Plan on c++ implementation (at least on base functions)

[LivInTheLookingGlass]

I don't know how to do network I/O in C++, but I do know it's possible to construct and interpret these messages, and I know that my callback structure will work.

At the very least, this can be built as a python C extension. Or more accurately, it could be entirely based in C++ and rely on python for I/O (I believe that's possible).

A definite thing to happen is that the python version will include C++ files at this point. So every import will be of the form:

try:
    import cbase as base
except ImportError:
    import base

And setup.py would have a flag for including C extensions. Roughly:

if sys.argv[-1] == "--universal":
    setup_without_extensions() 
else:
    try:
        setup_with_extensions() 
    except:
        setup_without_extensions() 

Most of that is untested, since I'm just now learning how to use C++ this way, but I believe that will work.

[LivInTheLookingGlass]

C++ implementation (not yet pushed) now has everything except compression methods. Has flags, protocol, pathfinding_message. This has, however, revealed some optimizations to make. Namely, as follows:

class example(object):
    def __init__(self, val_1, val_2):
        self.val_1 = val_1
        self.val_2 = val_2
        self.__cache = {}

    @property
    def id(self):
        if self.__cache.get('val_1') == self.val_1 and self.__cache.get('val_2') == self.val_2:
            return self.__cache.get('id')
        self.__cache.update({'val_1': self.val_1,
                             'val_2': self.val_2,
                             'id': self.__id_calc()})
        return self.__cache.get('id')

    def __id_calc(self):
        return hashlib.sha384(self.val_1 + self.val_2).hexdigest()

This saves CPU time in return for a mild hit to memory. In the case of a hash, this would likely save memory in the long run, as each hash calculation consumes more memory than the string result.

[LivInTheLookingGlass]

Now implemented and pushed. No compression is yet supported, though zlib is included as a submodule.

[LivInTheLookingGlass]

C++ is now finished (for the base layer). At some point it will get more documentation, as well as compression and networking.

At the moment, it's ~0.1 ms faster than python at constructing messages.