:rocket: sqlx4k dev update — what's landing next ...

# feed

[Yorgos S.]

🚀 sqlx4k dev update — what's landing next

Quick heads-up on two features coming to sqlx4k (the coroutine-first SQL toolkit for Kotlin Multiplatform — Postgres / MySQL / MariaDB / SQLite, JVM + Native + Android). Both lean hard into the project's core idea: catch it at compile time, not in prod. 👇

🔍 1. Query validations & optimizations (compile-time)

We already validate @Query SQL for syntax and schema.
The next step parses each query once and runs a set of extra safeguards + codegen rewrites — all through KSP, all on by default, each toggleable per module.

Validations that fail the build:
• Column validation — reference a column that doesn't exist on the entity (in SELECT / WHERE / GROUP BY / HAVING / ORDER BY / SET)? Build fails.
• Table validation — a single-table query pointing at the wrong table.
• Statement-kind check — find*/count*/exists* must be SELECT, delete* must be DELETE, etc.
• Projection check — a find* bound to the generated row mapper must select every column (no half-mapped rows at runtime).
• Stacked-statement guard — one @Query, one statement.

Optimizations the processor applies for you:
• SELECT * expansion — select * from users → select id, name, email from users in the generated statement.
• findOne* limiting — appends LIMIT 2 so the driver still detects multi-row results but transfers far less on wide tables.

Everything maps columns ↔️ properties with the same rules as the CRUD generator (snake_case / @Column), so it always stays consistent with the SQL that gets generated. Don't like one of them? arg("findone-limit", "false") in your KSP block and it's gone.

🧪 2. In-memory repositories (for unit testing) (experimental)

A companion module, sqlx4k-codegen-test, generates a thread-safe in-memory implementation of every @Repository — so you can unit-test the code around your repos with no database at all. For Sqlx4kRepository you get InMemorySqlx4kRepository.

val repo = InMemorySqlx4kRepository()
// A QueryExecutor is accepted to match the interface, but never touched.
repo.insert(db, Sqlx4k(id = 1, test = "test")).getOrThrow()
val all = repo.findAll(db).getOrThrow()

What you get out of the box:
• Full CRUD + save / batchInsert / batchUpdate, with auto-increment numeric ids.
• Whole-table findAll / countAll / deleteAll.
• Derived @Query methods — the in-memory behavior is derived from the query's actual SQL: the WHERE clause is parsed into a predicate over stored entities. @Query("select * from users where name = :name") just behaves like filter { it.name == name }. Supports = <> < <= > >=, IS [NOT] NULL, AND/OR, named params & literals.
• Test helpers — clear(), findAllStored(), and a withStore { } block for dirmap.

Anything the generator can't translate (LIKE, IN, subqueries, function callsws NotImplementedError — and since the class is open, you just override thosefew methods yourself. It's opt-in per KSP config (kspTest), so it never touches your production wiring.

Heads-up: this one's early/experimental — the generated API may still shift, so feedback is very welcome. 🙏

Repo & full docs: github.com/smyrgeorge/sqlx4k — would love to hear what you think, especially on the in-memory repos. 💬