Flask — using HMR in development¶

Use HMR to reload Python code in-place while developing Flask apps. HMR reruns changed modules and their dependents instead of restarting the whole process, shortening the edit-test loop and preserving in-memory objects when possible.

Install¶

pip install hmr

Run¶

hmr app.py
# or for package entry
hmr -m mypackage

What to Observe¶

Once the server is running, you can access the application at http://localhost:5000.

• Visit different endpoints and modify the response logic in your modules.
• Try modifying route handlers and refresh the browser to see changes applied instantly.
• Everything should work as expected, with a much smoother development experience than flask run --reload.

Important notes¶

• Disable Flask's builtin debug mode reloader. Do not run both HMR and Flask's reloader — run hmr directly.
• HMR preserves state only when modules are not re-executed. When a module is changed, it and its dependents are re-executed.
• Native extensions, C-level globals, or code that relies on process-level initialization may not be safe to hot-reload.

Best practices¶

• Move heavy initialization (DB pools, ML models) to modules you edit rarely.
• Make route registration idempotent to avoid double-registering on reload.
• Prefer small, pure route handlers and minimize module-level state.

Example sketch¶

# app.py
from flask import Flask, jsonify
from reactivity import signal

counter = signal(0)  # counter can be a reactivity signal

app = Flask(__name__)

@app.get("/count")
def get_count():
    return jsonify({"count": counter.get()})

If reloads behave oddly, reproduce in a minimal example and check for double registration or global mutable state. For protocol or C-level breaks, prefer a full restart.

See examples/flask/ for a concrete example.