drop rules_java dependency

Start by reading MODULE.bazel and locating the current rules_java dependency and the PlantUML targets that use it. The body proposes a standalone Bazel tool, but marks its draft as unreviewed and leaves both artifact checksums as placeholders. Confirm the intended approach and supported platforms with maintainers; done means the dependency is removed and PlantUML remains usable by the project’s documentation workflow.

rules_java is a FAT dependency. Investigate dropping it.
Unreviewed ChatGPT below.

---

PlantUML as a standalone Bazel tool

Provide PlantUML as a single executable Bazel target, bundling a pinned Java Runtime (JRE) and the PlantUML JAR.

Goals:

• No rules_java dependency
• No additional Bzlmod module dependencies
• Reproducible Java and PlantUML versions
• No system-wide Java installation required
• Usable through bazel run and as a tool in other Bazel targets
• Compatible with Sphinx / sphinxcontrib-plantuml

1. Download JRE and PlantUML

In MODULE.bazel:

http_archive = use_repo_rule(
    "@bazel_tools//tools/build_defs/repo:http.bzl",
    "http_archive",
)

http_file = use_repo_rule(
    "@bazel_tools//tools/build_defs/repo:http.bzl",
    "http_file",
)

http_archive(
    name = "plantuml_jre",
    urls = [
        "https://github.com/adoptium/temurin21-binaries/releases/download/jdk-21.0.9%2B10/OpenJDK21U-jre_x64_linux_hotspot_21.0.9_10.tar.gz",
    ],
    sha256 = "<JRE_SHA256>",
    strip_prefix = "jdk-21.0.9+10-jre",
    build_file_content = """
package(default_visibility = ["//visibility:public"])

exports_files(["bin/java"])

filegroup(
    name = "runtime",
    srcs = glob(["**"]),
)
""",
)

http_file(
    name = "plantuml_jar",
    urls = [
        "https://github.com/plantuml/plantuml/releases/download/v1.2026.9/plantuml-1.2026.9.jar",
    ],
    sha256 = "<PLANTUML_SHA256>",
    downloaded_file_path = "plantuml.jar",
)

Replace SHA-256 placeholders with verified checksums from the respective release artifacts.

Both dependencies are registered as external repositories without introducing additional Bzlmod module dependencies.

2. Create an executable Starlark rule

Create tools/plantuml.bzl:

def _runfile_path(file):
    path = file.short_path
    return path[3:] if path.startswith("../") else path

def _plantuml_impl(ctx):
    launcher = ctx.actions.declare_file(ctx.label.name)

    ctx.actions.write(
        output = launcher,
        is_executable = True,
        content = """#!/usr/bin/env bash
set -euo pipefail
root="${{RUNFILES_DIR:-$0.runfiles}}"
exec "$root/{java}" -Djava.awt.headless=true -jar "$root/{jar}" "$@"
""".format(
            java = _runfile_path(ctx.executable.java),
            jar = _runfile_path(ctx.file.jar),
        ),
    )

    runfiles = ctx.runfiles(
        files = [ctx.executable.java, ctx.file.jar],
        transitive_files = ctx.attr.jre[DefaultInfo].files,
    )

    return DefaultInfo(
        executable = launcher,
        runfiles = runfiles,
    )

plantuml_binary = rule(
    implementation = _plantuml_impl,
    executable = True,
    attrs = {
        "java": attr.label(
            default = "@plantuml_jre//:bin/java",
            executable = True,
            cfg = "exec",
        ),
        "jre": attr.label(
            default = "@plantuml_jre//:runtime",
        ),
        "jar": attr.label(
            default = "@plantuml_jar//file",
            allow_single_file = True,
        ),
    },
)

The rule:

1. Generates an executable shell launcher.
2. Includes the complete JRE in its runfiles.
3. Includes the PlantUML JAR.
4. Executes the bundled Java runtime with -jar.
5. Forwards all command-line arguments to PlantUML.

This implementation targets Linux with runfiles directories.

3. Expose PlantUML as a Bazel target

Create tools/BUILD.bazel:

load(":plantuml.bzl", "plantuml_binary")

plantuml_binary(
    name = "plantuml",
    visibility = ["//visibility:public"],
)

Usage:

bazel run //tools:plantuml -- -version

Generate an SVG:

bazel run //tools:plantuml -- -tsvg diagram.puml

4. Integrate with Sphinx

For sphinxcontrib-plantuml, the executable is configured through:

extensions = ["sphinxcontrib.plantuml"]

plantuml = "/path/to/bazel/plantuml"

The path must point to the executable inside the Bazel execution environment.

When integrating with a custom Bazel docs() rule:

1. Declare //tools:plantuml as an executable tool dependency.
2. Include the tool's runfiles in the documentation runtime.
3. Resolve the executable path at runtime.
4. Pass that path into the Sphinx configuration.

Sphinx does not need to know that PlantUML is implemented in Java.

5. Benefits and limitations

Property	Result
rules_java required	No
Additional Bzlmod dependencies	None
Java installation on host	Not required
Java version	Pinned
PlantUML version	Pinned
Usable with bazel run	Yes
Usable in other Bazel targets	Yes
Remote execution	Possible with appropriate tool/runfiles integration
Cross-platform	Requires platform-specific JREs
Graphviz	May be required for some diagram types

6. Alternative: Native PlantUML executable

PlantUML also publishes experimental native executables.

These could eliminate the Java Runtime entirely:

//tools:plantuml
    └── Native PlantUML executable

However, the JAR + pinned JRE approach is preferable if compatibility and reproducibility are more important than minimizing the runtime size.

References

• Bazel Repository Rules
• Bazel Runfiles
• Eclipse Temurin JRE releases
• PlantUML releases
• sphinxcontrib-plantuml