lmdb-safe: add prefix() cursor; use it in list/lookup/get

[Habbie]

Short description

Posted for first review. Bit draft still.

After updating lookup() and get() for the prefix method, I also had to update list() (as that also just prepares us for get()).

I don't think any other functions in lmdbbackend.cc need updating to work correctly; I do think a few more could benefit.

I copied the StringView define from lmdbbackend.cc but I'm not sure it's doing the right thing (in either place perhaps). It's also possible that we can fully rely on having the right Boost version these days.

Checklist

I have:

• read the CONTRIBUTING.md document
• compiled this code
• tested this code
• included documentation (including possible behaviour changes)
• documented the code
• added or modified regression test(s)
• added or modified unit test(s)
• checked that this code was merged to master

[coveralls]

Pull Request Test Coverage Report for Build 13587579875

Details

• 23 of 23 (100.0%) changed or added relevant lines in 2 files are covered.
• 11639 unchanged lines in 206 files lost coverage.
• Overall coverage increased (+3.5%) to 64.486%

---

Files with Coverage Reduction	New Missed Lines	%
ext/lmdb-safe/lmdb-typed.cc	1	89.66%
pdns/auth-catalogzone.hh	1	66.67%
pdns/dnsdistdist/dnsdist-kvs.hh	1	55.0%
pdns/dnsdistdist/dnsdist-lua-hooks.cc	1	94.59%
pdns/dnsdistdist/dnsdist-session-cache.cc	1	62.86%
pdns/dnswriter.hh	1	76.6%
pdns/dynlistener.hh	1	0.0%
pdns/inflighter.cc	1	90.48%
pdns/qtype.hh	1	86.96%
pdns/recursordist/nod.hh	1	92.59%

Totals
Change from base Build 13576694012:	3.5%
Covered Lines:	127630
Relevant Lines:	166927

---

💛 - Coveralls

[miodvallat]

The logic of the for loop in LMDBBackend::updateDNSSECOrderNameAndAuth might also be able to benefit from this, the for condition being now implied by the fact we didn't break out of it.

[Habbie]

The logic of the for loop in LMDBBackend::updateDNSSECOrderNameAndAuth might also be able to benefit from this, the for condition being now implied by the fact we didn't break out of it.

Indeed!

Unrelated thought: perhaps other cursor initializers should reset (empty) d_prefix - I don't know if we have callers that reuse cursors for different lookup purposes.

[Habbie]

this case (including type) does not even need to be a prefix query - a simple get by this key would yield the same result. But, after this PR, I'm not sure it matters much.

[miodvallat]

Resetting d_prefix sounds like something we want to do, indeed.

[Habbie]

Resetting d_prefix sounds like something we want to do, indeed.

Because I'm curious I'm going to assert emptiness first, see what dies :)

[miodvallat]

Because I'm curious I'm going to assert emptiness first, see what dies :)

[Habbie]

Because I'm curious I'm going to assert emptiness first, see what dies :)

nothing died, but it made me realise that lower_bound needs to be refactored so it can empty the prefix without breaking prefix(), which relies on lower_bound :)

[Habbie]

The logic of the for loop in LMDBBackend::updateDNSSECOrderNameAndAuth might also be able to benefit from this, the for condition being now implied by the fact we didn't break out of it.

done

[Habbie]

ready for another review round (I'll squash later)

[miodvallat]

Looking good.

[Habbie]

squashed, plus two new commits. Waiting for green :)

[miodvallat]

Good catch with the 3rd commit.

[Habbie]

Good catch with the 3rd commit.

I was renaming the short vars because clang-tidy complained and then it hit me. I'm sure the bruises won't stick around for too long.

[miodvallat]

Ah, yes, good old "commit by serendipity" technique.