Tuesday, January 20, 2015

Webcast Wednesday

Be there: http://www.oreilly.com/pub/e/3255

Of course, I've got too many slides. 58 slides for a 60 minute presentation. That's really about 2 hours of material. Unless people have questions, then it's a half-day seminar.


I think I've gone waaaay too far on this. But it's my first one, and I'd hate to burn through all eight slides, take a few questions and be done too soon.

If this goes well, perhaps I'll see if I can come up with other 1-hour topics.

I worry a great deal about rehashing the obvious.

On the other hand, I'm working with a room full of newbies, and I think I could spend several hours on each of their questions.

And straightening out their confusions.

Case in point. Not directly related to the webcast.

One of my colleagues had seen a webcast which described Python's &, |, and ~ operators, comparing  them with and, or and not.

I'm not 100% sure, but... I think that this podcast -- I'm getting this second-hand; it's just hearsay -- showed that there's an important equivalence between and and &.

This is true, but hopelessly obscure. Since & has a higher priority than the comparison operators, there will be serious confusion when one fails to parenthesize properly.

Examples like this abound:

>>> 3 == 3 & 4 < 5
>>> (3 == 3) & (4 < 5)


Further, the fact that & can't short-circuit had become confusing to the colleague. I figured out some of what was going on when trying to field some seemingly irrelevant questions on "Why are some operators more efficient?" and "How do you know which to use?"

Um. That's not really the point. There's no confusion if you set the bit-fiddling operators aside.

The point is that and, or, not, and the if-else conditional expression live in their own domain of boolean values. The fact that &, |, ^, and ~ will also operate on boolean values is a kind of weird duplication, not a useful feature. The arithmetic operators also work on booleans. Weirdly.

The Python rules are the rules; it makes sense for True&True to yield True. Results depend on the operands. It would be wrong in that sense for True&True to be 1. But it would also fit the concept of these operators a little better if they always coerced bool to int. This happens for * and +: True+True == 2.

Why can't it be true for & and |? It would reduce potential confusion. 

I'm sure the person who implemented __and__(), __or__(), __xor__(), and __invert__() was happy to create a parallel universe between and and &. I'm not sure I agree.

And perhaps I should have a webcast on Python logic. It seems like a rehash of fundamentals to me. But I have colleagues confused by fundamentals. So perhaps I'm way wrong about what's fundamental and what's useful information.