Home

Awesome

maestroqadev:pub-sub-lite

Lighter (Method-based) pub/sub for Meteor

Intro

Meteor publications are very resource-intensive by nature. For every subscribed client, Meteor create observers for tracking and sending real-time updates. Moreover, Meteor also maintains a copy of each client's data on the server-side (using SessionCollectionView), which is used to decide which portion of data should be reactively sent to clients. These features may put a huge load on server, especially at scale.

For this reason, publications should only be used when reactivity is essential for a given use case. When reactivity is not essential, it is better to retrieve data via a non-reactive fetch, such as a Meteor Method call.

However, using Method calls for fetching data has certain disadvantages:

The pub-sub-lite package aims to solve these problems by providing:

Use cases

pub-sub-lite will be useful when:

How to use the package

Installation

meteor add maestroqadev:pub-sub-lite

Converting existing pub/sub to Method

Because Meteor.publishLite and Meteor.subscribeLite have the same signature and simulate the behaviours of the original Meteor.publish and Meteor.subscribe, you usually do not need to refactor any existing code.

Meteor.publishLite('booksAndAuthors', function (/* arguments */) {
  /* Your original publish handler can be kept intact */

  // `this.userId` is still defined if the caller is logged in
  const userId = this.userId;

  // `this.added` can still be used. The added document will be included in the
  // final result data.
  this.added('books', 'CvwRfPxAoXQFi4txC', attrs);

  // Other low-level publish API, including 'this.changed', 'this.removed', 'this.ready',
  // 'this.onStop', 'this.error', and 'this.stop' will be disregarded, as they no longer
  // fit into the context of a Method-based data fetch.
  this.changed('books', 'CvwRfPxAoXQFi4txC', attrs); // no-op

  // As usual we can return a cursor or array of cursors
  return [Books.find(), Authors.find()];
});

/* Your original subscription and front-end logic can be kept mostly intact */

const subHandle = Meteor.subscribeLite(
  'booksAndAuthors',
  argument1,
  argument2,
  {
    onStop(error) {
      // If provided, run when subHandle.stop() is called, or when an error occurs
    },
    onReady() {
      // If provided, run when the Method's result data has been received.
      // Note: As with the original Meteor.subscribe signature, if the last
      // argument is a function, it will be interpreted as onReady.
    },
  }
);

Tracker.autorun(function () {
  // Once data has been received, subHandle.ready() will reactively return `true`,
  // re-triggering all reactive computations
  if (!subHandle.ready()) {
    console.log('Loading data...');
    return;
  }

  // Data can now be accessed in Minimongo just like with a normal pub/sub
  const books = Books.find().fetch();
  const authors = Authors.find().fetch();
});

// subscribeLite is also defined for Template instances
Template.BooksAndAuthors.onCreated(function () {
  this.subscribeLite('booksAndAuthors');
  this.autorun(() => {
    if (this.subscriptionsReady()) {
      const books = Books.find().fetch();
      const authors = Authors.find().fetch();
    }
  });
});

Enhanced Methods

If Meteor.publishLite and Meteor.subscribeLite are meant for quickly converting legacy pub/sub code, enhanced Methods will power up your Meteor Methods workflow.

Minimongo data merging

Meteor.methodsEnhanced produces Methods that support automatic Minimongo merging: If your Method handler function returns a cursor or array of cursors, it will be automatically restructured into a data format that can be merged into Minimongo on the client-side.

Meteor.methodsEnhanced({
  getBooksAndAuthors() {
    return [Books.find(), Authors.find()];
  },
});

// On the client, use Meteor.callEnhanced / Meteor.applyEnhanced. Once arrived
// the data will be merged into Minimongo automatically, and can be accessed
// in the Books and Authors collections from other parts of the front-end.
Meteor.callEnhanced('getBooksAndAuthors');

// The data (restructured by Meteor.methodsEnhanced) is also available as the
// result object
Meteor.callEnhanced('getBooksAndAuthors', function (error, result) {
  if (result) console.log(result);
  /*
    Output:
    {
      cacheMethodResultInMinimongo: true,
      books: [bookDocument1, bookDocument2],
      authors: [authorDocument1],
    }
  */
});

The data shape above is called a dictionary of collection names and their documents, together with a cacheMethodResultInMinimongo boolean key indicating that this result data can be merged into Minimongo. In fact, instead of using Meteor.methodsEnhanced on the server-side, you can use the original Meteor.methods and manually return the same dictionary shape. Meteor.callEnhanced and Meteor.applyEnhanced will still be able to automatically merge the data in that case.

