autosig 0.2.0

Define signatures to create beautiful APIs

Autosig

Experimental

Autosig allows to create classes that describe signatures of functions and common processing of arguments. This allows to:

• Model functions that share the same signature

• Model the commonalities between different signatures, e.g. sharing the first few arguments

• Model common processing normally associated with signatures, such as default values, type checking, validation and conversion, this both at the level of individual arguments and globally for a signature (e.g, check the the first argument is a DataFrame vs. check the that first two arguments are both of the same type)

Since a signature is modeled with class, inheritance can be used to capture commonalities and differences between signatures:

from autosig import *
# define a class with @signature decorator inheriting from a subclass of
# Signature (I know, a little redundant in this case)
# x and y are parameters with no special properties
@signature
class BinaryOp(Signature):
    x = param()
    y = param()

# define a parameter with an int annotation, converting to int if necessary)
int_param = param(converter=int, type=int)

# define another signature with two integer parameters, same annotation and
# conversion behavior
@signature
class BinaryIntOp(BinaryOp):
    x = int_param
    y = int_param


# define a binary operator
@autosig(sig=BinaryOp)
def add(x, y):
    return x + y

# define a binary operator with int parameters
@autosig(sig=BinaryIntOp)
def int_add(x, y):
    return x + y

add(2, 3) # 5
add(2, "3") # fails
int_add(2, "3") # 5

History

0.2.0 (2018-08-27)

• More stringent enforcement of signatures including defaults.

0.1.0 (2018-04-25)

• First release on PyPI.