SQL-first data access layer for PostgreSQL, with fluent query builders, automatic composite/enum/array type mapping, polymorphic dynamic DTOs, transaction plans, stored-procedure support, and LISTEN/NOTIFY.
An explicit, SQL-first data access layer for Kotlin & PostgreSQL
It's not an ORM. It's a ROME (Relational-Object Mapping Engine). Because all queries lead to ROME.
Just as Augustus brought order to a republic torn apart by the chaos of unchecked power, Octavius brings order to the chaotic republic of database interactions. The Senate of abstraction is dissolved. SQL rules supreme.
Octavius was built to bring order to the chaotic republic of database interactions. It rejects the unpredictable "magic" of traditional ORMs and returns the power to the rightful ruler: SQL.
| Principle | Description |
|---|---|
| Query is Imperator | Your SQL query dictates the shape of data — not the framework. |
Every design choice in Octavius is intentional. The reasoning behind them is laid out in the Design Philosophy.
// Define your data class — it maps directly to query results
data class Legionnaire(val id: Int, val name: String, val rank: String)
// Query with named parameters
val legionnaires = dataAccess.select("id", , )
.from()
.()
.orderBy()
.toListOf<Legionnaire>( to )
Automatic conversion between PostgreSQL and Kotlin types.
* Supports PostgreSQL infinity values (infinity, -infinity). See Type System for details.
Arrays of all standard types are supported and map to List<T>.
Extend the type system for any PostgreSQL type (e.g., circle, ltree) by implementing GlobalTypeHandler<T>. Handlers are automatically discovered via classpath scanning.
object PgCircleHandler : GlobalTypeHandler<PgCircle> {
override val pgTypeName = "circle"
override val kotlinClass = PgCircle::class
override val fromPgString = { s: String -> /* parse <(x,y),r> */ }
override val toPgString = { c: PgCircle -> "<(${c.x},),>" }
}
Need to change a mapping, bypass reflection, or return a composite as a Map for just one specific query? Use the .options() block without affecting global state:
val results = dataAccess.select("*").from("classified_reports")
.options {
registerTypeHandler(LegacyDateHandler)
returnCompositeAsMap("metadata")
}
.toListOf<Report>()
See Type System: Per-Query Configuration for full details.
dynamic_dto & dynamic_map)Octavius provides a powerful bridge between PostgreSQL and Kotlin's type system using the dynamic_dto (JSONB-based storage) and dynamic_map (ad-hoc projections) composite types.
They allow you to map complex, nested, or polymorphic data on the fly without creating strict database schema types for every nested object.
These types are automatically initialized in the public schema on startup.
Construct Kotlin objects directly in SQL using jsonb_build_object — no need to define PostgreSQL COMPOSITE types. Perfect for JOINs and projections where you want nested results without schema changes.
@DynamicallyMappable(typeName = "citizen_profile")
@Serializable
data class CitizenProfile(val tribe: String, val rights: List<String>)
( id: , name: String, profile: CitizenProfile)
citizens = dataAccess.rawQuery().toListOf<CitizenWithProfile>()
Why use this? Usually, to get a citizen with their profile in one query, you'd fetch flat columns (
citizen_id,citizen_name,profile_tribe...) and manually map them, create a database VIEW or COMPOSITE. With ad-hoc mapping, you construct the nested structure directly in SQL. The database does the packaging, Octavius does the unpacking — zero boilerplate.
Store different entity types in a single table and query them safely as a list of Kotlin interfaces.
Octavius stays true to its SQL-first philosophy. Invoke functions and procedures directly using native PostgreSQL syntax:
// Functions (SELECT * FROM func)
val result = dataAccess.select("*").from("calculate_tribute(@province, @year)")
.toField<Int>("province" to "Britannia", "year" to 43)
// Procedures (CALL proc)
val result = dataAccess.rawQuery("CALL register_conscript(@legion_id, @new_rank)")
.toSingleStrict(
to ,
to .withPgType()
)
Build complex WHERE clauses without SQL injection risks:
fun buildFilters = listOfNotNull(
name?.let { withParam ( to ) },
minRank?.let { withParam ( to it) },
province?.let { withParam ( to it) }
).join()
filter = buildFilters(name = , minRank = , province = )
senators = dataAccess.select()
.from()
.(filter.sql)
.toListOf<Senator>(filter.params)
Octavius supports two powerful interaction patterns for atomic operations.
