Compare Pjf

Compare dprint-plugin-java vs spotless:PJF

Compare the output of dprint-plugin-java against spotless:palantir-java-format on a real Java project to measure formatting parity.

Usage

# Via mise task runner:
mise run compare-pjf /path/to/java/project

# Directly:
bash skills/compare-pjf/scripts/compare.sh /path/to/java/project

If no project path is given, a usage message is printed.

What it does

1. Builds the WASM plugin (cargo build --release --target wasm32-unknown-unknown --features wasm)
2. Copies .java files from the target project (excluding build/ and .gradle/ dirs)
3. Formats with dprint using the freshly-built WASM plugin
4. Formats with spotless:PJF using a temporary Gradle wrapper project
5. Normalizes both outputs (strips java.lang.* imports, sorts imports, strips trailing whitespace)
6. Compares and reports match rate with unified diffs for mismatches

Requirements

• Rust stable toolchain with wasm32-unknown-unknown target
• dprint CLI installed
• Java 21 (auto-detected via mise where java@21 or $JAVA_HOME)
• python3
• A Gradle wrapper (gradlew + gradle/) — the script copies one from the target project if available

Output

=== PJF Comparison Results ===
Total files: 473
Matching:    471 (99.6%)
Differing:   2

Diff details for mismatching files are printed below.

Key notes

• Always compare from fresh source files, not re-formatted output. Re-formatting inflates match rates.
• Normalization removes java.lang.* imports (PJF strips these; our formatter intentionally keeps them) and sorts import blocks alphabetically.
• Current baseline on sdk-javav2 (473 files): 99.6% match rate.