Observer Pattern

1. What Is It?

The Observer pattern defines a one-to-many dependency between objects so that when one object (the Subject) changes state, all its dependents (Observers) are notified and updated automatically.

Without it, you either poll the subject for changes (wastes resources, adds coupling) or hard-code update calls directly into the subject (tight coupling, violates SRP — the subject shouldn't know who cares about its state). Observer decouples the thing that changes from the things that react to that change.

2. How It Works

Participants:

• Subject (also called Observable) — maintains a list of observers; provides attach(observer), detach(observer), and notify() operations
• ConcreteSubject — stores state; calls notify() when its state changes
• Observer — interface declaring the update() method
• ConcreteObserver — implements update(); pulls or receives new state from the subject

Step-by-step:

1. Define the Observer interface with update()
2. Define the Subject interface with attach, detach, notify
3. ConcreteSubject holds state; on each state change, calls notify() which loops over registered observers and calls their update()
4. ConcreteObserver registers itself with the subject; on update(), reads new state and reacts

Push vs. Pull:

• Push model: Subject sends state data directly in the update(event) call — observer doesn't need to query back
• Pull model: Subject calls update() with no data; observer calls subject.getState() — looser coupling but more roundtrips

Python:

from abc import ABC, abstractmethod
from dataclasses import dataclass
from typing import List


@dataclass
class StockEvent:
    ticker_symbol: str
    price: float


class StockObserver(ABC):
    @abstractmethod
    def update(self, event: StockEvent) -> None:
        pass


class StockTicker:
    def __init__(self, ticker_symbol: str) -> None:
        self._ticker_symbol = ticker_symbol
        self._price: float = 0.0
        self._observers: List[StockObserver] = []

    def attach(self, observer: StockObserver) -> None:
        self._observers.append(observer)

    def detach(self, observer: StockObserver) -> None:
        self._observers.remove(observer)

    def set_price(self, price: float) -> None:
        self._price = price
        self._notify()

    def _notify(self) -> None:
        event = StockEvent(self._ticker_symbol, self._price)
        for observer in self._observers:
            observer.update(event)


class PriceAlertObserver(StockObserver):
    def __init__(self, alert_threshold: float) -> None:
        self._alert_threshold = alert_threshold

    def update(self, event: StockEvent) -> None:
        if event.price > self._alert_threshold:
            print(f"ALERT: {event.ticker_symbol} hit ${event.price:.2f} "
                  f"(threshold: ${self._alert_threshold:.2f})")


class PriceDashboardObserver(StockObserver):
    def update(self, event: StockEvent) -> None:
        print(f"Dashboard updated: {event.ticker_symbol} = ${event.price:.2f}")


# Usage
ticker = StockTicker("AAPL")
ticker.attach(PriceDashboardObserver())
ticker.attach(PriceAlertObserver(alert_threshold=180.0))

ticker.set_price(175.0)
# Dashboard updated: AAPL = $175.00

ticker.set_price(185.0)
# Dashboard updated: AAPL = $185.00
# ALERT: AAPL hit $185.00 (threshold: $180.00)

Java:

import java.util.ArrayList;
import java.util.List;

// Event data (push model)
public record StockEvent(String tickerSymbol, double price) {}

// Observer interface
public interface StockObserver {
    void update(StockEvent event);
}

// Subject (ConcreteSubject)
public class StockTicker {
    private final String tickerSymbol;
    private double price;
    private final List<StockObserver> observers = new ArrayList<>();

    public StockTicker(String tickerSymbol) {
        this.tickerSymbol = tickerSymbol;
    }

    public void attach(StockObserver observer) {
        observers.add(observer);
    }

    public void detach(StockObserver observer) {
        observers.remove(observer);
    }

    public void setPrice(double price) {
        this.price = price;
        notifyObservers();
    }

    private void notifyObservers() {
        StockEvent event = new StockEvent(tickerSymbol, price);
        for (StockObserver observer : observers) {
            observer.update(event);
        }
    }
}

// ConcreteObserver A
public class PriceAlertObserver implements StockObserver {
    private final double alertThreshold;

    public PriceAlertObserver(double alertThreshold) {
        this.alertThreshold = alertThreshold;
    }