Emitting mutation update messages

Methods defined by Meteor.methodsEnhanced can also emit mutation update messages to the Method caller on the client-side. These DDP messages will update Minimongo automatically.

Meteor.methodsEnhanced({
  setBestSellingAuthor(authorId) {
    if (Meteor.isServer) {
      // Server-only logic to calculate number of books sold
      const numberOfBooksSold = ...;
    }

    // The changes to this author document will be automatically reflected in
    // client-side Minimongo, without any pub/sub!
    Authors.update(authorId, { $set: { isBestSelling: true, numberOfBooksSold } });
  }
});

In the example code above, even though isBestSelling can be simulated on the client-side with optimistic update, there is no way to determine numberOfBooksSold because its calculation logic is server-only. Without pub/sub, we would have to manually return numberOfBooksSold from our Method, and manually update this value in Minimongo. Enhanced Methods automate this whole process.

Besides updated documents, enhanced Methods can also emit messages for inserted and removed documents.

Caching and Minimongo data merging

Caching for Meteor.subscribeLite

By default, all Meteor.subscribeLite calls are cached for 5 minutes. This means duplicated calls having identical name and arguments will not be repeated. If there are subsequent calls having the same name but different arguments, they will still be carried out normally.

The default cache duration can be set with PubSubLite.setDefaultSubsCacheDurationMs:

import { PubSubLite } from 'meteor/maestroqadev:pub-sub-lite';

// Setting default cache duration to one minute
PubSubLite.setDefaultSubsCacheDurationMs(1000 * 60);

Cache duration can also be set individually for each Meteor.subscribeLite and will take priority over the default setting:

Meteor.subscribeLite(name, argument1, argument2, {
  onStop() {...},
  onReady() {...},
  cacheDurationMs: 1000 * 60,
});

If you want to turn off caching, call PubSubLite.disableSubsCache() to disable globally or set cacheDurationMs as 0 to disable for individual Meteor.subscribeLite.

Caching for Meteor.callEnhanced/applyEnhanced

By default, all Meteor.callEnhanced/applyEnhanced calls are cached for 5 minutes. Similar to the caching behaviour of Meteor.subscribeLite, Method calls having identical Method name and arguments are deduplicated.

Methods caching duration can also be customized:

import { PubSubLite } from 'meteor/maestroqadev:pub-sub-lite';

// Set default Methods cache duration
PubSubLite.setDefaultMethodCallCacheDurationMs(1000 * 60);

// Disable Methods caching globally
PubSubLite.disableMethodCallCache();

// Set individual cache duration for Meteor.callEnhanced by adding an object having
// the cacheDurationMs key as the final argument, or the last argument before
// callback. Note that this object will be omitted and will not be counted
// as a Method argument.
Meteor.callEnhanced(
  name,
  argument1,
  argument2,
  { cacheDurationMs: 1000 * 60 },
  callback
);

// Set individual cache duration for Meteor.applyEnhanced using the apply options object
Meteor.applyEnhanced(
  name,
  [argument1, argument2],
  {
    // cacheDurationMs can be defined alongside other native apply options
    wait: true,
    throwStubExceptions: true,
    cacheDurationMs: 0, // turning off caching for this Method call
  },
  callback
);

A powerful feature of enhanced Methods is that result data is also cached. This means when a duplicated Method is called and there is a cache hit, the Method will return immediately with the result data cached earlier. This allows enhanced Methods to be used for loading data without the risk of over-fetching.

// A cache is registered the first time Method is called, including its result data
Meteor.callEnhanced(name, argument1, argument2, function(error, result) {...});

// When called again with identical arguments, the Method returns immediately
// (no request is sent to server)
Meteor.callEnhanced(name, argument1, argument2, function(error, result) {
  // `result` will be the result data retrieved during the first call
});

Furthermore, if the cached result data was merged to Minimongo as illustrated above, it will be synced with Minimongo before being returned:

// Books and authors cached and merged into Minimongo
Meteor.callEnhanced('getBooksAndAuthors', function (error, result) {
  console.log(result);
  /*
    {
      cacheMethodResultInMinimongo: true,
      books: [...],
      authors: [{
        _id: 'NzrGsj9ooJnQwbDfZ',
        name: 'John',
        isBestSelling: false,
      }],
    }
  */
});

