enchannel-zmq-backend
Installation | Usage |
Contributors and developers |
Learn more about nteract
enchannel-zmq-backend offers the ZeroMQ backend implementation for
enchannel
.
Technical overview
As a refresher for the reader, enchannel details nteract's
lightweight, implementation-flexible specification for communication
between a user frontend and a backend, such as a language kernel. The
enchannel specification offers a simple description of "what"
messages may be passed between frontends and backends, while leaving
a developer freedom in "how" to achieve message communication.
enchannel-zmq-backend takes a classic design approach using ZeroMQ,
the foundation messaging protocol for the Jupyter project.
enchannel-zmq-backend implements backend support for the messaging
channels described in the Jupyter messaging specification. This
spec explains how front end clients should communicate with
backend language kernels which implement the Jupyter messaging
specification.
Our backend
enchannel-zmq-backend implements the "how" to communicate messages
to and from a backend.
We provide functions to create RxJS
Subjects (two way
Observables for
four of the channels described in the
Jupyter messaging specification):
That's it. Functions for four channels; simplicity in action.
Installation
Prerequisite: Node.js and npm
You may use whichever package manager (npm
or yarn
) best suits your workflow. The nteract
team internally uses yarn
.
npm install enchannel-zmq-backend
yarn add enchannel-zmq-backend
Usage
Creating messaging channels
To get access to all of the channels
for messaging (shell
, control
,
iopub
, and stdin
), import and use the createChannels
function:
import { createChannels } from 'enchannel-zmq-backend'
The createChannels
function accepts two things:
-
an identity
You'll want to set up your identity, relying on the node uuid
package:
const uuidv4 = require('uuid/v4');
const identity = uuidv4();
-
a runtime object, such as a kernel (which matches the on-disk JSON).
Using spawnteract
with
this project helps streamline spawning a kernel.
const runtimeConfig = {
stdin_port: 58786,
ip: '127.0.0.1',
control_port: 58787,
hb_port: 58788,
signature_scheme: 'hmac-sha256',
key: 'dddddddd-eeee-aaaa-dddd-dddddddddddd',
shell_port: 58784,
transport: 'tcp',
iopub_port: 58785
}
To create the channels object:
const channels = createChannels(identity, runtimeConfig)
const { shell, iopub, stdin, control } = channels;
enchannel-zmq-backend
also gives access to all of the channels
via a
single multipled channel exposed via createMainChannel
.
import { createMainChannel } from 'enchannel-zmq-backend';
Similar to the createChannels
function, the createMainChannel
function
accepts both an identity and a runtime object.
const channel = createMainChannel(identity, runtimeConfig);
Messages that are sent via the mutliplexed channel need to define a type
property that outlines which channel they should be sent under.
const body = {
header: {
msg_id: `execute_9ed11a0f-707e-4f71-829c-a19b8ff8eed8`,
username: "rgbkrk",
session: "00000000-0000-0000-0000-000000000000",
msg_type: "execute_request",
version: "5.0"
},
content: {
code: 'print("woo")',
silent: false,
store_history: true,
user_expressions: {},
allow_stdin: false
}
};
const message = { type: "shell", body };
enchannel-zmq-backend
also offers four convenience functions to
easily create the messaging channels for control
, stdin
, iopub
,
and shell
:
import {
createControlSubject,
createStdinSubject,
createIOPubSubject,
createShellSubject,
} from 'enchannel-zmq-backend';
Creating a subject for the shell
channel:
const shell = createShellSubject(identity, runtimeConfig)
Subscribing to messages
Here's an example about how to subscribe to iopub
messages:
const iopub = createIOPubSubject(identity, runtimeConfig);
var subscription = iopub.subscribe(msg => {
console.log(msg);
}
Since these channels are RxJS Observables, you can use filter
, map
,
scan
and many other RxJS operators:
iopub.filter(msg => msg.header.msg_type === 'execute_result')
.map(msg => msg.content.data)
.subscribe(x => { console.log(`DATA! ${util.inspect(x)}`)})
Sending messages to the kernel
Executing code will rely on sending an execute_request
to the shell
channel.
var message = {
header: {
msg_id: `execute_9ed11a0f-707e-4f71-829c-a19b8ff8eed8`,
username: 'rgbkrk',
session: '00000000-0000-0000-0000-000000000000',
msg_type: 'execute_request',
version: '5.0',
},
content: {
code: 'print("woo")',
silent: false,
store_history: true,
user_expressions: {},
allow_stdin: false,
},
};
Currently, you'll need to have at least one subscription activated
before you can send on a channel.
> shell.subscribe(console.log)
> shell.next(message)
> Message {
header:
{ username: 'rgbkrk',
msg_type: 'execute_reply',
msg_id: '0f6d37f3-56a2-41fd-b3ed-90cc189ac423',
version: '5.1',
session: '40472e70-e008-48d1-9537-55837a905c05',
date: '2016-01-12T00:39:44.686986' },
parent_header:
{ username: 'rgbkrk',
session: '00000000-0000-0000-0000-000000000000',
version: '5.0',
msg_id: 'execute_9ed11a0f-707e-4f71-829c-a19b8ff8eed8',
msg_type: 'execute_request' },
metadata:
{ dependencies_met: true,
engine: '34d73425-4f04-4b57-9bc7-b46e3100e1fd',
status: 'ok',
started: '2016-01-12T00:39:44.684534' },
content:
{ status: 'ok',
execution_count: 60,
user_expressions: {},
payload: [] } }
Contributors and developers
ZeroMQ Dependency
If you plan to contribute to this project or extend it, you will need
to have ZeroMQ installed on
your system. The easiest way to do this is to install nteract's
zmq-prebuilt
binary for your operating system.
Install a local development environment
To set up a development environment, you'll need to install:
Then, fork and clone this repo:
git clone https://github.com/nteract/enchannel-zmq-backend.git
cd enchannel-zmq-backend
yarn
Develop! We welcome new and first time contributors.
Learn more about nteract