    @Override
    public void update(StockEvent event) {
        if (event.price() > alertThreshold) {
            System.out.printf("ALERT: %s hit $%.2f (threshold: $%.2f)%n",
                    event.tickerSymbol(), event.price(), alertThreshold);
        }
    }
}

// ConcreteObserver B
public class PriceDashboardObserver implements StockObserver {
    @Override
    public void update(StockEvent event) {
        System.out.printf("Dashboard updated: %s = $%.2f%n",
                event.tickerSymbol(), event.price());
    }
}

// Client
public class Main {
    public static void main(String[] args) {
        StockTicker ticker = new StockTicker("AAPL");
        ticker.attach(new PriceDashboardObserver());
        ticker.attach(new PriceAlertObserver(180.0));

        ticker.setPrice(175.0);  // Dashboard updated: AAPL = $175.00
        ticker.setPrice(185.0);  // Dashboard + Alert
    }
}

3. Variants & Comparisons

Java: java.util.Observable (deprecated in Java 9+) and java.util.Observer were the original GoF implementation. Modern Java prefers PropertyChangeListener (from java.beans) or reactive streams (RxJava, Project Reactor). Python: Simple observer lists (as above) are common. The built-in signal module handles OS signals, not application events. __slots__ + property setters can trigger observers cleanly.

4. When to Use It (and When NOT To)

Use when:

• Multiple components need to react to state changes in another component (UI, event systems, stock feeds)
• You want to avoid polling
• The subject shouldn't know the concrete types of its observers

Don't use when:

• The chain of notifications is complex — cascading updates become hard to trace (who triggered what?)
• Observers are always the same set and won't change — just call them directly
• Notification order matters and must be guaranteed — Observer gives no ordering guarantees

Decision triggers:

• "When X changes, Y and Z need to know about it, but X shouldn't know about Y or Z" → Observer
• "I keep adding new notify*() methods to a class every time a new consumer appears" → Observer

Anti-patterns:

• Memory leaks: Forgetting to detach observers. A subject holding strong references to observers prevents GC. Use weak references or explicit lifecycle management.
• Cascading updates: Observer A's update triggers Subject B, which notifies Observer C, which loops back — hard to debug, can cause infinite loops.

5. Real-World Usage

1. Java Swing event listeners: Every Swing widget is a Subject. button.addActionListener(listener) registers an Observer. ActionListener.actionPerformed() is update(). Swing's entire event model is Observer.

2. Android LiveData / ViewModel: LiveData<T> is the Subject. UI controllers call liveData.observe(lifecycleOwner, observer). When data changes, all active observers are notified. The lifecycle owner ensures automatic detach on destruction — solving the memory leak anti-pattern.

3. RxJava / Project Reactor: The entire reactive programming model is Observer formalized. Observable/Flux are subjects; Observer/Subscriber receive events. They add back-pressure, threading control, and operator pipelines on top of the core pattern.

6. Interview Cheat Sheet

Key sentences:

1. "Observer decouples the thing that changes (Subject) from the things that react (Observers) — the subject only knows about the Observer interface, not concrete implementations."
2. "Push model sends data to observers; pull model lets observers query. Push is more decoupled; pull avoids sending irrelevant data."
3. "The classic pitfall is memory leaks — if you don't detach observers, the subject holds references forever, preventing garbage collection."
4. "In modern Java, PropertyChangeListener and reactive streams (Project Reactor, RxJava) are the production-grade evolution of this pattern."

Common follow-up questions:

Q: What's the difference between Observer and Event Bus? A: Observer requires observers to register directly on the subject. An Event Bus (Mediator variant) adds a central broker — subjects and observers never reference each other, at the cost of more indirection.

Q: How do you prevent memory leaks in Observer? A: Use weak references (Python weakref, Java WeakReference), enforce explicit detach() calls on lifecycle events, or use a framework that handles registration lifecycle (e.g., Android LiveData tied to a LifecycleOwner).

Q: What about thread safety? A: The basic pattern isn't thread-safe. If observers are added/removed concurrently, the observer list needs synchronization (e.g., CopyOnWriteArrayList in Java for safe iteration).

Connections:

• Observer is the foundation of Model-View-Controller — Model is the Subject, Views are Observers
• Observer + Mediator → Event Bus pattern
• Observer is the imperative precursor to Reactive Streams (RxJava, Project Reactor)