⚠️ GitHub issue #37199 has been automatically assigned in GitHub to PR creator.

General looks ok to me, cc @bkietz

I think it'd be better to avoid tying this to the concrete Array classes. The span would expose values masked by null bits which could lead to unfortunate surprises. FWIW these classes already expose begin, end, operator[] yields std::optional for usability.

I think it'd be more useful to expose such a helper at the buffer or array data level, for example as

• Buffer::span analogous to Buffer::data_as
• ArraySpan::GetSpan analogous to ArraySpan::GetValues
  (along with equivalents for mutable buffers)

I think it'd be more useful to expose such a helper at the buffer or array data level

You have a valid point and I think that makes a lot of sense.
cc @pitrou what are your thoughts on this as I might have misunderstood the original intent of this issue.

I think it'd be more useful to expose such a helper at the buffer or array data level, for example as

That sounds ok to me.

Sorry for the abuse of CI, my local env was messed up accidentally :(

After merging your PR, Conbench analyzed the 6 benchmarking runs that have been run so far on merge-commit 0552217.

There were no benchmark performance regressions. 🎉

The full Conbench report has more details. It also includes information about 2 possible false positives for unstable benchmarks that are known to sometimes produce them.