pyquchk 0.5.3

Already implemented

pyquchk combines two approaches to the arguments values generation in a single framework:

• QuickCheck-like (random)

  Random (fuzzy) testing runs a function for many randomly generated arguments values to make sure it’s generally correct. Some features (both from the original QuickCheck and unique to pyquchk)

• SmallCheck-like (serial)

  Serial testing means that values are generated starting with some edge cases and small ones. This exhausts all possible values up to some “size” and is deterministic, in contrast to the random approach. Serial testing nicely complements the random one and so gives more confidence that the function tested is correct.

Other notable features include:

• a lot of configurable built-in generators for different value types (see arbitraries for the list of currently implemented generators)

• custom generators are supported, both for built-in types and user-created ones

• easy-to-write and read syntax for simple property checking as well as for full tests to be run with a framework (nose or py.test for example)

• pyquchk attempts to give you the smallest available counterexample, always performing shrinkg of the generated value (and if a value is found with serial testing, it’s probably already minimal)

• assume function (allows so-called conditional properties and much more)

• Easy creation of Arbitraries combinations

  Examples: string from list; tuple with equally-sized strings.

Planned

pyquchk is in active development, and I have many features to implement here:

• Option to check for exists, not for_all

• Tests timing

  Show time spent in pyquchk code and in tested function

• Support generators arguments depending on other generators values

  Something like forAll in original QuickCheck, but everything is to be defined as the function default arguments. Example: integer from the range [0:len(list)].

• Size support

  Generation function takes size parameter which increases (from 0 to 1?) from the first to the last test. Explore sized, resize and so on from QuickCheck - whether they will fit here.

• Look for best way to express:

  • Modifiers like positive/negative/nonempty/…

  • Different numbers distributions like uniform/exponential/…

  • Structures like def f(a={int: [str_(...)]}, b=[float], c={'x': int, 'y': list})

• Any use for Python 3 annotations?

• Better testing pyquchk itself

• Smart shrink from QuickCheck, SmartCheck

• Collect, classify, cover

  Not only boolean parameters (label/no label), but numerical as well. Example: difference between approximate function result and the exact one.

• Verbosity settings

• Print seed and allow to change it

• Don’t fail fast

• Add timeout