Fix shrinking of floats with one bound

[Zac-HD]

Fixes #1212, fixes #1683 and fixes #1707.

Unfortunately fixing shrinking for floats() with two bounds will be much more involved. I'd prefer incremental to no progress though, so I've written up #1704 as a separate issue.

[DRMacIver]

I'm actually unsure whether I do prefer incremental progress here - given that we have a general fix proposed in #1704 and it doesn't share any code to speak of with this PR, do we actually still want to merge this?

[Zac-HD]

"incremental" is probably the wrong word - this is a full fix for the shrinking problem with one bound. floats() with two bounds uses a different SearchStrategy subclass internally, with completely different generation and shrinking behaviour. The change proposed in #1704 would complement but not replace this one.

[DRMacIver]

The change proposed in #1704 would complement but not replace this one.

Hmm. What I was imagining there, which maybe wasn't what you intended, was having a single wrapper strategy called say FloatShrinkingStrategy which wraps some underlying strategy and filter function (or set of float specific constraints) and automatically produces a float strategy that shrinks correctly. This would then commonize shrinking logic between all three float strategies, and potentially would allow us to (later) improve the generation of floats to boot.

[Zac-HD]

Oooh, that sounds great.

I'll open a new PR with just the incidental fixes which have... accumulated. TBH I'm probably violating the multi-purpose-PR rule again 😇

[DRMacIver]

TBH I'm probably violating the multi-purpose-PR rule again 😇

You are, but I decided not to make an issue of it until we'd decided whether you wanted to continue with this PR. 😉