Issue1631

Workaround as discussed in my email on jython-users:

from java.util import LinkedHashMap

def monkeypatch_method_if_not_set(cls):
    def decorator(func):
        if not hasattr(cls, func.__name__):
            setattr(cls, func.__name__, func)
        return func
    return decorator

@monkeypatch_method_if_not_set(LinkedHashMap)
def iteritems(self):
    return ((entry.getKey(), entry.getValue()) for entry in self.entrySet())

# equivalent functionality - we will do a similar type of guard
try:
    LinkedHashMap.update
except AttributeError:
    LinkedHashMap.update = LinkedHashMap.putAll
    LinkedHashMap.iterkeys = LinkedHashMap.keySet
    LinkedHashMap.itervalues = LinkedHashMap.values

# etc

Implement something like this directly in the proxy generation.

I don't see why this is a "bug" in Jython.

This is a bug because a goal of Jython is to create a tight integration (among others) between standard Java collections (Map, List, Set) with equivalent Python collection types (dict, list, set).

Another missing method is the keys method. But not __iter__ or values!

We should systematically test this, instead of the limited testing we have now. Maybe we can readily reuse code from test_dict or test_userdict?

I should also point out that this inconsistency came up in this excellent blog post
http://blogs.mulesoft.org/easily-optimizing-python-extending-with-java/

It would be really nice to fix this issue for beta 4

I'll have a patch by tonight.

Here you go.

Santoso, thanks for the patch! This looks like a good start. I think the next step is to look at how we can re-use existing "generic" map testing, per msg5921

@zyasoft Agreed. I have a bitbucket pull request [1] I can put further improvements on.

[1] https://bitbucket.org/jython/jython/pull-request/42/fix-issue-1631-javautilmap-derived-classes/diff

Adapting DictTest from test_dict to use java.util.HashMap, with multiple failing tests.

Not sure if it's in the scope of this issue to try to address all of them.

Santoso has pretty much completed this work, it's now just pending a detailed review.

Just wanted to follow up here in this bug: Santoso's PR is great and complete, however it's unfortunately exposing bugs in certain modules (such as threading) in Jython's stdlib implementation. These bugs rely on buggy behavior in the current support of java.util.Map.

I plan to address by adding a new jythonlib module that will allow for building dict, list, or set with arbitrary backing collections supporting respectively java.util.Map, java.util.List, and java.util.Set. An example is using Guava's MapMaker as a builder for dicts. Example:

import jythonlib
try:
    # jarjar-ed version
	from org.python.google.common.collect import MapMaker
except ImportError:
    # dev version from extlibs
	from com.google.common.collect import MapMaker

# note the use of the bound method
builder = jythonlib.dict_builder(MapMaker().concurrencyLevel(4).weakKeys().weakValues().makeMap)
# initialize a new dict using the above ConcurrentMap definition
d1 = builder(x=1, y=2, z=3)

Unblocked as of http://hg.python.org/jython/rev/7a96264b09ab

The corresponding PR has been merged in (https://hg.python.org/jython/rev/864bbec6ddb5), so at this point we are only lacking comparison methods other than __eq__.

Fixed as of https://hg.python.org/jython/rev/fc24f60fb32c