Event loop mechanics

Overview

The event loop is the heart of JavaScript’s concurrency model. It enables non-blocking I/O and asynchronous programming despite JavaScript being single-threaded. Understanding event loop mechanics is essential for writing performant, responsive applications.

The event loop continuously checks for work to do: executing scripts, processing events, and running queued callbacks. All JavaScript execution happens on a single thread, but the event loop orchestrates when different types of work get processed.

Event loop architecture

The event loop manages multiple task queues with different priorities and processing rules.

┌───────────────────────────┐
│   Call Stack (Running)    │
│   JavaScript Execution    │
└─────────────┬─────────────┘
              │
              v
┌─────────────────────────────────────────┐
│           Event Loop                    │
│                                         │
│  1. Execute script                      │
│  2. Process microtasks (drain queue)    │
│  3. Render (if needed)                  │
│  4. Process one macrotask               │
│  5. Process microtasks (drain queue)    │
│  6. Repeat from step 3                  │
└─────────────────────────────────────────┘
              |
    ┌─────────┴─────────┐
    v                   v
┌─────────┐       ┌──────────┐
│Microtask│       │Macrotask │
│ Queue   │       │  Queue   │
└─────────┘       └──────────┘

Processing phases

Execute current task

Run JavaScript code until the call stack is empty (script execution, event handler, or callback)

Process all microtasks

Drain the microtask queue completely, including any new microtasks added during processing

Render update (browser only)

If rendering is needed, run animation frame callbacks and update the display

Process next macrotask

Take one macrotask from the queue and execute it

Repeat

Go back to step 2 and continue the loop

Microtasks always run before the next macrotask or render. A microtask that continuously queues new microtasks will starve macrotasks and freeze the UI.

Microtask queue

Microtasks are high-priority tasks that execute immediately after the current task completes, before yielding to the browser.

What creates microtasks

Promise callbacks
queueMicrotask
MutationObserver

// Promise.then creates microtasks
console.log('1: Script start');

Promise.resolve().then(() => {
  console.log('3: Promise 1 microtask');
}).then(() => {
  console.log('4: Promise 2 microtask');
});

console.log('2: Script end');

// Output:
// 1: Script start
// 2: Script end
// 3: Promise 1 microtask
// 4: Promise 2 microtask

Key behaviors:

.then(), .catch(), .finally() schedule microtasks
Microtasks run in order they were queued
New microtasks execute before exiting microtask phase

// Explicit microtask scheduling
console.log('1: Start');

queueMicrotask(() => {
  console.log('3: Microtask 1');
  
  queueMicrotask(() => {
    console.log('4: Nested microtask');
  });
});

console.log('2: End');

// Output:
// 1: Start
// 2: End
// 3: Microtask 1
// 4: Nested microtask

Use queueMicrotask() for low-level scheduling when you need microtask timing without promises.

// MutationObserver callbacks are microtasks
const observer = new MutationObserver((mutations) => {
  console.log('3: DOM mutation observed');
  // Runs as microtask after DOM changes
});

const element = document.getElementById('target');
observer.observe(element, { childList: true });

console.log('1: Before mutation');
element.appendChild(document.createElement('div'));
console.log('2: After mutation');

// Output:
// 1: Before mutation
// 2: After mutation
// 3: DOM mutation observed

Microtask queue draining

The microtask queue is fully drained before moving to the next phase:

function demonstrateDraining() {
  console.log('1: Start');
  
  // Queue 3 microtasks
  Promise.resolve().then(() => console.log('3: Microtask 1'));
  Promise.resolve().then(() => console.log('4: Microtask 2'));
  Promise.resolve().then(() => {
    console.log('5: Microtask 3');
    // This creates a NEW microtask during draining
    Promise.resolve().then(() => console.log('6: Nested microtask'));
  });
  
  // Queue a macrotask
  setTimeout(() => console.log('7: Macrotask'), 0);
  
  console.log('2: End');
}

// Output:
// 1: Start
// 2: End
// 3: Microtask 1
// 4: Microtask 2
// 5: Microtask 3
// 6: Nested microtask  <- Runs before macrotask!
// 7: Macrotask

The nested microtask at step 6 runs before the macrotask because the entire microtask queue must drain before processing macrotasks.

