Currently, we support indexing with floats in most places by rounding the float value to an integer. That's been called into question a number of times. There are three obvious behaviors we could go with here:

• Keep the current float-rounding approach for non-integer indices
• Only allow integer indices and throw a no method error when someone indexes with floats
• Hybrid: allow indexing with floats, but only if they're integer-valued.

We need to decide on this before v1.0; the safest approach is to disallow non-integer indexing, since that gives us the option of changing our minds later without breaking existing code.