Vercel React Best Practices

Vercel React Best Practices — StarMapper

Performance optimization guide for React 19 + Next.js 16.2 on Vercel free tier.

When to Apply

• Writing new React components or Next.js pages
• Implementing data fetching (chunk loop, cache reads)
• Reviewing components for performance issues
• Optimizing bundle size (MapLibre is heavy at ~900KB)
• Refactoring existing code

Rule Categories by Priority

Quick Reference

1. Eliminating Waterfalls (CRITICAL)

• async-parallel — Use Promise.all() for independent operations

  // ✅ parallel: fetch repo info and check cache simultaneously
  const [repoInfo, cached] = await Promise.all([
    fetchRepoInfo(owner, repo),
    checkStargazerCache(owner, repo),
  ]);
  

• async-api-routes — Start promises early, await late

  // ✅ start DB lookup early
  const cachePromise = db.stargazerCache.findUnique({ where: { owner_repo: { owner, repo } } });
  // ... other sync work ...
  const cached = await cachePromise;
  

• async-defer-await — Only await in the branch that needs it

2. Bundle Size Optimization (CRITICAL for MapLibre)

• bundle-dynamic-imports — MapLibre GL is ~900KB, always dynamic:

  // ✅ correct — StarMapper uses this pattern
  import dynamic from "next/dynamic";
  const StargazerMapDynamic = dynamic(() => import("./stargazer-map"), { ssr: false });
  

• bundle-barrel-imports — Import directly, avoid index.ts barrel files:

  // ❌ barrel (pulls in everything)
  import { geocode, geocodeBatch } from "@/lib";
  // ✅ direct import
  import { geocode } from "@/lib/geocoder";
  

• bundle-defer-third-party — Load analytics/non-critical scripts after hydration

• bundle-conditional — Only load MapLibre when user navigates to map page (Next.js App Router handles this naturally via route segments)

3. Server-Side Performance (HIGH)

• server-cache-react — Use React.cache() for per-request deduplication in Server Components:

  import { cache } from "react";
  const getCachedRepoInfo = cache(async (owner: string, repo: string) => {
    return fetchRepoInfo(owner, repo);
  });
  

• server-serialization — Minimize data passed to client components:

  // ❌ pass entire object with 50 fields
  <StargazerMap repoData={fullRepoData} />
  // ✅ only what's needed
  <StargazerMap owner={owner} repo={repo} stars={stars} />
  

• server-parallel-fetching — Parallel Server Component data fetching for map page

4. Re-render Optimization (MEDIUM)

• rerender-memo — memo() for StargazerMap (re-renders on every points update):

  export const StargazerMap = memo(({ points, ... }: Props) => {
    // ...
  });
  

• rerender-dependencies — Use primitive dependencies in effects:

  // ❌ object reference changes every render
  useEffect(() => { ... }, [repoData]);
  // ✅ primitive values
  useEffect(() => { ... }, [owner, repo]);
  

• rerender-memo — Extract GeoJSON feature array into useMemo:

  const geojsonData = useMemo(() => ({
    type: "FeatureCollection" as const,
    features: points.map(pointToFeature),
  }), [points]);
  

• rerender-functional-setstate — Use functional setState for chunk loop accumulation:

  setPoints(prev => [...prev, ...newPoints]);
  

5. Rendering Performance (MEDIUM)

• rendering-hydration-no-flicker — Use inline script for dark/light theme (StarMapper already does this in layout.tsx):

  // ✅ inline script prevents FOUC on theme load
  

• rendering-conditional-render — Use ternary, not && for conditional rendering:

  // ❌ risky with 0/falsy values
  {count && <Badge count={count} />}
  // ✅ explicit
  {count > 0 ? <Badge count={count} /> : null}
  

6. JavaScript Performance (LOW-MEDIUM)

• js-index-maps — Build Map for repeated lookups (e.g., country name normalization):

  // Build once, lookup many times
  const countryMap = new Map(countries.map(c => [c.code, c.name]));
  

• js-set-map-lookups — Use Set for has() checks over includes():

  const unmappedSet = new Set(unmapped.map(u => u.login));
  const isUnmapped = unmappedSet.has(login); // O(1) vs O(n)
  

• js-early-exit — Return early from geocoding functions

StarMapper-Specific Notes

• MapLibre GL bundle (~900KB) is the single biggest bundle concern. Dynamic import with ssr:false is mandatory and already in place via stargazer-map-dynamic.tsx.
• Chunk loop state accumulates progressively — setPoints(prev => [...prev, ...newPoints]) is correct functional setState pattern.
• GeoJSON construction is O(n) with n=stars — useMemo on points array is important for large repos.
• Vercel free = no Edge Runtime needed — standard Node.js serverless functions work fine.