Intro
Decorators were introduced in Python 2.4 but remain a somewhat obscure feature of the language. A decorator essentially is a function that you apply to another function, replacing it with the return value of the decorator function:
def d(f):
...
@d
def g():
...
This is equivalent to:
def d(f):
...
def g():
...
g = d(g)
More background
here.
The use case
As part of my research project, I’m testing several ways to explain side effects of drugs. So I made a class that loads all the relevant data into memory and that contains methods to fit the data. (I thought about creating subclasses replacing the “fit” method of the base class. However, the data that’s being loaded into memory is the same for all instances of the subclasses, so I don’t want to reload it every time. Thus, I put multiple methods into the base class.)
At first, I used introspection to identify the relevant functions by name (using dir) and then calling them. This worked fine, but it doesn’t make it particularly easy to disable some of the functions. Enter decorators: I wrote a decorator that takes a function, puts it into a list (which is a global variable) and returns the function unchanged. To call every fitting function, I can just iterate over the list. If I want to disable a function, I just comment out one line. Yay!
# with the help of decorators, keep track of the functions
# we would like to use for fitting
fit_functions = []
def fit_func(f):
fit_functions.append(f)
return f
class C(object):
def __init__(self):
...
@fit_func
def fit1(self):
"""This is the first function"""
...
@fit_func
def fit2(self):
"""This is the second function"""
...
## This function is disabled by putting the decorator in a comment
# @fit_func
def fit3(self):
"""This is the third function"""
...
def main():
c = C()
for f in fit_functions:
fit = f(c)
print fit, f.__doc__
...
Another neat trick is that the
docstring of each function is accessible as an attribute, so that each fit comes together with a description.
Labels: python
