Ian Barfield

it looks like it was written that way from the first version submitted by @jkff. To clarify for the lazy, the difference is between `ceil(e/epsilon)` and the current `ceil(2/epsilon)`, where...

The paper referenced in the javadoc uses `e` (as per the wayback machine anyway), but in any event, since there is no serialization issue, we could probably change it if...

Can't think of any intentional reason other than possibly reducing code noise. It seems reasonable to add and shouldn't cause any backwards compatibility issues that I can think of. I...

The algorithm and implementation look broadly correct, but there's some formatting and minor implementation details that would be nice to be addressed. eg. If merge isn't supported, it is fine...

I don't mind handling such mundane things directly, so if you'd rather, I can just pull and then clean it up myself. Just lmk

We use intellij (long live it's glory!), so we mostly just have a formatting style we import/export to share. Deviating from it is fine, but it usually covers most minor...

As a quick kinda hint, generally the start of a file goes "copywriter/license headers", "package/imports", "class javadoc". I always have intellij fold (aka kinda hide) those first two components, so...

(That said, today is kind of tightly scheduled -- think I'm already late for things -- so you might not see results from my promises until tomorrow)

I've created monstrosity like graphite targets that implement this kind of binding into a single target before. This was using grafana to make things easier, but it emits a huge...