2022-09-26 11:41:44 +05:30

9.5 KiB

Raw Permalink Blame History

Continuation-Local Storage

Continuation-local storage works like thread-local storage in threaded programming, but is based on chains of Node-style callbacks instead of threads. The standard Node convention of functions calling functions is very similar to something called "continuation-passing style" in functional programming, and the name comes from the way this module allows you to set and get values that are scoped to the lifetime of these chains of function calls.

Suppose you're writing a module that fetches a user and adds it to a session before calling a function passed in by a user to continue execution:

// setup.js

var createNamespace = require('continuation-local-storage').createNamespace;
var session = createNamespace('my session');

var db = require('./lib/db.js');

function start(options, next) {
  db.fetchUserById(options.id, function (error, user) {
    if (error) return next(error);

    session.set('user', user);

    next();
  });
}

Later on in the process of turning that user's data into an HTML page, you call another function (maybe defined in another module entirely) that wants to fetch the value you set earlier:

// send_response.js

var getNamespace = require('continuation-local-storage').getNamespace;
var session = getNamespace('my session');

var render = require('./lib/render.js')

function finish(response) {
  var user = session.get('user');
  render({user: user}).pipe(response);
}

When you set values in continuation-local storage, those values are accessible until all functions called from the original function – synchronously or asynchronously – have finished executing. This includes callbacks passed to process.nextTick and the timer functions (setImmediate, setTimeout, and setInterval), as well as callbacks passed to asynchronous functions that call native functions (such as those exported from the fs, dns, zlib and crypto modules).

A simple rule of thumb is anywhere where you might have set a property on the request or response objects in an HTTP handler, you can (and should) now use continuation-local storage. This API is designed to allow you extend the scope of a variable across a sequence of function calls, but with values specific to each sequence of calls.

Values are grouped into namespaces, created with createNamespace(). Sets of function calls are grouped together by calling them within the function passed to .run() on the namespace object. Calls to .run() can be nested, and each nested context this creates has its own copy of the set of values from the parent context. When a function is making multiple asynchronous calls, this allows each child call to get, set, and pass along its own context without overwriting the parent's.

A simple, annotated example of how this nesting behaves:

var createNamespace = require('continuation-local-storage').createNamespace;

var writer = createNamespace('writer');
writer.run(function () {
  writer.set('value', 0);

  requestHandler();
});

function requestHandler() {
  writer.run(function(outer) {
    // writer.get('value') returns 0
    // outer.value is 0
    writer.set('value', 1);
    // writer.get('value') returns 1
    // outer.value is 1
    process.nextTick(function() {
      // writer.get('value') returns 1
      // outer.value is 1
      writer.run(function(inner) {
        // writer.get('value') returns 1
        // outer.value is 1
        // inner.value is 1
        writer.set('value', 2);
        // writer.get('value') returns 2
        // outer.value is 1
        // inner.value is 2
      });
    });
  });

  setTimeout(function() {
    // runs with the default context, because nested contexts have ended
    console.log(writer.get('value')); // prints 0
  }, 1000);
}

cls.createNamespace(name)

return: {Namespace}

Each application wanting to use continuation-local values should create its own namespace. Reading from (or, more significantly, writing to) namespaces that don't belong to you is a faux pas.

cls.getNamespace(name)

return: {Namespace}

Look up an existing namespace.

cls.destroyNamespace(name)

Dispose of an existing namespace. WARNING: be sure to dispose of any references to destroyed namespaces in your old code, as contexts associated with them will no longer be propagated.

cls.reset()

Completely reset all continuation-local storage namespaces. WARNING: while this will stop the propagation of values in any existing namespaces, if there are remaining references to those namespaces in code, the associated storage will still be reachable, even though the associated state is no longer being updated. Make sure you clean up any references to destroyed namespaces yourself.

process.namespaces

return: dictionary of {Namespace} objects

Continuation-local storage has a performance cost, and so it isn't enabled until the module is loaded for the first time. Once the module is loaded, the current set of namespaces is available in process.namespaces, so library code that wants to use continuation-local storage only when it's active should test for the existence of process.namespaces.

Class: Namespace

Application-specific namespaces group values local to the set of functions whose calls originate from a callback passed to namespace.run() or namespace.bind().

namespace.active

return: the currently active context on a namespace

namespace.set(key, value)

return: value

Set a value on the current continuation context. Must be set within an active continuation chain started with namespace.run() or namespace.bind().

namespace.get(key)

return: the requested value, or undefined

Look up a value on the current continuation context. Recursively searches from the innermost to outermost nested continuation context for a value associated with a given key. Must be set within an active continuation chain started with namespace.run() or namespace.bind().

namespace.run(callback)

return: the context associated with that callback

Create a new context on which values can be set or read. Run all the functions that are called (either directly, or indirectly through asynchronous functions that take callbacks themselves) from the provided callback within the scope of that namespace. The new context is passed as an argument to the callback when it's called.

namespace.runAndReturn(callback)

return: the return value of the callback

Same as namespace.run() but returns the return value of the callback rather than the context.

namespace.bind(callback, [context])

return: a callback wrapped up in a context closure

Bind a function to the specified namespace. Works analogously to Function.bind() or domain.bind(). If context is omitted, it will default to the currently active context in the namespace, or create a new context if none is currently defined.

namespace.bindEmitter(emitter)

Bind an EventEmitter to a namespace. Operates similarly to domain.add, with a less generic name and the additional caveat that unlike domains, namespaces never implicitly bind EventEmitters to themselves when they're created within the context of an active namespace.

The most likely time you'd want to use this is when you're using Express or Connect and want to make sure your middleware execution plays nice with CLS, or are doing other things with HTTP listeners:

http.createServer(function (req, res) {
  writer.bindEmitter(req);
  writer.bindEmitter(res);

  // do other stuff, some of which is asynchronous
});

namespace.createContext()

return: a context cloned from the currently active context

Use this with namespace.bind(), if you want to have a fresh context at invocation time, as opposed to binding time:

function doSomething(p) {
  console.log("%s = %s", p, ns.get(p));
}

function bindLater(callback) {
  return writer.bind(callback, writer.createContext());
}

setInterval(function () {
  var bound = bindLater(doSomething);
  bound('test');
}, 100);

context

A context is a plain object created using the enclosing context as its prototype.

copyright & license

See LICENSE for the details of the BSD 2-clause "simplified" license used by continuation-local-storage. This package was developed in 2012-2013 (and is maintained now) by Forrest L Norvell, @othiym23, with considerable help from Timothy Caswell, @creationix, working for The Node Firm. This work was underwritten by New Relic for use in their Node.js instrumentation agent, so maybe give that a look if you have some Node.js performance-monitoring needs.

9.5 KiB Raw Permalink Blame History Unescape Escape