Macrotask queue

Macrotasks (also called tasks) are lower-priority work items that execute one at a time, with microtask processing and rendering between each.

What creates macrotasks

setTimeout/setInterval
I/O operations
setImmediate (Node.js)
Message channel

console.log('1: Start');

setTimeout(() => {
  console.log('4: Timeout 1');
}, 0);

setTimeout(() => {
  console.log('5: Timeout 2');
}, 0);

Promise.resolve().then(() => {
  console.log('3: Microtask');
});

console.log('2: End');

// Output:
// 1: Start
// 2: End
// 3: Microtask (runs first!)
// 4: Timeout 1 (macrotask)
// 5: Timeout 2 (macrotask)

Timing guarantees:

Minimum delay, not exact timing
Browser may clamp to 4ms for nested calls
Background tabs may be throttled to 1000ms

// Node.js file I/O
const fs = require('fs');

console.log('1: Start reading');

fs.readFile('data.txt', (err, data) => {
  console.log('3: File read complete');
  // This callback runs as a macrotask
});

Promise.resolve().then(() => {
  console.log('2: Microtask');
});

// Output:
// 1: Start reading
// 2: Microtask
// 3: File read complete

// Node.js only - schedules macrotask
console.log('1: Start');

setImmediate(() => {
  console.log('4: setImmediate');
});

process.nextTick(() => {
  console.log('3: nextTick (microtask)');
});

console.log('2: End');

// Output:
// 1: Start
// 2: End
// 3: nextTick (microtask)
// 4: setImmediate (macrotask)

setImmediate executes after I/O events but before timers in Node.js’s event loop phases.

// MessageChannel for fast macrotask scheduling
const channel = new MessageChannel();

channel.port1.onmessage = () => {
  console.log('3: Message macrotask');
};

console.log('1: Start');
channel.port2.postMessage(null);

Promise.resolve().then(() => {
  console.log('2: Microtask');
});

// Output:
// 1: Start
// 2: Microtask
// 3: Message macrotask

MessageChannel provides faster macrotask scheduling than setTimeout(fn, 0).

Macrotask processing

Unlike microtasks, only one macrotask runs per event loop iteration:

function demonstrateMacrotaskProcessing() {
  console.log('1: Start');
  
  // Queue 3 macrotasks
  setTimeout(() => {
    console.log('3: Macrotask 1');
    Promise.resolve().then(() => console.log('4: Microtask after macro 1'));
  }, 0);
  
  setTimeout(() => {
    console.log('6: Macrotask 2');
    Promise.resolve().then(() => console.log('7: Microtask after macro 2'));
  }, 0);
  
  setTimeout(() => {
    console.log('9: Macrotask 3');
  }, 0);
  
  console.log('2: End');
}

// Output:
// 1: Start
// 2: End
// 3: Macrotask 1       <- First macrotask
// 4: Microtask after macro 1  <- Drain microtasks
// (potential render)
// 6: Macrotask 2       <- Second macrotask
// 7: Microtask after macro 2  <- Drain microtasks
// (potential render)
// 9: Macrotask 3       <- Third macrotask

Long-running macrotasks block rendering and user input. Break work into smaller chunks using setTimeout or requestIdleCallback.

Animation frame callbacks

Animation frame callbacks run before rendering, synchronized with the display refresh rate (typically 60Hz).

requestAnimationFrame timing

function animateWithRAF() {
  console.log('1: Script start');
  
  requestAnimationFrame(() => {
    console.log('4: Animation frame 1');
  });
  
  requestAnimationFrame(() => {
    console.log('5: Animation frame 2');
  });
  
  Promise.resolve().then(() => {
    console.log('3: Microtask');
  });
  
  setTimeout(() => {
    console.log('6: Macrotask');
  }, 0);
  
  console.log('2: Script end');
}

// Output:
// 1: Script start
// 2: Script end
// 3: Microtask
// 4: Animation frame 1  <- Before render
// 5: Animation frame 2  <- Before render
// (render happens here)
// 6: Macrotask          <- After render

