oc2_hlapi_v2 0.2.0

A client for the new version of OpenComputers II's HLAPI
Documentation

Crates.io Version docs.rs Crates.io MSRV

OC2 HLAPI interface crate

This crate provides a client for OpenComputers II's HLAPI, used to interface with high-level peripheral devices like redstone interfaces, robot movement APIs, and transposers, among many others.

This crate is specifically for the new 1.21.1 version of base OC2! The original OC2 for 1.18.2 and the community 1.20.1 port, OC2R, use an earlier version of the HLAPI that is incompatible with this crate. Support for their version of the HLAPI will very likely not be added, as crates to interface with both are already available, see miku-rpc and oc2-hlapi.

The implemented HLAPI is that of OC2 v0.10.1, the latest version at the time this version of the crate was made.

This crate is divided into three different layers of abstraction. From lowest to highest level, they are:

  • Transports: Transports provide the raw byte streams over which HLAPI messages are sent, and may even adapt other transports, like the client for the HLAPI bus daemon.
  • Codecs: Codecs handle message framing, (de)serialization and connection management, allowing for the sending and receiving of parsed Rust data structures instead of raw bytes. Only one is provided with this crate for now, as only one main codec exists.
  • The HlapiClient + device wrappers: HlapiClient handles listing devices, obtaining device wrappers, and reading device events. Wrappers are Rust structures that provide a typed API for HLAPI devices, instead of the dynamic typing that the HLAPI natively exposes. Devices without an available wrapper can be used either via the GenericDevice wrapper, which exposes the raw method call API, or a custom wrapper type can be written (this is the recommended approach).

Cross-platform usage

Most of this crate is platform-independent, and should work on any platform with std, though it has only been tested on Linux, with a cross-build for Windows confirmed successful. The LinuxVirtio and BusdUnixSocket transports are Linux-only, a custom transport is required to use the client on other platforms.

Basic usage

The following is a basic usage example of this crate. It opens a connection (via the bus daemon if available, otherwise via the virtio ports), gets a wrapper for a redstone device, and sets the left output on.

use oc2_hlapi_v2::{
    client::HlapiClient,
    devices::{Side, redstone::Redstone},
};
use std::error::Error;

let client = HlapiClient::connect()?;
let redstone = client.find_device_by_type::<Redstone>()?.ok_or_else(||
    Box::<dyn Error>::from("No redstone device found".to_owned())
)?;
redstone.set_redstone_output(Side::Left, 15)?;

Device events

Device events in this crate are handled somewhat differently than the Lua library. The biggest change is that waiting on an event for one device will not automatically discard events for other devices.

To get events from a device, obtain an EventStream for the device by calling Device::event_stream. Events are dropped for a device if no EventStreams exist for it, so the stream will start out empty. Call EventStream::get_queued_event to obtain events waiting to be processed, or EventStream::get_event to wait for an event from the device.

EventStreamGroup allows waiting for events from any one of a group of devices all at once. All devices in a given EventStreamGroup must be on the same client. Add streams to the group with EventStreamGroup::add and wait for an event on any stream with EventStreamGroup::wait_for_event.

EventStream::get_event and EventStreamGroup::wait_for_event both block the client from being accessed from other threads while they are running, so if multiple threads are being used, threads that wait for device events should have their own connection to the HLAPI bus.

Writing a custom device wrapper

Writing a custom device wrapper requires implementing the DeviceBase trait. You should have at least two fields in your struct, one for a client and one for a device UUID. new() should fill them in with the passed parameters and give default values for the other fields, uuid_base() should get the UUID, and client() should get the client. A simple example is below.

use oc2_hlapi_v2::{
    client::{DeviceBase, HlapiClient, HlapiClientError},
    uuid::Uuid,
};

use std::convert::Infallible;

#[derive(Debug, Clone)]
pub struct RedstoneDevice {
    client: HlapiClient,
    uuid: Uuid,
}

impl RedstoneDevice {
    // Note: for simplicity side is a string here, in the real wrapper it is an enum for type
    // safety. The value of strength is also not validated, unlike in the real wrapper.
    pub fn set_redstone_output(&self, side: &str, strength: u8) -> Result<(), HlapiClientError> {
        self.call::<()>("setRedstoneOutput", &[&side as _, &strength as _], &[])?;
        Ok(())
    }
}

impl DeviceBase for RedstoneDevice {
    const TYPE: &'static str = "redstone";

    type Event = Infallible;

    fn new(client: HlapiClient, uuid: Uuid) -> Result<Self, HlapiClientError> {
        Ok(Self { client, uuid })
    }

    fn uuid_base(&self) -> Uuid {
        self.uuid
    }

    fn client(&self) -> &HlapiClient {
        &self.client
    }
}

AI usage disclaimer

TL;DR: AI-assisted, not vibecoded. AI was primarily a helper checking that the code matched the specs and providing bugfixes, and I wrote the vast majority of the code. The only pure AI bit is the test suite. The docs are 100% human-written.

Library code (src/) is 95% human-written, with some chunks of code from Claude that have been reviewed and understood by me before using. Every feature started from a base of fully human-written code. The test suite (tests/ + ingame/) is entirely AI-generated, as I am incredibly bad at writing tests, and something is better than nothing. It appears to cover everything quite well, and Claude was not allowed to make changes that broke the library to make tests pass.

The documentation is 100% human written. If any bit of the docs feel AI-generated, well, I happen to be autistic and the writing of us autistic folks very often gets falsely flagged as AI by detectors...

Claude was also used to do a, quite considerable, code review of the crate before it was published, which is where the AI-generated chunks of code and test suite came from. Any and all design changes (not many) as a result I agreed with, Claude did not make any changes without my OK.

License

Licensed under either of

at your option.

Contribution

Unless you explicitly state otherwise, any contribution intentionally submitted for inclusion in the work by you, as defined in the Apache-2.0 license, shall be dual licensed as above, without any additional terms or conditions.