Practical Performance Optimization in React

Practical Performance Optimization in React Kemet Dugue kemetdugue kdugue

Agenda 1. Brief Introduction 2. Sensible Optimization & Popular Use Cases a. Unnecessary Re-rendering & Expensive Computations b. Lists & Virtualization 3. Tools for Debugging (React Dev Tools) 4. Performance Best Practices 5. Q+A

About Me ● Engineer at ● Interests: ○ lo-fi house ○ tennis ○ cats Whatever brings you joy!

Opinions and Q+A

Sensible Optimization aka When is the right time to optimize? ● App functionality slowness ○ Scrolling ○ Updating of views ○ Click events ● React component design (use best practices) ● Don’t prematurely optimize!

Popular Optimization Use Cases ● Unnecessary re-renders ● Expensive Computations ● Collections (small and large) ● Large App Bundle*

Component to DOM (Two Phases) 1. Render phase ○ JSX transformed ○ Can be called multiple times for a given change (re-render) i. Component state change ii. Props Change iii. (in case of children) Parent re-renders ○ 2. Reconciliation: i. “Diffing” algorithm: determines parts of the node tree that have to be updated ii. Current virtual DOM to previous virtual DOM Commit phase ○ changes to actual DOM (commits to changes resulted from reconciliation); ○ Called only once for a given change

● Assume props are rarely changing

Memoize Expensive Calculations/ Components ● Use for Expensive calculations ● React.memo for functional components ● useMemo to memoize values ● PureComponent for class components

Rendering Collections ● Using index as key can be sometimes deceptive ● Still, index as key can negatively affect performance Credit: Robin Pokorny

Large Collection Solution: List Virtualization (Windowing) ● Rendering 1000’s of rows ● Virtualization: rendering only what is in the viewport ● Popular Libraries: ○ react-window ○ react-virtualized ● Tradeoff: pay in rendering time later rather than upfront (initial load) Credit: web.dev

Todo App with React Dev Tools!

Best Practices ● Make sure optimization is a sensible one! ○ Every optimization comes with a tradeoff ○ Measure performance first ● Don’t pass new references ( arrays, objects, functions) as props to components. ● Use a unique and stable key value for a list. Minimum usage is index (be cautious) ● Pass only the props that are needed by the component (no “…props”) ● Consider PureComponents (class) or React.memo (functional) for the following scenarios: props that rarely change, components that renders often ● useMemo to memoize expensive calculations

Best Practices (cont’d) ● useCallback* to maintain callback reference ○ EX: child component is already memoized and takes in that callback as prop ● Use React Dev Tools for debugging components. ● Don’t query DOM directly (EX: document.querySelector) ● Keep constructor() light

Resources ● React official docs: https://reactjs.org/ ● React Dev Tools Tutorial: https://react-devtools-tutorial.vercel.app/ ● List Virtualization Libraries ○ react-window: https://github.com/bvaughn/react-window ○ react-virtualized: https://github.com/bvaughn/react-virtualized ● Code-Splitting (for large app bundles): https://reactjs.org/docs/code-splitting.html

Questions????

Practical Performance Optimization in React Kemet Dugue kemetdugue kdugue