Robert Kern

`vectorize` uses it mostly because `vectorize` was made to look like a function. The fact that it is implemented as a class is an implementation detail. It even describes itself...

> `vectorize` uses it mostly because `vectorize` was made to look like a function. To clarify (as this is ambiguous): because the class _itself_, not its instances, was made to...

I'm mostly not in favor of making more aliases. As a general rule, Chaco-specific tools are exposed in `chaco`, more general tools that can apply to any Enable component are...

Well, if passing NaNs to Agg causes a C++ exception, then it's _also_ a `kiva.agg` bug that our wrapper doesn't check for NaNs and raise a Python exception.

`bool(np.array([True]))` will also probably change to be consistent with `bool(np.array([True, True, ...]))` in the future, so this needs to be fixed. Additionally, `float(np.array([x]))` might also change or at least start...

It's probably the `marker_size` trait, which is `Either(Float, Array)`

Can you reduce this to a complete minimal example that the rest of us can run? And [format it so that the indentation is preserved](https://help.github.com/articles/basic-writing-and-formatting-syntax/#quoting-code)? Thanks. If I had to...

The problem is still likely to be `tf.app.run()`.

The stream might have a different encoding. If the stream is `sys.stdout`, then it might have an `.encoding` attribute that can be inspected to find this out. UTF-8 cannot be...

Yeah, only the main process can be profiled.