Too many refresh events published from ConfigWatch

[dahamr]

We have a reasonably complex set of services all of which are configured to watch for consul changes. Because of #183, I've been looking for ways to reduce the number of refresh events.

What I've noticed is that refresh events are issued when no KVs related to the service have changed. I found that if a prefix does not exist, the X-Consul-Index value could match the ModifyIndex of a key in a completely different area. In that case, the query made to consul returns a 404 along with new X-Consul-Index. The refresh events I am questing all seem to be related to missing prefixes.

I'm wondering if it is necessary to publish a refresh event in this scenario where the value is null and the query to check for the prefix returns a 404? I tested the minor change to ConfigWatch to skip the refresh if the value is null and it seems to work like I'd want.

[spencergibb]

Sweet, how about a PR?

[dahamr]

The code change isn't really interesting but I submitted it. We are at 1.0.2.RELEASE but the PR is for master.