Replace the in-tree YAML emitter with the yaml-flock dependency

[ctgnz]

Replaces jmsfx-generator's in-tree YAML emitter with a dependency on yaml-flock.

Five files go — JmsfxYamlFactory, JmsfxYamlGenerator, JmsfxPrettyPrinter, YamlFlowStyle, YamlForceQuote. JmsfxParser stays, rebuilt on FlockYamlFactory, since it configures a mapper for jmsfx's own model.

The swap is mechanical because the library's API is compatible by construction: YamlForceQuote is identical, YamlFlowStyle widens @Target from TYPE to {TYPE, FIELD, METHOD} so every existing use still compiles, and the factory constructor signature matches. Import swap plus one class name, across 19 files.

⚠️ This cannot pass CI yet, by design

yaml-flock is still 1.0.0-SNAPSHOT and exists only in a local .m2. A GitHub runner has no way to resolve it, so the build here will fail on dependency resolution until one of:

1. yaml-flock 1.0.0 is published to Central and this PR switches yaml.flock.version to 1.0.0 — this is what Replace the in-tree YAML emitter with the yaml-flock dependency #155 always planned, and the simplest path. The PR waits.
2. yaml-flock snapshots get published to Central's snapshot repository and jmsfx adds that <repository> — more moving parts, and it makes jmsfx's build depend on a snapshot, which is not something to carry long-term.

Option 1 unless there's a reason to prefer otherwise. The version sits behind a yaml.flock.version property so that switch is one line.

Correcting a premise in #155

The issue anticipated replacing JmsfxPrettyPrinter's two hardcoded field lists with a @YamlBlockStyle annotation, and described country as a latent trap that removing one @YamlFlowStyle would spring.

Neither holds. JmsfxPrettyPrinter has zero references anywhere in the project — never instantiated, never installed on a mapper. Jackson consults a DefaultPrettyPrinter for JSON output only, and YAMLGenerator emits through SnakeYAML without ever asking it. So both field lists were unreachable, not merely inert, and country was never going to spring: removing that @YamlFlowStyle would change nothing.

So this adds no annotation at all. Adding one would introduce behaviour that never existed. (This is the same discovery that killed the equivalent class in yaml-flock itself, where a probe showed identical output for every shape with the printer installed and removed.)

Acceptance criteria, both verified

All three model.yml files round-trip byte-identically through the swapped emitter:

IDENTICAL  jmsfx-standard     (451788 bytes)
IDENTICAL  jmsfx-historical   (40018 bytes)
IDENTICAL  jmsfx-battleorder  (1043 bytes)

Checked by reading each real file and calling writeLibraryModel in memory, comparing bytes — deliberately nothing written into the repo. Their on-disk SHA-256 hashes are also unchanged after a full regeneration of all three libraries.

Regeneration plus one spotless:apply lands on the committed sources. Regenerating left 133 files differing in raw output; one spotless:apply returned every one to the committed source, leaving a clean tree. That is the property #127 established, still holding.

Build

• mvn -Pstandard verify — BUILD SUCCESS
• mvn -Phistorical verify — BUILD SUCCESS

Spotless check included in both, so the re-sorted imports are compliant (io.github.ctgnz.jmsfx.* sorts before io.github.ctgnz.yamlflock.* in the trailing group).

Addresses #155 — not closing it, since the switch to 1.0.0 from Central is part of the same issue.

🤖 Generated with Claude Code