RxJTango

Reactive Streams for Tango Controls — a spec-compliant implementation of reactive-streams-jvm on top of ezTangoAPI.

No Maven. No build step. Every example runs directly with jbang.

---

What is this?

Tango Controls gives you DeviceProxy — imperative, blocking, one-call-at-a-time. RxJTango wraps it in Publisher<T> so you can compose device interactions the same way you compose streams of data:

interval(500ms) → read double_scalar → ×2 + 1.5 → write double_scalar_w

becomes a single, readable chain instead of a loop with try/catch and manual threading.

The library itself depends only on org.reactivestreams interfaces. Examples use RxJava3 (Flowable, Single) as the composition layer, but any reactive-streams-compatible library works.

---

Prerequisites

• jbang — sdk install jbang or curl -Ls https://sh.jbang.dev | bash
• Java 11+
• Docker (for the Tango stack)
• GitHub Packages access for JTango artifacts (see below)

JTango artifacts — one-time setup

The Tango client library is hosted on GitHub Packages and requires a classic PAT with read:packages scope.

Add this to ~/.m2/settings.xml:

<settings>
  <servers>
    <server>
      <id>jtango</id>
      <username>YOUR_GITHUB_USERNAME</username>
      <password>YOUR_CLASSIC_PAT</password>
    </server>
  </servers>
</settings>

The id must be jtango — it matches the //REPOS alias in every example script.

---

Start the Tango stack

docker compose up -d

Three containers start: MariaDB → DatabaseDS (port 10000) → TangoTest (sys/tg_test/1). Each waits for the previous to pass its health check before starting.

Wait about 10 seconds, then verify:

docker compose ps

---

Quick demo

jbang examples/ReadAttribute.java tango://localhost:10000/sys/tg_test/1 double_scalar long_scalar

Or use the catalog aliases (run from the repo root):

jbang read-attribute@. tango://localhost:10000/sys/tg_test/1 double_scalar long_scalar
jbang stats@.          tango://localhost:10000/sys/tg_test/1
jbang pipeline@.       tango://localhost:10000/sys/tg_test/1

---

Library architecture

org.tango.client.rx
├── RxTango<T>                       Abstract base — Publisher<T>, single-shot
│   ├── RxTangoAttribute<T>          Read one attribute
│   ├── RxTangoAttributeWrite<T>     Write one attribute (emits Void)
│   └── RxTangoCommand<T,V>          Execute one command (input V, result T)
├── RxTangoAttributeChangePublisher<T>  Push publisher for CHANGE/PERIODIC/ARCHIVE events
└── TangoClient                      Fluent builder — chains RxJava Singles

RxTango subclasses are single-shot: one item per subscription. Combine with Flowable.interval(...).flatMapSingle(...) for continuous polling.

RxTangoAttributeChangePublisher is multi-value / push: lives as long as the underlying Tango event subscription.

---

Spec compliance

All publishers are verified against the reactive-streams TCK:

jbang examples/VerifySpec.java

12 applicable tests pass. Skips are expected for a single-shot publisher (the TCK has tests for publishers that emit 2+ items; those are correctly skipped).

---

Examples

See examples/README.md for the full demo playbook — ordered from the simplest read to the showstopper fluent pipeline, with narrative explanations and exact jbang commands you can run directly from IntelliJ IDEA.

---

Project layout

src/          Library source (no jbang headers — included via //SOURCES)
examples/     Runnable jbang scripts + README.md playbook
docker-compose.yml   SKAO Tango stack (MariaDB + DatabaseDS + TangoTest)
jbang-catalog.json   Catalog aliases (read-attribute, stats, pipeline, ...)