decorators tutorials

[Scott David Daniels]

Josiah Carlson wrote:
> ... I try to limit my use of decorators whenever possible, both because I 
> still have to support Python 2.3 (which doesn't support the syntax), and 
> because I find that they obfuscate what the code is doing more often 
> than not.  I will admit that they are useful as a metaprogramming 
> technique.  Just be careful.

I find them very useful in debugging (for example writing a decorator
'traced'), and have also used them to simplify code that puts a lot
of functions into a dictionary.  In the latter case, the function can
be returned unadorned, or can be stored, and only the locater for the
function returned:

     function_table = []
     def enumerated(function):
         function_table.append(function)
         return len(function_table) - 1

     @enumerated
     def triple(a, b, c):
         blah, blah, blah

     @enumerated
     def double(a, b):
         blah, blah, blah

     @enumerated
     def another(a, b, c):
         blah, blah, blah

After all of that, the identifiers triple, double, and another refer
to 0, 1, and 2 respectively, and you can call functions with the
following structure:
     function_table[double]('pi', 3.14)
This can be very useful for certain kinds of table-driven code (where
you may want to be able to efficiently store lists of functions to call
in data files).

--Scott David Daniels
Scott.Daniels at Acm.Org