Don't parse timestamp from message unless we're going to check it

[mattrobenolt]

No description provided.

[kr]

Can you give a little background on the motivation for this change? I'd love to understand better why it would be needed. Thanks!

[mattrobenolt]

No real reason other than it's unnecessary to do so if we're short circuiting with no TTL. The short circuit logic was introduced in #13. And with that, if we're opting to not check TTL, we can make this more efficient.

It's also combining the slight change from #14, to match the docstrings. A TTL of 0 doesn't really make sense.

[mattrobenolt]

The >= thing also bit me in a project since I set our TTL to 0 since that's what the docs suggest, and couldn't figure out why all messages were being dropped which lead to me digging into the source.

[kr]

Ah, I didn't catch the >= change before! I went ahead and committed that separately, since it's definitely a bug where the behavior doesn't match what the docs say.

For this PR, it seems there's no behavior change, and it's purely a refactor/performance improvement. Generally, when performance-motivated changes cause an increase in complexity, I tend to shy away unless there are measurements showing that it'll make a noticeable difference for an application.

[mattrobenolt]

I mean, I'd hardly say this makes things more complex, and to me, it makes it more clear that the ts bit is only needed if the ttl bit is used. On top of that, it simplifies the if statement.

But this is bikeshedding at this point. :)