How is lmdb-store different from node-lmdb?

[detj]

Although, there is a lot of valuable information in the README. It was not very clear to me how lmdb-store is adding value compared to node-lmdb.

Does adding a summary list of features make sense?

Features

• feature 1 bla bla bla
• feature 2 bla bla bla
• and so on

As a user of lmdb-store, how much understanding of node-lmdb is required? Is MessagePack added by lmdb-store?

These are the few questions I'm wondering.

[kriszyp]

The most simplified distinction would be that node-lmdb is a low-level interface designed to directly expose the LMDB API functions, whereas lmdb-store is designed to provide a higher level interface for more efficient and idiomatic JS interaction. However, lmdb-store isn't really built "on top off" node-lmdb, since many/most of the improvements that lmdb-store makes actually involve extending/modifying the C++ code, for example adding the ability to coordinate batches of operations and commit (and compress) them in a separate thread, and perform multi-threaded coordination of asynchronous transactions, which is mostly done in C++. lmdb-store is written as a "branch" or fork of node-lmdb, building from the node-lmdb C++ code.

Basically the whole set of features listed at the beginning of the readme are features in lmdb-store that aren't available in node-lmdb, are lmdb-store further optimizes:

• High-performance translation of JS values and data structures to/from binary key/value data: lmdb-store provides a mechanism for deserialization that reuses buffers directly with the MessagePack deserializer and is significantly faster than using the low-level methods and passing buffers around to a deserializer.
• Queueing asynchronous off-thread write operations with promise-based API
• Simple transaction management: Automated handling of creating and managing read transactions. Also asynchronous transactions are only available in lmdb-store.
• Iterable queries/cursors
• Automated database growth
• Record versioning and optimistic locking for scalability/concurrency
• Optional native off-main-thread compression with high-performance LZ4 compression

I also help with maintenance of node-lmdb, but since many of the features needed to really accelerate modern applications require the higher level integrations of lmdb-store, lmdb-store is just getting a lot more dev now.

Using lmdb-store shouldn't require any knowledge of node-lmdb. It may be of interest to familiarize yourself with LMDB itself and the LMDB docs provide more information on certain settings/options if you get more into optimizing, but lmdb-store docs are designed to be pretty standalone.

And yes, lmdb-store adds integration of MessagePack (again, by internalizing the buffer reuse, it is much faster than returning buffers and having a separate step of deserializing using JSON.parse or anything else).

Anyway, hope that helps, let me know if you have any other questions about the differences.

[detj]

Understood. Thanks for the detailed & fantastic answer.