v0.3.0

This release adds the returning of unused DB objects in mdb.Add() and mdb.Delete(). This can help downstream callers to cleanup any unused connections, by closing them.

Refactor

A major refactor to simplify the package. Simpler code translates in higher performance!

1. Use the new Row.Err() method to be error aware in QueryRow variants.
2. Benchmark and optimize the core multiplexer functions to run with less allocation and overhead. Improvement of 5-8x on bare runs.
3. Remove re-connection logic, as it was creating additional locking and overhead. The core SQL library already replaces errored connections in the pool.
4. Ditch the Node and Tx type. Instead use sql.DB and sql.Tx types where possible. This reduces the amount of duplicated code to maintain.
5. An effort to reduce the amount of (mutex) locks.

Due to core type changes, this release is not backward-compatible. This package is still considered to be in beta.

Rc2

This upgrades dependencies.

Rc1

• Use errors.Is() for consistent error checking
• Exclude context cancelled error from disconnect statistics
• Fix: (#7) properly close unused Rows and wait for all query to finish in a Multi transaction

Beta release

We have tested the use of this library in our projects and feel comfortable saying that this project is now in Beta stage.

This release carries a single bugfix: #3.

First release

The intended design is fully implemented and unit tested with the race detector. Actual implementations using this package still need to prove its stability and usability.

Knows issues