[Python-3000] Draft pre-PEP: function annotations (original) (raw)

Talin talin at acm.org
Sat Aug 12 06:49:52 CEST 2006

Phillip J. Eby wrote:

Not at all. A and B need only use overloadable functions, and the problem is trivially resolved by adding overloads. The author of C can add an overload to "A" that will handle objects with 'next' attributes, or add one to "B" that handles tuples, or both.

I'm still not sure what you are talking about - what is being overloaded here?

Let me give you a better example. Suppose I have a 'docstring' annotation and a 'getopt' annotation. The docstring annotation associates a string with each argument, which can be inspected by an external documentation scanner to produce documentation for that argument.

Thus:

def myfunc( x : "The x coordinate", y : "The y coordinate" )
   ...

The 'getopt' annotation is used in conjunction with the 'getopt' decorator, which converts from command-line arguments to python method arguments. The idea is that you have a class that is acting as a back end to a command-line shell. Each method in the class corresponds to a single command. The annotations allow you to associate specific flags or switches with particular arguments. So:

class MyHandler( CommandLineHandler ):

@getopt
def list( infile:"i" = sys.stdin, outfile:"o" = sys.stdout ):
   ...

With the getopt handler in place, I can type the following shell command:

list -i <infile> -o <outfile>

If either the -i or -o switch is omitted, then the corresponding argument is either stdin or stdout.

Additionally, the getopt module can generate 'usage' information for the function in question:

Usage: list [-i infile] [-o outfile]

Now, what happens if I want to use both docstrings and the getopt decorator on the same function? The both expect to see annotations that are strings! How does the doc extractor and the getopt decorator know which strings belong to them, and which strings they should ignore?

-- Talin

More information about the Python-3000 mailing list