// Perform a mutation that set John's isBestSelling to true and automatically
// update that value in Minimongo
Meteor.callEnhanced('setBestSellingAuthor', 'NzrGsj9ooJnQwbDfZ');

// Later when 'getBooksAndAuthors' is called again, the previously saved data is
// synced with Minimongo before being returned
Meteor.callEnhanced('getBooksAndAuthors', function (error, result) {
  console.log(result);
  /*
    {
      cacheMethodResultInMinimongo: true,
      books: [...],
      authors: [{
        _id: 'NzrGsj9ooJnQwbDfZ',
        name: 'John',

        // We have the updated value, even though this second Method call was a
        // cache hit and never reached the server!
        isBestSelling: true,
      }],
    }
  */
});

Additional helpers

If your Method returns cursor(s) or data in the dictionary shape (with the cacheMethodResultInMinimongo key), caching and result data merging with Minimongo will be automatically carried out for you as illustrated above. In other cases, the package provides helpers for handling custom data:

PubSubLite.cacheMethodResult

This helper can cache Method data in any arbitrary values, and will not perform Minimongo merging:

Meteor.callEnhanced(name, argument1, argument2, function (error, result) {
  if (result) {
    PubSubLite.cacheMethodResult({
      name,
      args: [argument1, argument2],
      data: result,
      durationMs: 1000 * 60,
    });
  }
});

PubSubLite.cacheMethodResultInMinimongo

This helper can cache and perform Minimongo merging for result data in the following format:

Meteor.callEnhanced(name, argument1, argument2, function (error, result) {
  if (result) {
    PubSubLite.cacheMethodResultInMinimongo({
      name,
      args: [argument1, argument2],
      data: result,
      // When data is a document or array of documents, a collectionName must be provided
      collectionName: 'books',
      durationMs: 1000 * 60,
    });
  }
});

PubSubLite.mergeDataIntoMinimongo

This helper makes it convenient to merge data from arbitrary sources (e.g. data fetched via Apollo) into Minimongo. The data need to be structured as a dictionary of collection names and their documents:

const data = {
  books: [bookDocument1, bookDocument2],
  authors: [],
};
PubSubLite.mergeDataIntoMinimongo(data);

Integration with other packages

publishComposite

If you have existing composite publications, simply replace publishComposite with Meteor.publishCompositeLite to have the publication (and any child publications) converted into a Method.

// Similar to the original publishComposite, the second argument can be either
// an object literal or callback function
Meteor.publishCompositeLite('authorsAndTheirBooks', {
  find() {
    return Authors.find();
  },
  children: [
    {
      find(author) {
        return Books.find({ authorId: author._id });
      },
    },
  ],
});

Meteor.subscribeLite('authorsAndTheirBooks', function () {
  const authors = Authors.find().fetch();
  const books = Books.find().fetch();
});

ValidatedMethod

To use pub-sub-lite with ValidatedMethod, replace mdg:validated-method with maestroqadev:validated-method. More information can be found here.

Miscellaneous

Requirements for MongoDB Change Streams

The package uses MongoDB Change Streams to detect changes in update operations. Because each stream may open a separate MongoDB connection, the package tries to minimize the number of streams to at most 1 per collection. This means the maximum number of streams opened at once can potentially be equal to the number of collections (although in practice this number will likely be smaller, as streams are closed as soon as we capture the necessary changes).

The Node.js MongoDB driver sets the value to 5. This value was set for legacy reasons only, and is too small for the connections potentially opened by pub-sub-lite. The package sets this value to 100 by default. This number was inspired by the default value in the Python MongoDB driver, and should be enough for the package to function.

To customize this value, edit mongoConnectionPoolSize in the CUSTOM_HARD_CODED_CONFIG object in config.js.

Notes:

The __PubSubLite__updateOperationId field

During update operations, enhanced Methods set an additional field on the affected document(s) to keep track of what changes were triggered from what operations. This field is called __PubSubLite__updateOperationId and is not automatically removed from the document(s) after each operation (because we do not want to have another manual update operation just to unset it).

The field will be automatically omitted from all document(s) queried from normal collections created with Mongo.Collection, and will not affect your existing application in most cases. However you should still be aware of its existance, as the field might be present in certain cases, such as when documents are queried with native MongoDB APIs.

Todos

Contributors