Function Mangling
This module provides utilities for extracting information about python functions.
AUTHORS:
Bases: object
This class provides functionality to normalize the arguments passed into a function. While the various ways of calling a function are perfectly equivalent from the perspective of the callee, they don’t always look the same for an object watching the caller. For example,
sage: def f(x = 10):
... return min(1,x)
the following calls are equivalent,
sage: f()
1
sage: f(10)
1
sage: f(x=10)
1
but from the perspective of a wrapper, they are different:
sage: def wrap(g):
... def _g(*args,**kwargs):
... print args, kwargs
... return g(*args, **kwargs)
... return _g
sage: h = wrap(f)
sage: t = h()
() {}
sage: t = h(10)
(10,) {}
sage: t = h(x=10)
() {'x': 10}
For the purpose of cached functions, it is important not to distinguish between these uses.
INPUTS:
EXAMPLES:
sage: from sage.misc.function_mangling import ArgumentFixer
sage: def wrap2(g):
... af = ArgumentFixer(g)
... def _g(*args, **kwargs):
... print af.fix_to_pos()
... return g(*args,**kwargs)
... return _g
sage: h2 = wrap2(f)
sage: t = h2()
((10,), ())
sage: t = h2(10)
((10,), ())
sage: t = h2(x=10)
((10,), ())
sage: class one:
... def __init__(self, x = 1):
... self.x = x
sage: af = ArgumentFixer(one.__init__.__func__, classmethod=True)
sage: af.fix_to_pos(1,2,3,a=31,b=2,n=3)
((1, 2, 3), (('a', 31), ('b', 2), ('n', 3)))
Normalize the arguments with a preference for named arguments.
INPUT:
OUTPUT:
We return a tuple
\((e_1, e_2, ..., e_k), ((n_1, v_1), ... , (n_m, v_m))\)
where \(n_1, ... , n_m\) are the names of the arguments and \(v_1, ..., v_m\) are the values passed in; and \(e_1, ..., e_k\) are the unnamed arguments. We minimize \(k\).
The defaults are extracted from the function and filled into the list K of named arguments. The names \(n_1, ..., n_t\) are in order of the function definition, where \(t\) is the number of named arguments. The remaining names, \(n_{t+1}, ..., n_m\) are given in alphabetical order. This is useful to extract the names of arguments, but does not maintain equivalence of
A,K = self.fix_to_pos(...)
self.f(*A,**dict(K))`
and
self.f(...)
in all cases.
EXAMPLE:
sage: from sage.misc.function_mangling import ArgumentFixer
sage: def sum3(a,b,c=3,*args,**kwargs):
... return a+b+c
sage: AF = ArgumentFixer(sum3)
sage: AF.fix_to_named(1,2,3,4,5,6,f=14,e=16)
((4, 5, 6), (('a', 1), ('b', 2), ('c', 3), ('e', 16), ('f', 14)))
sage: AF.fix_to_named(1,2,f=14)
((), (('a', 1), ('b', 2), ('c', 3), ('f', 14)))
Normalize the arguments with a preference for positional arguments.
INPUT:
Any positional or named arguments
OUTPUT:
We return a tuple
\((e_1, e_2, ..., e_k), ((n_1, v_1), ... , (n_m, v_m))\)
where \(n_1, ... , n_m\) are the names of the arguments and \(v_1, ..., v_m\) are the values passed in; and \(e_1, ..., e_k\) are the unnamed arguments. We minimize \(m\).
The commands
A,K = self.fix_to_pos(...)
self.f(*A,**dict(K))
are equivalent to
self.f(...)
though defaults are extracted from the function and appended to the tuple A of positional arguments. The names \(n_1, ..., n_m\) are given in alphabetical order.
EXAMPLE:
sage: from sage.misc.function_mangling import ArgumentFixer
sage: def do_something(a,b,c=3,*args,**kwargs):
... print a,b,c, args, kwargs
sage: AF = ArgumentFixer(do_something)
sage: A,K = AF.fix_to_pos(1,2,3,4,5,6,f=14,e=16); print A,K
(1, 2, 3, 4, 5, 6) (('e', 16), ('f', 14))
sage: do_something(*A,**dict(K))
1 2 3 (4, 5, 6) {'e': 16, 'f': 14}
sage: do_something(1,2,3,4,5,6,f=14,e=16)
1 2 3 (4, 5, 6) {'e': 16, 'f': 14}