Feature Flag Implementation Examples | Generated by AI

Home 2025.12

Question: How do you actually use feature flags in code, and can you show concrete examples in a few popular languages?

Answer:

Feature flags let you merge code to main and deploy it to production while keeping the feature completely hidden (or partially visible) until you’re ready. You control the toggle externally (dashboard, CLI, or config file) — no new deploy needed to turn it on/off.

Core principles when writing flagged code

  1. Always safe by default (flag OFF = old, proven behavior).
  2. Keep flag checks cheap and synchronous at startup or request time.
  3. Clean up the flag + dead code within 30–90 days.
  4. Name flags clearly: new-checkout-flow-2025q1, ai-search-backend-v2, etc.

Real-world examples

1. JavaScript / TypeScript (React + LaunchDarkly example)

import { useFlags, useLDClient } from 'launchdarkly-react-client-sdk';

function CheckoutButton() {
  const { newCheckoutFlow2025q1 } = useFlags();

  if (newCheckoutFlow2025q1) {
    return <NewFancyCheckout />;
  }

  return <OldCheckout />;
}

Percentage rollout example:

const variant = ldClient.variation('new-checkout-flow-2025q1', { key: user.id }, 'control');
if (variant === 'new') { /* ... */ }

2. Python (Flask/Django + Unleash or simple Redis)

from flask import Flask
import requests

app = Flask(__name__)

def is_feature_enabled(feature_name, user_id=None):
    # Simple Unleash example
    response = requests.get(f"https://unleash/api/features/{feature_name}")
    data = response.json()
    return data['enabled']

@app.route("/dashboard")
def dashboard():
    if is_feature_enabled("new-dashboard-v2", user_id=current_user.id):
        return render_template("new_dashboard.html")
    return render_template("old_dashboard.html")

Even simpler with a local config service:

if settings.FEATURE_FLAGS.get('ai-search-enabled', False):
    results = call_openai_search(query)
else:
    results = call_legacy_search(query)

3. Go (with Unleash Go client)

if client.IsEnabled("payment-retry-logic-v3", unleash.WithContext(ctx)) {
    doNewRetryLogic()
} else {
    doOldRetryLogic()
}

// Percentage rollout to 10% of users
if client.IsEnabled("new-onboarding", unleash.WithFallback(false), unleash.WithContext(userContext)) {
    showNewOnboarding()
}

4. Ruby on Rails (with Flipflop or Rollout)

# config/features.yml or dashboard toggle
if Flipflop.new_checkout_flow?
  render "checkouts/new_flow"
else
  render "checkouts/legacy"
end

# Gradual rollout to 5% of users + all staff
Flipflop.new_checkout_flow? do |user|
  actor.staff? || Random.rand(100) < 5
end

5. Feature-flagged database migration pattern

You often need this for schema changes:

# Step 1: Add column with flag off (no code uses it yet)
add_column :users, :preferred_language, :string

# Step 2: Deploy code that writes to both old and new places when flag is on
if feature_enabled?(:new-language-storage)
  user.update(preferred_language: lang)           # new column
end
user.update(language: lang)                         # old column (still used)

# Step 3: Back-fill data when 100% rolled out
# Step 4: Flip flag permanently on, remove old code/column

Common rollout strategies (done in the flag dashboard, no code change)

Strategy Use case
Boolean on/off Internal tools, admin features
% of users Canary releases (0.1% → 100%)
Specific users/emails Beta testers, dogfooding
Environment (prod/staging) Keep it off in staging if you want
Country / tenant Launch only in EU first

Real examples from big companies

Once you get comfortable, you’ll start treating feature flags as the primary way you release anything non-trivial — and deployments truly become boring.

References:

Back

x-ai/grok-4.1-fast

Donate