The latest version of Libmdbx, version 0.13.6 (MDBX), has been released. Libmdbx offers a high-performance compactly built-in data database with key-meaning. The code is distributed under the Apache 2.0 license and supports all relevant operating systems and architectures, including Russian Elbrus 2000. Libmdbx provides a developed API for C++, with bindings supported by enthusiasts for Rust, Haskell, Python, Nodejs, Ruby, GO, NIM, Deno, and Scala. Projects such as isar, erigon, retah, as well as companies like starkware and positive technologies are known to use libmdbx.
Historically, Libmdbx is a deep processing DBMDB DBMS that surpasses its ancestor in reliability, capabilities, and performance. It focuses on code quality, API stability, testing, and automatic checks. Libmdbx offers ACID, strict serialization of changes, non-relaxing reading with linear scaling along CPU cores, car sequence, automatic database size control, and Range Query Estimation.
The main changes in this release include:
- Change in behavior for mdbx_cursor_bind () function.
- Avoidance of potential issues by not using Copy_File_RANGE() on Linux 5.3 – 5.18.
- STD stuffing :: Invalid_argument now provides a clear message mdbx_einval.
- Removal of remember usage in reading transactions for DBI = 0 (AKA GC/Freedb).
- Elimination of Eagain error regression when copying the database on NFS and CIFS/SMB.
- Addition of file blocking capture in the API copy function.
- Fixed merge/rebase error in mdbx_txn_release_all_cursors_ex().
- Fixed Assert Care in transaction completion.
- Possibility of returning unexpected error mdbx_busy from mdbx_txn_lock (Dont_wait = false).
- Change in function attribute indication for compatibility with GCC 15.x in C23 mode.