Issue1783803

I have attached a patch which fixes this issue.Now the in  and not in keywords will call check if the object is an instance of PyDictionary .If so has_key() will be called instead of __contains__()

hi charles.
i have looked at the patch.
its working nice.!
thanks

Your patch does indeed fix the original issue, but checking for an instance of PyDictionary in PyObject isn't the right solution here. PyObject doesn't need to know anything about its PyDictionary subclass when we can utilize polymorphism to make PyDictionary do the right thing

PyObject subclasses are made to be able to override PyObject methods when they want to implement their own custom versions (such as for performance reasons, like this issue) of Python methods

So PyDictionary needs to override __contains__ to call its own dict___contains__. PyObject's in/notin can stay the same; simply call __contains__

The following patch does this instead (I can't attach it here):

http://underboss.org/~pjenvey/patches/dict_contains_r3455.diff

It also changes dict___contains__ to call has_key instead of the final dict_has_key method -- so if a subclass overrides has_key, __contains__ will utilize the overridden version 

a note of warning, Python doesn't properly define what other methods behavior is influenced by overriding some other methods, indeed in CPython the usual answer in most cases is none:

Python 2.4.3 (#1, Apr  7 2006, 10:54:33) 
[GCC 4.0.1 (Apple Computer, Inc. build 5250)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> class X(dict):
...    def has_key(self, k):
...        return (k % 2) == 0 
...                         
>>> 
>>> x=X()
>>> x.has_key(4)
True
>>> 4 in x
False
 
the risk in not following CPython in this and try to be more useful, is to easily create cases where in some situations you get what you wanted but in other overriding cases you get infinite recursions.

Thanks for the valuable tip pjenvy.I have attached your patch .It does make a lot of sense to utilise polymorphic capabilties of java (though there are several places in jython where this rule is not followed).
I also ran Pedroni's code sample with both my patch and pjenvy's patch in jython as also python.I get the same output in all cases .
File Added: dict_contains_r3455.diff

thanks for catching that pedronis

applied in r3473 minus the has_key change

Could this patch be applied to 2.2 branch so that possible 2.2.1 will have the fix also?

Using "key in dictionary" format is considered more pythonic than "dictionary.has_key(key)" and on CPython it's also faster (see e.g. [1]). There's thus probably a lot of code using that format and it's a shame that it's so much slower in Jython 2.2. 

[1] http://python.net/~goodger/projects/pycon/2007/idiomatic/handout.html#use-in-where-possible-1

Charlie has merged this to the 2.2 branch in r3484