don't send None to statsd daemon as metric value

[dcrosta]

If you do, you'll get errors like this in /var/log/messages:

Metric value must be a number: metric.name, None

[yannmh]

Thanks @dcrosta for fixing this. I am adding your contribution to our 0.11.0 milestone.

[dcrosta]

Cool. Is there a projected date for 0.11.0?

[yannmh]

@dcrosta, we are planning a release in about two weeks 🚢 .

Additionally, I just noticed that some recent merges are conflicting with your branch: would you mind doing a quick rebase of your changes on top of the current state of the master branch please ?

[dcrosta]

@yannmh ok, it's rebased -- waiting on Travis now.

[yannmh]

@dcrosta what would you think about concatenating those tests to a single one ?

[dcrosta]

I typically do one assertion per test -- then if the test fails, you know from the test name alone what the problem is. If you guys use a different style, I can definitely consolidate them, let me know what your preference is.

[yannmh]

Out of curiosity @dcrosta, could you share with us the use case that led to send None values via statsd ?

[dcrosta]

We have a value we pass to gauge() which is sometimes None -- our code doesn't (currently) check whether it's None before calling gauge(). Obviously we could fix this, but I thought it might make more sense to share this with others who might have the problem.

[yannmh]

Thanks for your reactivity @dcrosta.

Let's merge your changes 🚢 ! (oops seems like an other rebase is needed)

[dcrosta]

OK, rebased again.

[yannmh]

Thanks again !