The simplest way to execute multiple operations. Transactions follow a fail-fast policy: they are automatically rolled back if the block returns DataResult.Failure or throws an exception.
val result = dataAccess.transaction {
val citizenId = insertInto("citizens")
.value("name")
.returning("id")
.toField<Int>("name" to "Marcus Aurelius")
.getOrElse { return@transaction it }
insertInto()
.values(listOf(, ))
.execute( to citizenId, to )
.getOrElse { it }
DataResult.Success(citizenId)
}
Execute multi-step operations with complex dependencies between steps. Results from previous steps can be referenced in subsequent steps without nested callbacks or manual state management.
val plan = TransactionPlan()
// Step 1: Record the edict, get handle to future ID
edictIdHandle = plan.add(
dataAccess.insertInto()
.values(listOf(, ))
.returning()
.asStep()
.toField<>(mapOf( to consulId, to tribute))
)
(item levyItems) {
levyItem: Map<String, Any?> = mapOf(
to edictIdHandle.field(),
to item.provinceId,
to item.amount
)
plan.add(
dataAccess.insertInto()
.values(levyItem)
.asStep()
.execute(levyItem)
)
}
dataAccess.executeTransactionPlan(plan)
Subscribe to PostgreSQL channels and receive real-time notifications as a Kotlin Flow:
// Send a notification
dataAccess.notify("legion_dispatch", "legion_id:VII")
// Listen on a dedicated connection (outside the HikariCP pool)
dataAccess.createChannelListener().use { listener ->
listener.listen("legion_dispatch", "senate_decrees")
listener.notifications()
.collect { notification ->
when (notification.channel) {
"legion_dispatch" -> handleDispatch(notification.payload)
"senate_decrees" -> handleDecree(notification.payload)
}
}
}
Each PgChannelListener holds its own dedicated JDBC connection, separate from the query pool. Notifications sent inside a transaction are only delivered after commit.
Octavius distinguishes between Database Execution Errors (returned safely) and Fatal Developer Errors (thrown).
val result = dataAccess.insertInto("citizens")
.value("name")
.returning()
.toField<>( to )
result
.onSuccess { id -> println() }
.onFailure { error ->
(error) {
ConstraintViolationException -> println()
DataOperationException -> println()
TransactionException -> println()
-> println()
}
}
See Error Handling for the full exception hierarchy and debugging tips.
Create a database.properties file in src/main/resources:
db.url=jdbc:postgresql://localhost:5432/roma
db.username=augustus
db.password=spqr
db.schemas=public,cursus_honorum
db.packagesToScan=com.roma.domain,com.roma.dto
# Custom HikariCP settings
db.hikari.maximumPoolSize=20
db.hikari.minimumIdle=5
# Optional settings
db.setSearchPath=true
db.dynamicDtoStrategy=AUTOMATIC_WHEN_UNAMBIGUOUS
db.disableCoreTypeInitialization=false
Load it in your application:
// From properties file
val dataAccess = OctaviusDatabase.fromConfig(
DatabaseConfig.loadFromFile("database.properties")
)
val dataAccess = OctaviusDatabase.fromConfig(
DatabaseConfig(
dbUrl = "jdbc:postgresql://localhost:5432/roma",
dbUsername = "augustus",
dbPassword = "spqr",
dbSchemas = listOf("public"),
packagesToScan = listOf("com.roma.domain"),
hikariProperties = mapOf("maximumPoolSize" to "20")
)
)
// From existing DataSource
val dataAccess = OctaviusDatabase.fromDataSource(existingDataSource, ...)
