Strange (?) list comprehension behavior

I am a Python newbie, using IDLE, writing a lexical analyzer for a small
expression language in Python 2.2.3.

class PSILex:
ops = ["~", "!", "@", ... "|", "||", ... ]
...
def __findAtom(self ):
result = ""
ops = [op for op in PSILex.ops if self.text.start swith(op)]

.... results in a traceback and "TypeError: expected a character buffer
object." In trying to figure out what was going on, in the Python Shell I
subsequently tried
[color=blue][color=green][color=darkred]
>>>[op for op in PSILex.ops if "|| hello".startswi th(op)][/color][/color][/color]

with the same result. Printing type(self.text) immediately prior ot the
above line proves that that attribute's type has not been compromised, so
the problem appears to be with 'op'.
[color=blue][color=green][color=darkred]
>>>[op for op in PSILex.ops if op.startswith(" || hello")][/color][/color][/color]

yields "AttributeError : 'int' object has no attribute 'startswith',"
however:
[color=blue][color=green][color=darkred]
>>>[type(op) for op in PSILex.ops][/color][/color][/color]

produces a list of <type 'str'> objects, as would be expected. So, what is
op? Is it a string, an int, or something else? It appears that the
interpreter may be confused. I persuaded the interpreter to cooperate by
injecting an explicit evaluation of type(op):

ops = [op for op in PSILex.ops if type(op) == type("") and \
self.text.start swith(op)]

and this does what I want, and what I would expect without 'type(op) ==
type("") and '.

Can someone explain what is happening here? Or should I send my code to
ww.python.org as a "defect report?"

Regards,
George

_______________ _______________ _______________ _______________ _____
Protect your PC - get McAfee.com VirusScan Online