Performance Benchmarks

This page provides a performance reference for Oro, GORM, XORM, and Bun under the same benchmark scenarios and database drivers. Use it to compare relative ORM overhead, common business-path performance, and optimization direction.

:::note[Reading the Numbers] The main tables report the median from 10 consecutive go test -bench runs, not the fastest single run. Median is less sensitive to scheduler noise, database cache churn, and background tasks than a raw average. :::

Environment

MySQL and PostgreSQL use the local duxorm database. Networked database results are affected by database configuration, connection pooling, background load, and cache state.

Versions

The benchmark lives in the standalone Go module benchmarks/ormbench.

Fairness Rules

• SQLite: Oro, GORM, XORM, and Bun all use mattn/go-sqlite3.
• MySQL: all libraries use go-sql-driver/mysql.
• PostgreSQL: all libraries use pgx.
• Oro and GORM both disable default write transactions for high-throughput path comparison.
• Table shape, unique index, soft-delete field, and data size are kept equivalent.
• Create, query, update, and delete benchmarks execute real database operations.
• Each database is tested with -count=10 -benchtime=1s -benchmem.

Commands

SQLite with mattn/go-sqlite3:

source ~/.gvm/scripts/gvm && gvm use go1.27rc1
cd benchmarks/ormbench
ORO_BENCH_DRIVER=sqlite ORO_BENCH_SQLITE_DRIVER=mattn \
  go test -bench=. -benchmem -run '^$' -benchtime=1s -count=10 -timeout=0

MySQL:

source ~/.gvm/scripts/gvm && gvm use go1.27rc1
cd benchmarks/ormbench
ORO_BENCH_DRIVER=mysql \
  go test -bench=. -benchmem -run '^$' -benchtime=1s -count=10 -timeout=0

PostgreSQL:

source ~/.gvm/scripts/gvm && gvm use go1.27rc1
cd benchmarks/ormbench
ORO_BENCH_DRIVER=pgsql \
  go test -bench=. -benchmem -run '^$' -benchtime=1s -count=10 -timeout=0

Scenarios

Read and update benchmarks seed data before timing. Delete benchmarks seed enough rows so each operation deletes a unique row.

SQLite

SQLite uses the same CGO driver for all libraries: mattn/go-sqlite3.

SQLite Allocations

MySQL

MySQL uses go-sql-driver/mysql.

MySQL Allocations

PostgreSQL

PostgreSQL uses pgx.

PostgreSQL Allocations

Summary

• Oro is in the first tier for batch insert, indexed reads, list queries, updates, and deletes across common Go ORM workloads.
• Under the fair SQLite driver setup, Oro performs well on CreateMany100, FirstByCode, WhereList, and DeleteByCode.
• On MySQL, Oro leads this run for batch insert, read, update, and delete medians; single-row Create is not the fastest.
• On PostgreSQL, Oro is close to or ahead of the compared libraries for single-row create, batch insert, read, update, and delete; Bun is faster on WhereList in this run.
• Oro keeps allocation counts low on batch write and query paths, which indicates the model scan plan, SQL compilation cache, and batch insert path are working as intended.

How to Read This Page

• Compare relative trends on the same machine, database, and driver.
• Prefer the 10-run median over the fastest single run.
• Read B/op and allocs/op together with latency to understand ORM-layer allocation cost.
• Always state the SQLite driver; pure-Go and CGO SQLite drivers have very different costs.
• Treat these numbers as benchmark reference data, not a fixed production performance guarantee.