Surprising table equality behavior

[dubois]

import lupa
r = lupa.LuaRuntime()
r.globals().test == r.globals().test
False

I can supply a patch that adds richcmp and hash to _LuaTable.

[scoder]

Sounds like a good idea.

[dubois]

I never got around to adding the patch :-P. Here it is!
https://gist.github.com/dubois/9478877

[scoder]

Thanks. A couple of comments:

• looks very good functionality wise
• lacks tests
• I prefer pull requests :)
• "richcmp()" is bidirectional, so you can't be sure the first argument is "self", it might be the second (which is, I guess, why you thought you'd have to type the first argument explicitly)
• "uintptr_t" appears to be C99. Not sure I want to require that. There's also the "stdint.h not in MSVC" issue. Tricky ground.
• the comparison operators should use the CPython constants (from cpython.object cimport Py_EQ, ...), not plain integers. There's also no need to explicitly return True/False, just return the comparison result.
• there's a special type "Py_hash_t" that Cython defines as return value of "hash()" (and that it backports for Py<3.2)

[SoulSen]

I'm pretty sure this occurs with all _LuaObject, since I've experienced this with _LuaFunction too

[scoder]

Don't know why this was closed. Doesn't look like it's resolved.
PR welcome, as described above.