debugging-orm-queries

galihcitta 的头像作者 galihcitta

将 ORM 调用转换为原始 SQL 并分析查询性能。检测 N+1 查询、缺失索引和其他反模式。在调试慢查询、跟踪 ORM 生成的 SQL 或优化 Sequelize、Prisma、TypeORM (Node.js)、GORM、sqlc、sqlx、ent (Go) 或 SQLAlchemy、Django ORM、Peewee (Python) 的数据库性能时使用。

📖 中文导读:将 ORM 调用转换为原始 SQL 并分析查询性能。检测 N+1 查询、缺失索引和其他反模式。在调试慢查询、跟踪 ORM 生成的 SQL 或优化 Sequelize、Prisma、TypeORM (Node.js)、GORM、sqlc、sqlx、ent (Go) 或 SQLAlchemy、Django ORM、Peewee (Python) 的数据库性能时使用。 以下正文为该技能的原始 SKILL.md 内容(英文),上方安装命令可直接使用。

name: debugging-orm-queries description: Converts ORM calls to raw SQL and analyzes query performance. Detects N+1 queries, missing indexes, and other anti-patterns. Use when debugging slow queries, tracing ORM-generated SQL, or optimizing database performance for Sequelize, Prisma, TypeORM (Node.js), GORM, sqlc, sqlx, ent (Go), or SQLAlchemy, Django ORM, Peewee (Python).

Debugging ORM Queries

References

Language ORMs Reference
Node.js Sequelize, Prisma, TypeORM references/nodejs.md
Go GORM, sqlc, sqlx, ent references/golang.md
Python SQLAlchemy, Django, Peewee references/python.md
Anti-patterns N+1, indexes, pagination references/anti_patterns.md

Scripts

# Pretty-print SQL
python scripts/sql_formatter.py "SELECT..."

# Detect N+1, missing WHERE, duplicates from log file
python scripts/query_analyzer.py queries.log [--json]

# Parse EXPLAIN output (PostgreSQL or MySQL)
psql -c "EXPLAIN ANALYZE ..." | python scripts/explain_parser.py --postgres
mysql -e "EXPLAIN FORMAT=JSON ..." | python scripts/explain_parser.py --mysql

# Node.js request-scoped query tracking (see script for setup)
# Integrates with Sequelize, Prisma, TypeORM

Quick Patterns

Enable logging: Check reference for ORM-specific config (usually logging: true or echo=True)

ORM → SQL: Enable logging, run query, capture output

SQL → ORM: Map clauses to methods:

  • SELECT columns → specify fields/attributes
  • WHERE → where/filter with operators
  • JOIN → include/preload/prefetch_related
  • ORDER BY → order/orderBy
  • LIMIT/OFFSET → limit/offset or cursor-based

Performance issues: Run query_analyzer.py on logs, check anti_patterns.md for solutions