On Oct 27, 3:14 am, Andrew West <[EMAIL PROTECTED]> wrote:
> Probably a bit of weird question. I realise decorators shouldn't be
> executed until the function they are defined with are called, but is
> there anyway for me to find all the decorates declared in a file when
> I import it? Or perhaps anyway to find the decorators by loading the
> file by other methods (with out simply parsing it by hand).
There is no need for having *parser Angst* in Python. Once you
understand the syntactical structure of the language it's easy to move
on. I show you how to deal with it using EasyExtend for grepping parse
trees ( which is much easier than using regexps for that purpose ).
Alternatively you can use the compiler package. It hides some low
level stuff being present in EE on this level. You need to decide how
you want to reduce mental overhead and achieve simplicity that "fits
your brain".
EasyExtend is available at www.fiber-space.de
Here is a small tutorial:
# purpose -- seeking for decorators and decorated functions in Python
2.5 stdlibs contextlib.py
import inspect
import parser
import symbol
import token
# just import contextlib, keep source and create a parse tree
import contextlib
src = inspect.getsource(contextlib) # keep source
cst = parser.suite(src).tolist() # create concrete syntax tree as
a nested list
# so far its all standard lib stuff...
# ... now something new:
from EasyExtend.csttools import find_node, find_all
functions = find_all(cst, symbol.funcdef)
# You need to know how these nodes are structured to get more
information out of them
# This can be looked up in EasyExtend/Grammar which contains a copy of
Pythons Grammar or
# in the same file in Pythons source distribution. Here are the
relevant rules:
# decorator: '@' dotted_name [ '(' [arglist] ')' ] NEWLINE
# decorators: decorator+
# funcdef: [decorators] 'def' NAME parameters ':' suite
# We see immediately that it's not all that simple. Decorators can be
methods and the first name
# in dotted-name can be the name of an object. For demonstration
purposes we simplify our
# assumptions and suppose to have plain function decorators
func_info = {}
for f in functions:
decorator_info = set() # store decorator information here
decorators = find_node(f, symbol.decorators) # we do *not* want
all decorators
# since there might
be those related to
# a closure that is
handled separately
if decorators:
decos = find_all(decorators, symbol.decorator)
for deco in decos:
# token.NAME has structure [1, 'name', line_no]
deco_name = find_node(deco, token.NAME)[1]
args = find_node(deco, symbol.arglist) # ... we
analyze args later
decorator_info.add((deco_name, args))
f_names = find_all(f, token.NAME, level = 1) # set level
information otherwise you
# get all names
defined anywhere in
# funcdef
func_name = f_names[1][1] # the first name is 'def', the
second one the func_name
func_info[func_name] = decorator_info
# Note: you can move from syntax tree representation straightforward
back to a textual
# representation. This requires not much work but the following
import is nevertheless
# conceptual overhead I do not try to explain in detail here:
from EasyExtend.fibers.zero.fiber import unparse # zero represents
*Python* in the larger
# context of
EasyExtend applications
# unparse
transforms a cst of a particular
# grammar back to
source code
for func_name, deco_info in func_info.items():
for (deco_name, deco_args) in deco_info:
print func_name, deco_name, (unparse(deco_args) if deco_args
else "()")
# I get the rather unspectacular result
# nested contextmanager ()
# You might verify this manually...
--
http://mail.python.org/mailman/listinfo/python-list