Render timing and optimization

60fps animation
Layout thrashing
Compositor animations

// Smooth 60fps animation loop
function animate() {
  const element = document.getElementById('box');
  let position = 0;
  
  function frame(timestamp) {
    // Update state
    position += 2;
    
    // Apply changes (batched by browser)
    element.style.transform = `translateX(${position}px)`;
    
    // Schedule next frame
    if (position < 500) {
      requestAnimationFrame(frame);
    }
  }
  
  requestAnimationFrame(frame);
}

requestAnimationFrame automatically:

Synchronizes with display refresh (60fps or higher)
Pauses in background tabs to save resources
Batches DOM reads/writes to avoid layout thrashing

// BAD: Layout thrashing (forces multiple reflows)
function badAnimation() {
  const boxes = document.querySelectorAll('.box');
  
  requestAnimationFrame(() => {
    boxes.forEach(box => {
      const height = box.clientHeight; // Read (causes layout)
      box.style.height = height + 10 + 'px'; // Write (invalidates layout)
      // Next read will force another layout calculation!
    });
  });
}

// GOOD: Batch reads and writes
function goodAnimation() {
  const boxes = document.querySelectorAll('.box');
  
  requestAnimationFrame(() => {
    // Batch all reads
    const heights = Array.from(boxes).map(box => box.clientHeight);
    
    // Batch all writes
    boxes.forEach((box, i) => {
      box.style.height = heights[i] + 10 + 'px';
    });
  });
}

// Best performance: Compositor-only properties
function compositorAnimation() {
  const element = document.getElementById('box');
  let x = 0;
  
  function frame() {
    // Only animate transform and opacity (compositor properties)
    x += 2;
    element.style.transform = `translateX(${x}px)`;
    
    if (x < 500) {
      requestAnimationFrame(frame);
    }
  }
  
  requestAnimationFrame(frame);
}

Compositor-only properties (no layout/paint):

transform
opacity
filter (some)

Properties that trigger layout/paint:

left, top, width, height
color, background-color
Most other visual properties

Render pipeline

Event Loop Iteration:

1. Execute macrotask
   └─> Call stack empties

2. Process all microtasks
   └─> Drain microtask queue

3. Check if render needed
   └─> If frame budget available:
       ├─> Run animation frame callbacks (rAF)
       ├─> Recalculate styles
       ├─> Layout (reflow)
       ├─> Paint
       ├─> Composite
       └─> Display to screen

4. Process next macrotask

The browser may skip rendering frames if the previous frame took too long, dropping to 30fps or lower to maintain responsiveness.

Idle callbacks and scheduling

Idle callbacks run during idle periods when the browser has completed high-priority work.

requestIdleCallback

// Run low-priority work during idle time
function processLowPriorityWork() {
  const tasks = getTasks();
  
  function processTasksWhenIdle(deadline) {
    // Process tasks while time remains
    while (deadline.timeRemaining() > 0 && tasks.length > 0) {
      const task = tasks.shift();
      processTask(task);
    }
    
    // Schedule more work if tasks remain
    if (tasks.length > 0) {
      requestIdleCallback(processTasksWhenIdle);
    }
  }
  
  requestIdleCallback(processTasksWhenIdle, { timeout: 2000 });
}

Deadline API
Use cases
Caveats

requestIdleCallback((deadline) => {
  // Check remaining time in frame
  console.log('Time remaining:', deadline.timeRemaining());
  
  // Check if forced by timeout
  console.log('Did timeout:', deadline.didTimeout);
  
  if (deadline.timeRemaining() > 10) {
    // Enough time for expensive operation
    performExpensiveWork();
  } else {
    // Defer to next idle period
    requestIdleCallback(callback);
  }
});

deadline.timeRemaining():

Returns milliseconds until frame deadline
Typically 16.6ms - (work done so far)
May be 0 if frame is already late