Octavius provides an optional integration with Flyway for schema migrations via the :flyway-integration module.
val dataAccess = OctaviusDatabase.fromConfig(
config = config,
migrationRunner = FlywayMigrationRunner.create(
schemas = config.dbSchemas,
baselineVersion = "1"
)
)
See Flyway Migrations in the configuration guide for details.
For detailed guides and examples, see the full documentation:
| Object is a Vessel | A data class is simply a type-safe container for query results. |
| Explicitness over Magic | No lazy-loading, no session management, no dirty checking. |
COMPOSITE, ENUM, ARRAY and Custom Type Handlers (Global & Per-Query) ↔ Kotlin typesdynamic_dto and dynamic_mapWHERE clauses with QueryFragment// SELECT with pagination
val senators = dataAccess.select("id", "name", "province")
.from("senate")
.where("active = true")
.orderBy("appointed_at DESC")
.limit(10)
.offset(20)
.toListOf<Senator>()
// INSERT with RETURNING
val newId = dataAccess.insertInto("citizens")
.value("name")
.value("tribe")
.returning("id")
.toField<Int>(mapOf("name" to "Marcus Aurelius", "tribe" to "Cornelia"))
// UPDATE with expressions
dataAccess.update("legion_supplies")
.setExpression("quantity", "quantity - 1")
.where("id = @id")
.execute("id" to supplyId)
// DELETE
dataAccess.deleteFrom("expired_mandates")
.where("expires_at < NOW()")
.execute()
| PostgreSQL | Kotlin | Notes |
|---|
int2, smallserial | Short | |
int4, serial | Int | |
int8, bigserial | Long | |
float4 | Float | |
float8 | Double | |
numeric | BigDecimal | |
text, varchar, char | String | |
bool | Boolean | |
uuid | Uuid | kotlin.uuid.Uuid |
bytea | ByteArray | |
json, jsonb | JsonElement | kotlinx.serialization.json |
void | Unit | Return type of void functions (e.g. pg_notify) |
date | LocalDate | kotlinx.datetime * |
time | LocalTime | kotlinx.datetime |
timestamp | LocalDateTime | kotlinx.datetime * |
timestamptz | Instant | kotlin.time * |
interval | Duration | kotlin.time * |
// PostgreSQL COMPOSITE TYPE → Kotlin data class
data class Province(val name: String, val capital: String, val governor: String)
// PostgreSQL ENUM → Kotlin enum
enum class Magistrature { Quaestor, Aedile, Praetor, Consul, Censor }
// Works seamlessly in queries
data class Senator(val id: Int, val rank: Magistrature, val homeProvince: Province)
val senators = dataAccess.select("id", "rank", "home_province")
.from("senate")
.toListOf<Senator>() // Types converted automatically
// 1. Define a sealed interface
sealed interface MonumentRecord
data class Inscription(val text: String, val lang: String) : MonumentRecord
data class Relief(val subject: String) : MonumentRecord
// Database: CREATE TABLE monument_records (id INT, record dynamic_dto);
// 2. Fetch directly to a list of your interface
val records = dataAccess.select("record")
.from("monument_records")
.toColumn<MonumentRecord>()
// Returns: [Inscription(...), Relief(...), Inscription(...)]
DataResult.Failure(error) instead of throwing. This forces explicit handling of expected database errors like constraint violations, lock timeouts, or missing records.WHERE clause in a DELETE, or a Kotlin type mapping mismatch), Octavius throws a standard exception (FatalDatabaseException). It fails fast because these errors represent broken code that should be caught and fixed during development.QueryContext that provides a clean visualization of the high-level SQL, the low-level JDBC query, and the exact parameters involved (great for logging!)..options() and builder modes| Module | Platform | Description |
|---|
api | Multiplatform | Common: Annotations & DTOs (JVM/JS). JVM-only: Query & Transaction interfaces. |
core | JVM | Zero-dependency core engine. Pure JDBC & HikariCP. |
spring-integration | JVM | Optional integration for Spring Boot (@Transactional support). |
flyway-integration | JVM | Optional migration runner integration. |
Surfaced from shared tags and platforms — no rankings paid for.