performance-audit

针对Hone进行性能问题审计。适用于优化数据库查询、分析Rust异步代码、审视React渲染逻辑，或排查运行缓慢的操作时使用。

SKILL.md

--- frontmatter

name: performance-audit
description: Audits Hone for performance issues. Use when optimizing database queries, analyzing Rust async code, reviewing React rendering, or investigating slow operations.
allowed-tools: Read, Grep, Glob, Bash

Performance Audit for Hone

Rust Backend (hone-core, hone-server)

Database Performance (db.rs)

•N+1 Queries: Watch for loops that execute queries
•Missing Indexes: Ensure indexes on transactions(account_id), transactions(date), subscriptions(merchant)
•Connection Pooling: Using r2d2, verify pool size is appropriate
•Batch Operations: Prefer batch inserts over individual inserts in loops

rust

// BAD - N+1 query pattern
for account in accounts {
    let txns = get_transactions_for_account(account.id)?; // Query per account!
}

// GOOD - Single query with JOIN or IN clause
let txns = get_all_transactions_for_accounts(&account_ids)?;

Async Performance (hone-server)

•Blocking in Async: Never use std::thread::sleep in async handlers
•Database Access: SQLite is synchronous - use spawn_blocking for heavy queries
•Ollama Calls: Ensure HTTP client reuses connections

rust

// BAD - blocks the async runtime
async fn handler() {
    std::thread::sleep(Duration::from_secs(1)); // Blocks!
}

// GOOD - use tokio's async sleep
async fn handler() {
    tokio::time::sleep(Duration::from_secs(1)).await;
}

Memory & Allocations

•Avoid .clone() in hot paths when references work
•Use &str instead of String for function parameters when possible
•Pre-allocate vectors with Vec::with_capacity() when size is known

React Frontend (ui/)

Rendering Performance

•Unnecessary Re-renders: Use React DevTools Profiler
•Missing Keys: List items need stable, unique keys
•Heavy Computations: Wrap in useMemo if expensive

tsx

// BAD - recalculates on every render
const total = transactions.reduce((sum, t) => sum + t.amount, 0);

// GOOD - only recalculates when transactions change
const total = useMemo(
  () => transactions.reduce((sum, t) => sum + t.amount, 0),
  [transactions]
);

Bundle Size

•Tailwind CSS tree-shakes unused styles (good)
•lucide-react supports tree-shaking (import specific icons)
•Check for accidentally importing entire libraries

API Efficiency

•Paginate large transaction lists
•Debounce search/filter inputs
•Cache responses where appropriate

Detection Algorithms (detect.rs)

Zombie Detection

•Group transactions efficiently (use HashMap)
•Avoid re-scanning entire transaction history

Price Increase Detection

•Index subscriptions by merchant for quick lookup
•Compare only recent transactions (last N months)

Duplicate Detection

•Pre-categorize subscriptions once, not per-check

Quick Checks

bash

# Check for .clone() usage
rg "\.clone\(\)" crates/

# Check for blocking sleep
rg "std::thread::sleep" crates/

# Check bundle size
cd ui && npm run build && ls -la dist/assets/