AllInvalid arrays have non empty distinct info

[robert3005]

This is required to handle cases where we are attempting to dict compress and
a sample ends up being all null

Signed-off-by: Robert Kruszewski github@robertk.io

[codspeed-hq]

Merging this PR will improve performance by 29.94%

⚡ 2 improved benchmarks
✅ 1120 untouched benchmarks
⏩ 1455 skipped benchmarks¹

Performance Changes

	Mode	Benchmark	BASE	HEAD	Efficiency
⚡	Simulation	take_10k_random	251.2 µs	193.3 µs	+29.94%
⚡	Simulation	take_10k_contiguous	315 µs	257 µs	+22.56%

---

_{Comparing rk/fixcompressor (b485cc0) with develop (ae906c7)}

Footnotes

1. 1455 benchmarks were skipped, so the baseline results were used instead. If they were deleted from the codebase, click here and archive them to remove them from the performance reports. ↩

[connortsui20]

distinct count has to be 0, this is an invariant that other code relies on.

Let me check where

[connortsui20]

we use it in the constant schemes where if the distinct count is not greater than 1, it must be exactly 1.

I think it is actually fine though since that exact code path wont happen, but I would prefer if we just set distinct count to 1 and then add a big TODO and red flag that this is a hack. Once I finish the vector search stuff I can revisit this and come up with a more robust solution while adding all the things that would be good to have in the compressor (#7216)

[robert3005]

not sure why setting any value is meaningful here, the idea is that they will not look at it and we just need to have a present value. This is also internally consistent where all null arrays aren't constant

[connortsui20]

its just that I rely on that fact at other callsites, so it might panic there instead. I guess it doesn't really matter either way since both of these fixes are technically incorrect

eh

[connortsui20]

i will fix this next week