Add benchmarks to inform performance-oriented refactoring · launchbadge/sqlx#158

(0 comments) (1 reaction) (0 assignees)Rust (13,916 stars) (1,307 forks)batch import

enhancementhelp wanted

Description

Establish a new connection.
Checkout an existing ( and idle ) connection from the pool.
Ping a connection.
Small query, small results ( following the specs over at tech empower: https://github.com/TechEmpower/FrameworkBenchmarks/wiki/Project-Information-Framework-Tests-Overview#single-database-query ). Do nothing with the results.
Small query, huge results ( 10K rows with at least 5 columns where 2 of the columns are String and Vec<u8> ). Do nothing with the results.

~~Preferably these can run locally with a single script that uses docker to spin up/down database servers as needed to test all supported databases and versions.~~

I'm imagining a bench/ or perf/ folder that contains a run.sh script that does the "magic" but I'm pretty open here.

Update: We now have a docker-compose file that can spin up a database/version pair. This was one of the blockers here.

Contributor guide

Tech stack: rustpostgresqlmysqlsqlite
Domain: backenddatabaseperformance
Issue type: feature
Difficulty: 3
Estimated time: 1-2 days
Activity status: stale
Clarity: clear
Prerequisites: Rust knowledgesqlx basic usageDocker basics
Newbie friendliness: 60
Research direction: Review the existing sqlx repository structure to understand how database connections are handled. Use the docker compose file mentioned in the issue to spin up database servers for PostgreSQL, MySQL, and SQLite. Then create a `bench/` folder with a `run.sh` script that executes benchmarks for each scenario listed in the issue: establishing a connection, checking out and pooling idle connections, pinging a connection, a single small query, and a small query with large results. The benchmarks should be written in Rust using a benchmarking crate like `criterion`.