Is it reasonable for the "MetricRegistry.histogram()" method to state that it uses a ExponentiallyDecayingReservoir?

[David Karr]

I've just started looking at using the DropWizard Metrics library (odd that this entity is referred to as both "DropWizard" and "codehale") because another team in my org is using it successfully, and I'll likely adopt it for other work.

I submit doc PRs for almost every framework I've ever used, and I already submitted one for this.  However, I'm sometimes cautious to make assumptions about what's reasonable in new frameworks.

I saw the mention of the "histogram" method on the Core doc page. I also looked at this method in the javadoc, and also at the Histogram class. After digging into the code a bit, I saw that the "histogram" method creates a Histogram using the ExponentiallyDecayingReservoir, among the several choices of a Reservoir implementation.  Of the choices, this seems like the most reasonable choice, but my question is: Would it be reasonable to mention this in the Core doc page, and in the javadoc for the "histogram" method?

Again, I can just submit PRs for these changes, but I'm not going to go down this path if these are considered "trivial" and "obvious" details for typical new users of this framework.

[Marshall Pierce]

A lot of people get caught off guard by this, FYI, because that reservoir implementation has some odd behavior. Maybe it would be helpful to expose that detail in `histogram()` and friends? Or maybe that would just be further cementing a default choice of arguable validity. Anyway, one of the goals for metrics v4 is to make such defaults configurable, so the problem will be at least different in v4.

> --
> You received this message because you are subscribed to the Google Groups "metrics-user" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to metrics-user...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.