meteor-pubsub

Meteor publications and subscriptions

Publications stream a set of documents to subscribed clients and keep them live. The publication is the only place server-side authorization can filter rows before they reach the client.

Decision flow

1. Does the client need this data reactively? If no, prefer a method (one-shot

read). If yes, use a publication.

1. Is the data user-specific? Filter by this.userId inside the publish

function. Without that filter, documents leak across users.

1. Can the publication be expressed as a single cursor? Return it directly.
2. Does it need one async lookup before choosing a cursor? Use an async

publish handler, await the lookup, then return the cursor.

1. Does it need per-document async joins, custom aggregation output, or an

external reactive source? Drop to the low-level this.added / this.changed / this.removed API.

Scaffold

import { Meteor } from "meteor/meteor";
import { Items } from "/imports/api/items";

Meteor.publish("items.mine", function () {
  if (!this.userId) {
    return this.ready();
  }
  return Items.find(
    { ownerId: this.userId },
    { fields: { title: 1, qty: 1, updatedAt: 1 }, sort: { updatedAt: -1 } },
  );
});

Project fields whenever the collection contains columns the subscriber must not receive.

Async publish handlers may also return a cursor:

Meteor.publish("items.byTeam", async function (teamId) {
  const member = await Memberships.findOneAsync({
    teamId,
    userId: this.userId,
  });
  if (!member) return this.ready();
  return Items.find({ teamId }, { fields: { title: 1, qty: 1 } });
});

Meteor awaits the handler before processing the returned cursor.

Subscribing

import { Meteor } from "meteor/meteor";

const handle = Meteor.subscribe("items.mine");
// React / Blaze / Svelte hooks observe handle.ready() and the local cursor.
handle.stop();

Publication strategies

Set per-collection with Meteor.server.setPublicationStrategy. Three options:

import { DDPServer } from "meteor/ddp-server";

Meteor.server.setPublicationStrategy(
  "items",
  DDPServer.publicationStrategies.NO_MERGE,
);

See references/publication-strategies.md.

Low-level publish API

For joins or async work:

Meteor.publish("feed", async function () {
  const cursor = Posts.find({}, { fields: { title: 1, authorId: 1 } });
  const observer = await cursor.observeChangesAsync({
    added: async (id, doc) => {
      const author = await Users.findOneAsync(doc.authorId, {
        fields: { username: 1 },
      });
      this.added("feed", id, { ...doc, authorName: author?.username });
    },
    changed: (id, changes) => this.changed("feed", id, changes),
    removed: (id) => this.removed("feed", id),
  });

  this.ready();
  this.onStop(() => observer.stop());
});

Note: cursor transform functions remain synchronous. Use a separate publication or the low-level API for per-document async joins.

Live observer delivery does not wait for one async callback before later changes arrive. If ordering or backpressure matters, serialize the work with a per-subscription Promise queue. Attach an error policy and stop the observer from this.onStop; see references/low-level-publish-api.md.

Anti-patterns

• Publish without a this.userId filter when data is user-specific.
• Publish sensitive or unnecessary columns. Add a fields projection when

the full document is not part of the publication contract.

• Confuse an async publish handler with an async cursor transform. Meteor

awaits the handler Promise, but each transform callback must return a document synchronously.

• Use cursor transforms for async joins. Transforms are synchronous; the

low-level API is the right tool.

• Subscribe to large unbounded collections. Page the data with limit/skip.

See also

• references/publication-strategies.md
• references/low-level-publish-api.md
• references/eval-cases.md