Node.js Evaluation Loop (NEL)

NEL is an npm module for running Node.js REPL sessions.

NEL is a spin-off library from IJavascript. This fact explains some of the design decisions in NEL, such as returning results in MIME format, and the functionality provided for completion and inspection of Javascript expressions. See the section on usage for more details.

NEL is used by the following Jupyter kernels:

• IJavascript
• ITypescript
• jp-babel
• jp-coffeescript

Main Features

• Run Javascript code within a Node.js session. The result can be formatted as:
  • plain text
  • other MIME formats (HTML, SVG, PNG...)
• Run Javascript code asynchronously.
• Generate a list of completion options for an incomplete piece of Javascript code.
• Inspect a Javascript expression and return information such as type or even documentation (currently, only for Javascript builtins).

Announcements

• NEL v1.3.0: API enhancement (added option awaitExecution)
• NEL v1.2.0: API enhancement (Session#transpile may return a Promise)
• NEL v1.1.0: API enhancement (added $$.clear({wait}))
• NEL v1.0.0: Stable API
• NEL v0.5.6: API enhancement (added $$.input() and onRequest callback)
• NEL v0.5.5: Accept Promises as output
• NEL v0.5.4: API enhancement (added $$.display() and onDisplay callback)
• NEL v0.5: API enhancement (added transpile option)
• NEL v0.4: API enhancement (added onStdout and onStderr callbacks)
• NEL v0.3: New API (simplify API by hiding type module:nel~Task)
• NEL v0.2: API change (removed Session#executionCount)
• NEL v0.1.1: API enhancement (experimental $$mimer$$ and $$defaultMimer$$)
• NEL v0.1: Output change (changed function output)
• NEL v0.0: Initial release

Install

npm install nel

Usage

The documentation generated by JSDoc can be found here.

Hello, World!

// Load `nel` module
var nel = require("nel");

// Setup a new Javascript session
var session = new nel.Session();

// Example of an execution request
// Output:
// { mime: { 'text/plain': '\'Hello, World!\'' } }
var code = "['Hello', 'World!'].join(', ');";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

Exceptions

// Example of throwing an exception
// Output:
// { error:
//  { ename: 'Error',
//    evalue: 'Hello, World!',
//    traceback:
//     [ 'Error: Hello, World!',
//       '    at evalmachine.<anonymous>:1:7',
//       '    at run ([eval]:182:19)',
//       '    at onMessage ([eval]:63:41)',
//       '    at process.EventEmitter.emit (events.js:98:17)',
//       '    at handleMessage (child_process.js:318:10)',
//       '    at Pipe.channel.onread (child_process.js:345:11)' ] } }
code = "throw new Error('Hello, World!');";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

stdout and stderr

// Example of use of console.log()
// Output:
// Hello, World!
//
// { mime: { 'text/plain': 'undefined' } }
code = "console.log('Hello, World!');";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
    onStdout: console.log,
    onStderr: console.error,
});

onDisplay callback

The Jupyter messaging protocol introduces the concept of display. A display is very much like an execution result. It is associated with an execution request and in protocol version 5.1 and above it can be assigned an ID for subsequent updates. Here's an example of the support provided by NEL:

// Example using onDisplay callback
// Output:
// { mime: { 'text/plain': 'Hello, World!' } }
code = "$$.display().text('Hello, World!');";
session.execute(code, {
    onDisplay: console.log,
});

// Example using a display ID
// Output:
// { display_id: 'test', mime: { 'text/plain': 'Hello, World!' } }
code = "$$.display('test').text('Hello, World!');";
session.execute(code, {
    onDisplay: console.log,
});

onRequest callback and $$.input(options, callback)

The Jupyter messaging protocol defines an stdin socket, so that a kernel can request an input from the user. NEL defines $$.input(options, callback) to create such a request.

Here are two examples (first one passing a callback to $$.input; second one using a Promise returned by $$.input()):

// Example passing a callback to $$.input()
// Output:
// { mime: { 'text/plain': '\'opensesame\'' } }
code = "$$.input({prompt:'?', password: true}, function(error, reply) {$$.done(reply)});";

session.execute(code, {
    onRequest: function(request, onReply) {
        assert(request.input.prompt === "?");
        assert(request.input.password === true);

        onReply({input: "opensesame"});
    },
    onSuccess: console.log,
});

// Example using the Promise returned by $$.input()
// Output:
// { mime: { 'text/plain': '\'opensesame\'' } }
code = "(function($$) {$$.input({prompt:'?', password: true}).then($$.done);})($$);";

session.execute(code, {
    onRequest: function(request, onReply) {
        assert(request.input.prompt === "?");
        assert(request.input.password === true);

        onReply({input: "opensesame"});
    },
    onSuccess: console.log,
});

onRequest callback and $$.clear(options)

The Jupyter messaging protocol defines the message clear_output for kernels to request the output of a cell to be cleared. NEL provides $$.clear(options) to implement such a request.