// Analytics batching
const analyticsQueue = [];

function trackEvent(event) {
  analyticsQueue.push(event);
  
  requestIdleCallback(() => {
    if (analyticsQueue.length > 0) {
      sendAnalyticsBatch(analyticsQueue.splice(0));
    }
  });
}

// Preloading/prefetching
requestIdleCallback(() => {
  const images = document.querySelectorAll('img[data-src]');
  images.forEach(img => {
    const loader = new Image();
    loader.src = img.dataset.src;
  });
});

// Background computation
requestIdleCallback((deadline) => {
  while (deadline.timeRemaining() > 0 && hasMoreWork()) {
    processChunk();
  }
});

Idle callback limitations:

Not guaranteed to run (browser may never be idle)
Not suitable for time-critical work
Limited browser support (no Safari as of 2024)
Use timeout option for fallback execution

// Fallback for browsers without requestIdleCallback
const requestIdleCallback = window.requestIdleCallback || 
  function(callback) {
    const start = Date.now();
    return setTimeout(() => {
      callback({
        didTimeout: false,
        timeRemaining: () => Math.max(0, 50 - (Date.now() - start))
      });
    }, 1);
  };

Task prioritization

Modern browsers support task prioritization through the Scheduler API and manual chunking strategies.

Scheduler API (experimental)

// Priority-based task scheduling
async function schedulePrioritizedWork() {
  // User-blocking (highest priority)
  scheduler.postTask(() => {
    updateUIForUserInput();
  }, { priority: 'user-blocking' });
  
  // User-visible (high priority)
  scheduler.postTask(() => {
    renderVisibleContent();
  }, { priority: 'user-visible' });
  
  // Background (low priority)
  scheduler.postTask(() => {
    prefetchOffscreenContent();
  }, { priority: 'background' });
}

Priority levels
Manual chunking
Time slicing

Priority	Use case	Examples
`user-blocking`	Immediate response to input	Click handlers, keyboard input
`user-visible`	Visible updates	Rendering content, animations
`background`	Deferrable work	Analytics, prefetching

// Abort lower-priority work
const controller = new TaskController();

scheduler.postTask(() => {
  expensiveBackgroundWork();
}, {
  priority: 'background',
  signal: controller.signal
});

// Cancel if high-priority work arrives
document.addEventListener('click', () => {
  controller.abort(); // Cancel background work
  handleClick();
});

// Break long task into chunks
async function processLargeDataset(items) {
  const CHUNK_SIZE = 100;
  
  for (let i = 0; i < items.length; i += CHUNK_SIZE) {
    // Process chunk
    const chunk = items.slice(i, i + CHUNK_SIZE);
    processChunk(chunk);
    
    // Yield to event loop
    await scheduler.yield(); // Or: await new Promise(r => setTimeout(r, 0))
  }
}

scheduler.yield():

Yields to higher-priority work
Maintains task priority after resuming
Better than setTimeout(fn, 0) for prioritization

// Time-budget-based chunking
function timeSlicedProcessing(items, timeBudget = 5) {
  let index = 0;
  
  function processChunk() {
    const start = performance.now();
    
    while (index < items.length) {
      processItem(items[index++]);
      
      // Check time budget
      if (performance.now() - start > timeBudget) {
        // Exceeded budget, yield
        scheduler.postTask(processChunk, { priority: 'user-visible' });
        return;
      }
    }
    
    // Done processing
    onComplete();
  }
  
  processChunk();
}

Long task detection

// Monitor for long tasks (>50ms)
const observer = new PerformanceObserver((list) => {
  for (const entry of list.getEntries()) {
    if (entry.duration > 50) {
      console.warn('Long task detected:', {
        duration: entry.duration,
        startTime: entry.startTime,
        name: entry.name
      });
      
      // Consider breaking this task into chunks
    }
  }
});

observer.observe({ entryTypes: ['longtask'] });

Tasks over 50ms block user input and cause jank. Break long tasks into 50ms chunks to maintain 60fps responsiveness.

