Daniel Martí

cc @josharian :)

Friendly ping @nvanbenschoten :) Let me know if I should do anything else to help this along. Note that https://github.com/cockroachdb/apd/pull/139 caused conflicts so I dropped the CI changes from this...

I did provide benchmarks in the commit messages with https://github.com/mvdan/benchinit, which as far as I can tell is the best way we have for now to measure init time cost....

Unfortunately the benchmark suite is rather vast when including the sub-benchmarks, so trying to run them all with `-count=6 -benchtime=1s` literally timed out after ten minutes, so I gave up...

Fair enough. Half a millisecond is still unfortunate for downstream CLI users, where every millisecond wasted in startup really does matter. But I don't know what to suggest here given...

@mxey it depends on what you mean by "same struct"; please provide a small example here because the examples at #3969 are quite large. If you mean "same struct value",...

As of the release of v0.14.0, `CUE_EXPERIMENT=toposort` is always on and can no longer be disabled. Hence, given that the experiment is concluded, I think we can close this issue...

what do you mean by "search by terms"? we could either do substring searches or "full text search" as described in https://www.sqlite.org/fts5.html.

Any particular reason why you are mixing uppercase and lowercase letters? That might be tripping up the search. We could normalize the input for this.

Ah I didn't know this, SGTM. Yes, I'll do a fix.