Here's an example showing the use:

// Example using $$.clear(options)
// Output:
// { clear: { wait: true } }
code = "$$.clear({wait: true});";

session.execute(code, {
    onRequest: console.log
});

MIME output

A session may return results in MIME formats other than 'text/plain'.

// HTML example
// Output:
// { mime: { 'text/html': '<div style=\'background-color:olive;width:50px;height:50px\'></div>' } }
code = "$$html$$ = \"<div style='background-color:olive;width:50px;height:50px'></div>\";";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

// SVG example
// Output:
// { mime: { 'image/svg+xml': '<svg><rect width=80 height=80 style=\'fill: orange;\'/></svg>' } }
code = "$$svg$$ = \"<svg><rect width=80 height=80 style='fill: orange;'/></svg>\";";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

// PNG example
code = "$$png$$ = require('fs').readFileSync('image.png').toString('base64');";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

// JPEG example
code = "$$jpeg$$ = require('fs').readFileSync('image.jpg').toString('base64');";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

// MIME example
code = "$$mime$$ = {\"text/html\": \"<div style='background-color:olive;width:50px;height:50px'></div>\"};";
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
});

Promises are accepted as output

When the result of an execution request is a Promise, NEL enables asynchronous execution automatically and waits for the Promise to resolve:

// Example returning a Promise
// Output:
// { mime: { 'text/plain': '\'Hello, World!\'' } }
code = "Promise.resolve('Hello, World!');";

session.execute(code, {
    onSuccess: console.log,
});

Generate a completion list

NEL can parse simple Javascript variable expressions and generate a list of completion options:

session.complete(
    "set",     // code
    3,         // cursorPos
    {
        onSuccess: console.log,
        onError: console.error,
    }
);

// Output:
// { completion:
//    { list: [ 'setImmediate', 'setInterval', 'setTimeout' ],
//      code: 'set',
//      cursorPos: 3,
//      matchedText: 'set',
//      cursorStart: 0,
//      cursorEnd: 3 } }

Note that the cursor position can be located anywhere within the Javascript code:

session.complete(
    "set",     // code
    2,         // cursorPos
    {
        onSuccess: console.log,
        onError: console.error,
    }
);

// Output:
// { completion:
//    { list: [ 'setImmediate', 'setInterval', 'setTimeout' ],
//      code: 'set',
//      cursorPos: 2,
//      matchedText: 'se',
//      cursorStart: 0,
//      cursorEnd: 3 } }

Inspect an expression

NEL can parse simple Javascript variable expressions and inspect their value:

code = "var a = [1, 2, 3];";
session.execute(code, null, onError);
session.inspect(
    code,      // code
    5,         // cursorPos
    {
        onSuccess: console.log,
        onError: console.error,
    }
);

// Output:
// { inspection:
//    { string: '[ 1, 2, 3 ]',
//      type: 'Array',
//      constructorList: [ 'Array', 'Object' ],
//      length: 3,
//      code: 'var a = [1, 2, 3];',
//      cursorPos: 5,
//      matchedText: 'a' } }

NEL can also provide relevant documentation (currently only available for Javascript builtins):

session.inspect(
    "parseInt", // code
    8,          // cursorPos
    {
        onSuccess: console.log,
        onError: console.error,
    }
);

// Output:
// { inspection:
//    { string: '[Function: parseInt]',
//      type: 'Object',
//      constructorList: [ 'Function', 'Object' ],
//      length: 2,
//      code: 'parseInt',
//      cursorPos: 8,
//      matchedText: 'parseInt' },
//   doc:
//    { description: 'The parseInt() function parses a string argument and returns an integer of the specified radix (the base in mathematical numeral systems).',
//      url: 'https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/parseInt',
//      usage: 'parseInt(string, radix);' } }

Callbacks beforeRun and afterRun

var beforeRun = function() { console.log("This callback runs first"); }
code = "'I run next'";
var afterRun = function() { console.log("This callback runs last"); }
session.execute(code, {
    onSuccess: console.log,
    onError: console.error,
    beforeRun: beforeRun,
    afterRun: afterRun,
});

// Output:
// This callback runs first
// { mime: { 'text/plain': '\'I run next\'' } }
// This callback runs last

Contributions

First of all, thank you for taking the time to contribute. Please, read CONTRIBUTING.md and use the issue tracker for any contributions: support requests, bug reports, enhancement requests, pull requests, ...

TODO

• Add tests for customising output
• Add Node.js documentation