Complete event loop example

Here’s a comprehensive example demonstrating all queue types:

function comprehensiveEventLoopDemo() {
  console.log('1: Script start');
  
  // Macrotask
  setTimeout(() => {
    console.log('8: setTimeout macrotask');
    
    Promise.resolve().then(() => {
      console.log('9: Microtask after setTimeout');
    });
  }, 0);
  
  // Microtask
  Promise.resolve()
    .then(() => {
      console.log('3: Promise microtask 1');
      return Promise.resolve();
    })
    .then(() => {
      console.log('5: Promise microtask 2');
    });
  
  // Another microtask
  queueMicrotask(() => {
    console.log('4: queueMicrotask');
  });
  
  // Animation frame
  requestAnimationFrame(() => {
    console.log('6: Animation frame');
  });
  
  // Idle callback
  requestIdleCallback(() => {
    console.log('10: Idle callback');
  });
  
  console.log('2: Script end');
}

// Output:
// 1: Script start
// 2: Script end
// 3: Promise microtask 1
// 4: queueMicrotask
// 5: Promise microtask 2
// 6: Animation frame (before render)
// (render occurs)
// 8: setTimeout macrotask
// 9: Microtask after setTimeout
// 10: Idle callback (when browser is idle)

Best practices

Choose the right queue
Avoid starvation
Optimize for 60fps

Task type	Use
Critical state updates	Microtasks (`Promise.then`)
Async operations	Macrotasks (`setTimeout`)
Visual updates	`requestAnimationFrame`
Low-priority work	`requestIdleCallback`
Prioritized work	`scheduler.postTask`

// BAD: Microtask starvation
function infiniteMicrotasks() {
  Promise.resolve().then(() => {
    infiniteMicrotasks(); // Never yields!
  });
}

// GOOD: Yield periodically
function cooperativeMicrotasks(count = 0) {
  Promise.resolve().then(() => {
    doWork();
    
    if (count < 100) {
      cooperativeMicrotasks(count + 1);
    } else {
      // Yield to macrotask queue
      setTimeout(() => cooperativeMicrotasks(0), 0);
    }
  });
}

// Target: 16.6ms per frame for 60fps
const FRAME_BUDGET = 16;

function performanceAwareProcessing(items) {
  let index = 0;
  
  function processFrame(timestamp) {
    const frameStart = performance.now();
    
    while (index < items.length) {
      processItem(items[index++]);
      
      // Check frame budget
      if (performance.now() - frameStart > FRAME_BUDGET * 0.8) {
        // Used 80% of frame budget, yield
        requestAnimationFrame(processFrame);
        return;
      }
    }
  }
  
  requestAnimationFrame(processFrame);
}

Parsing & Execution

Runtime

Overview

Event loop architecture

Processing phases

Microtask queue

What creates microtasks

Microtask queue draining

Macrotask queue

What creates macrotasks

Macrotask processing

Animation frame callbacks

requestAnimationFrame timing

Render timing and optimization

Render pipeline

Idle callbacks and scheduling

requestIdleCallback

Task prioritization

Scheduler API (experimental)

Long task detection

Complete event loop example

Best practices

Next steps

Garbage collection

JIT optimization

Build docs developers (and LLMs) love

Parsing & Execution

Runtime

​Overview

​Event loop architecture

​Processing phases

​Microtask queue

​What creates microtasks

​Microtask queue draining

​Macrotask queue

​What creates macrotasks

​Macrotask processing

​Animation frame callbacks

​requestAnimationFrame timing

​Render timing and optimization

​Render pipeline

​Idle callbacks and scheduling

​requestIdleCallback

​Task prioritization

​Scheduler API (experimental)

​Long task detection

​Complete event loop example

​Best practices

​Next steps

Garbage collection

JIT optimization

Build docs developers (and LLMs) love

Overview

Event loop architecture

Processing phases

Microtask queue

What creates microtasks

Microtask queue draining

Macrotask queue

What creates macrotasks

Macrotask processing

Animation frame callbacks

requestAnimationFrame timing

Render timing and optimization

Render pipeline

Idle callbacks and scheduling

requestIdleCallback

Task prioritization

Scheduler API (experimental)

Long task detection

Complete event loop example

Best practices

Next steps