caddis-plugin: Instructions file for GitHub Copilot

Performance Optimization Guidelines

General Principles

1. Measure first — Profile before optimizing. Use real data and production-like conditions.
2. Optimize the common case — Focus on hot paths and frequently executed code.
3. Avoid premature optimization — Write clear, correct code first. Optimize when measurements prove a bottleneck.
4. Set performance budgets — Define acceptable latency, throughput, and memory targets before starting.
5. Optimize at the right level — Algorithm > architecture > code micro-optimization.

---

Backend Performance

Algorithm and Data Structure Selection

• Choose appropriate data structures: dict/set for O(1) lookups, sorted containers for range queries
• Prefer O(n log n) or better algorithms; avoid O(n²) in hot paths
• Use generators and iterators for large data pipelines to avoid loading everything into memory

Caching

from functools import lru_cache

# In-memory cache for expensive, repeated computations
@lru_cache(maxsize=256)
def get_exchange_rate(currency_code: str, date: str) -> float:
    """Cached exchange rate lookup — date string ensures daily invalidation."""
    return fetch_rate_from_service(currency_code, date)

• Cache at the right granularity (per-request, per-user, global)
• Always set TTL (time-to-live) — stale data causes subtle bugs
• Implement cache invalidation on writes
• Protect against cache stampede with locking or staggered expiry

Async I/O and Concurrency

import asyncio
import aiohttp

async def fetch_all_endpoints(urls: list[str]) -> list[dict]:
    """Fetch multiple endpoints concurrently."""
    async with aiohttp.ClientSession() as session:
        tasks = [session.get(url) for url in urls]
        responses = await asyncio.gather(*tasks, return_exceptions=True)
        results = []
        for resp in responses:
            if isinstance(resp, Exception):
                results.append({"error": str(resp)})
            else:
                results.append(await resp.json())
        return results