More efficient working with headers

[plokhotnyuk]

Currently, it is quite expensive in terms of burned CPU and eaten memory bandwidth during parsing, validation

...and serialization

This PR reduces that overheads and improves working with headers.

[plokhotnyuk]

@adamw @kciesielski Could you please review and merge this PR?

Would you be interested in subsequent improving working with headers and paths (as example replacing regexp and split calls by more efficient implementations)?

[kciesielski]

@plokhotnyuk Thank you! Performance improvements like this one are much appreciated, and we are definitely interested in subsequent PRs. Unfortunately, the core code has to compile for Scala 2.12 as well. Could you adjust your changes to this requirement?

[plokhotnyuk]

@kciesielski Sorry, I had missed to check that sbt +core/test doesn't run tests across Scala versions.

I've force pushed an updated commit with fixes that I had tested locally by the following sequence of commands:

sbt ++2.12.18 core/test
sbt ++2.13.11 core/test
sbt ++3.3.0 core/test

[kciesielski]

@plokhotnyuk We use the sbt-projectmatrix plugin, which generates dynamic projects for each scala version. The core project is for Scala 2.13, core2_12 for 2.12 and core3 for Scala 3. The failed compilation happens on core2_12/compile (compile scope).

[plokhotnyuk]

@kciesielski Thanks, for the heads up! Now core2_12/compile compiles without any errors.

[kciesielski]

Awesome, thanks again!

[adamw]

regarding the TODO - yeah it looks weird, why would we need to reverse; if anything, we want to preserve the original ordering?

[plokhotnyuk]

Originally reversing was introduced by recursive version of previous implementation.

So my question in the TODO message was about need of reversion for the result list.

[adamw]

There was another tail-recursion with an accumulator which reversed the list - I moved the reverse to the more logical place: ea43d96

Thanks for the PR, btw :)