Josh Bleecher Snyder

@moreati I just discovered this was still open and unmerged. Sorry about that. Any reason I shouldn't merge now?

I think it’s done enough to be extracted without much pain. The name is a bummer though. Ideas: * wide.Uint128 (wide data types) * c2.Uint128 (twos-complement) * universe.Uint128 (things that...

Another reason to do this (eventually) is that they'd be useful in finishing up a SMART implementation. My personal favorites, some days later, are `wide.Uint128`, `uint128.T`, and `int128.U`.

Both this (return an error instead of swallowing invalid input to IPSetBuilder) and the direction in #167 (fix an invalid input to IPPrefixFrom rather than return an error) feel locally...

Oh, and I see this is also now being discussed in #178. :)

I also hit this, and am motivated to fix it. Another workaround is to use `pq.Array`, even though everything else is using pgx. Unfortunately, this requires editing the generated code,...

Scratch that, it doesn't quite compile cleanly out of the box. But this does: https://github.com/josharian/pqarray.

Another workaround is to use a column go_type override with an appropriate type from github.com/josharian/pqarray, and then write your own conversion helpers. Ugly, but still retains a bit of type...

> Maybe we can just improve the implementation? Perhaps. The reason I went with a callback is that this is in some quite performance sensitive code. Writing everything into a...

Well, the API for the models is well-defined: It takes in a set of pixels in a particular format and returns an output in a particular format. You can substitute...