unroll is 10x slower than ealier

[yegor256]

Take a look: objectionary/hone-maven-plugin#118

For some reason we are 10x slower (in particular, the unroll step). I suggest we get rid of it entirely.

[yegor256]

@maxonfjvipon @volodya-lombrozo there are some ideas how we can do this. Let's keep this ticket open and put them all here. Probably, a number of PRs will be required.

[yegor256]

@volodya-lombrozo my first suspicion is that unroll uses XSD validation (or even linting) of XML files:

[INFO] >> [WARNING] XSL 'StUnhex' took 10s (over 500ms)
[INFO] >> [WARNING] XSL 'StUnhex' took 14s (over 500ms)
[INFO] >> [WARNING] XSL 'StUnhex' took 18s (over 500ms)
[INFO] >> [WARNING] XSL 'StUnhex' took 20s (over 500ms)

This is what I see on my laptop. This definitely is not a problem of XSL performance, since the files are pretty small. Seems to be a synchronisation issue.

[maxonfjvipon]

@volodya-lombrozo I'm trying to figure out why do we need explicit-data.xsl in unroll.
It converts such XMIR:

<o base="org.eolang.number">42-00-....-00</o>

to

<o base="org.eolang.number">
  <o base="org.eolang.bytes">
    42-00-....-00
  </o>
</o>

After just unphi the XMIR will be like the 2nd one (with number and bytes).
Is the 2nd XMIR okay to you for convenient converting it to back to bytecode?