Skip to main content

Read-Your-Writes

When building collaborative applications, you often run into a consistency problem: a user performs an action (like sharing a document or adding a team member), and immediately navigates to a screen that checks that new permission.

If your authorization storage uses asynchronous replication or has not yet committed the transaction, the subsequent check() call might fail because the new permission tuple hasn't reached the database.

polizy solves this with Read-Your-Writes validation using contextualTuples.

This guide shows you how to pass temporary, uncommitted facts directly to a permission check.

Performance & Consistency

To learn about how polizy caches reads, wraps operations in scopes, and implements strong consistency levels, read Consistency & Caching.

The Solution: contextualTuples​

Instead of waiting for database synchronization or blocking the thread, you can pass uncommitted relation tuples in the contextualTuples parameter of a check(), checkMany(), or explain() call.

polizy evaluates these contextual tuples in memory as if they were already written to storage. They are never saved to your persistent database.

Example: Checking a Freshly-Written Grant​

Imagine Alice just shared a document with Bob, and your app needs to verify access immediately before the database commit completes.

import { AuthSystem, InMemoryStorageAdapter } from "polizy";

// ... initialize authz ...

const bob = { type: "user", id: "bob" };
const doc = { type: "document", id: "confidential-plan" };

// Check if Bob can view the document, assuming the write completes successfully
const canBobView = await authz.check({
  who: bob,
  canThey: "view",
  onWhat: doc,
  
  // Pass uncommitted facts here
  contextualTuples: [
    {
      subject: bob,
      relation: "viewer",
      object: doc,
    },
  ],
});

console.log(canBobView); 
// => true (evaluated immediately in memory, even if the database is empty)

When to Use Contextual Tuples​

  1. Optimistic UI updates: Checking permissions for a page transition immediately after a user requests access.
  2. Multi-step setup flows: Verifying a complex hierarchy or setup before finalizing and committing the configuration to the database.
  3. Draft environments: Simulating temporary access policies during previews without polluting production storage.