Using hasattr for trait_names instead of just looking for it directly/using __dir__?

[jtratner]

This issue just came up with IPython 2.0-dev (appears to not be an issue for IPython 1.0) where our DictWrapper object raised KeyError on bad __getattr__ (which screws up hasattr). It's a badly behaved object, but it's been around for a while. We're going to fix this in 0.13, which we're releasing soonish, but this probably affects a number of versions. pandas-dev/pandas#5182

I'm not clear why defining a __dir__ method isn't enough, but a simple fix would be to go the EAFP route and just try to do it in try/except rather than doing the hasattr check. I'm sure there are other places where this could be an issue

cc @d10genes who pointed out the problematic interaction with IPython 2.0-dev.

[jtratner]

For clarity - I bring this up because I wanted to point out that objects misbehave and it's probably surprising to users; however, but it's not unreasonable to think that most things should work right with hasattr ;p

[takluyver]

We have a safe_hasattr() function in IPython.core.oinspect which can be used for this.

[jtratner]

okay, I can put up a fix for this if you want...

[takluyver]

Sure. The function should probably be moved into utils somewhere, because stuff in utils shouldn't rely on things in core.

[jtratner]

So just move from oinspect and then add an import from utils?