s a need to remember a local loop variable.