OctaviusClient

Hands out queries and sessions, and runs transactions over them. That is all it is.

A query taken from here - rawQuery, select, insertInto, update, deleteFrom - is a RunnableQuery, which carries the terminal family and finds its own session when one of them is called. Nothing has to be opened around it, which is what keeps a single query to a single expression.

Where the work is not a query - copy, largeObjects, notifications, or several statements that have to share one session - execute hands over the driver's own OctaviusSessionOperations and gets out of the way. Neither path wraps or renames anything the driver named.

val db = OctaviusClient.fromDataSource(hikariDataSource)

val senators = db.rawQuery("SELECT id, cognomen FROM senators WHERE province_id = @p")
    .fetchObjects<Senator>("p" to 7)

What it buys over calling dataSource.getOctaviusSession() by hand is the thing a data layer of any size ends up needing: a repository function can open a scope without knowing whether it is already inside a transaction, and be right either way. A session opened here joins the transaction running on this thread if there is one, so the same function works standalone and as a step in a larger unit of work, without a session in its signature.

Failures arrive as exceptions, which is what the driver raises and what a try/catch expects. Where a failure should be a value instead, the result style has a door for each width: asResult for one query, transactionResult for a transaction, and dbResult for anything else.

Inheritors