Smash Lab is a SQLite database for a badminton racket-restringing shop. It tracks customers, staff, strings, orders and individual stringing jobs, from racket drop-off through to contactless locker pickup.
Result: indexes on the shop's most-filtered columns cut the seven everyday queries from 5.6 s to 31 ms (about 180× faster) on a test dataset of 90,000 stringing jobs. See benchmark results.
A customer drops off one or more rackets, which creates an order. Each racket in that order is a string job, done by one staff member using one string from the shop's catalog. When every job in an order is done, the order moves to a locker, and the customer collects it with a locker number and password.
.
├── sql/
│ ├── schema.sql # tables, constraints and indexes
│ ├── seed.sql # sample customers, staff, strings, orders and jobs
│ └── queries.sql # the shop's seven everyday queries
├── benchmark/
│ ├── bench.py # times the queries with and without indexes
│ └── RESULTS.md # benchmark output
└── docs/
├── DESIGN.md # design document: scope, entities, optimizations, limitations
└── images/
└── er-diagram.png
SQLite comes preinstalled on macOS and most Linux systems. From the repo root:
sqlite3 smashlab.db < sql/schema.sql
sqlite3 smashlab.db < sql/seed.sql
sqlite3 -box smashlab.db < sql/queries.sqlTo browse the tables in a GUI, open smashlab.db in DB Browser for SQLite.
| Entity | Description |
|---|---|
customers |
Customer details such as name, contact number and signup date. |
staff |
Employees who carry out the stringing jobs. |
strings |
Catalog of string types, specifications and prices. |
orders |
A customer's drop-off, containing one or more string jobs, with its locker and collection status. |
string_job |
One racket's stringing job: string, colour, tension, cost and status. |
Three indexes cover the columns the everyday queries filter on:
| Index | Speeds up |
|---|---|
string_job(order_id) |
Looking up an order's jobs, totals and cost breakdown |
string_job(job_status) |
The pending job queue |
orders(collection_status) |
Locker status checks and finding orders ready for collection |
Benchmark on synthetic data (50,000 orders, 90,034 string jobs), median time per query:
| Query | No indexes | With indexes |
|---|---|---|
| Orders ready for collection | 5,532 ms | 0.74 ms |
| Job specs for an order | 4.8 ms | 0.011 ms |
| Pending job queue | 7.0 ms | 0.80 ms |
| All seven queries | 5,585 ms | 31 ms |
Rerun it with python3 benchmark/bench.py (standard library only, takes a few minutes). It rewrites benchmark/RESULTS.md, which also has results at two other dataset sizes and the query plans.
- Constraints for data integrity: tension limited to 22–34 lbs, allowed values for statuses and string colours, unique contact numbers
- Default statuses for new jobs (
Pending) and orders (In-Store) - Contactless pickup through locker number and password
- Order totals calculated from their jobs rather than stored twice
The full design document, covering scope, functional requirements, each entity's columns, and known limitations, is in docs/DESIGN.md.
- Database: SQLite
- Benchmark: Python 3 (standard library)
- Tools: DB Browser for SQLite, VS Code

