ViperDB is a lightweight embedded key-value store written in pure Python. It has been designed for being extremely simple while efficient.

• tiny: the main db file consists of just ~300 lines of code.
• highly coverage: thanks to the small codebase, every single line of code is tested.
• log-structured: ViperDB takes design concepts by log-structured databases such as Bitcask.
• written in pure Python: no external dependency needed.

foo@bar:~$ pip3 install viperdb

ViperDB has been tested with Python 3.8.

ViperDB simply consists of two files: a key log file and a value log file. The first is used to maintain information about values (e.g. offset, size, etc...) which are actually stored in the value log. This layout allows to speed-up db initialization, which consists in loading the pointers to the entire key-space from the key-file to a dictionary. For simplicity, the key file is treated as a text file, with each line containing a json-encoded entry. The value file is viewed as a raw sequence of bytes. Before being written to the value file, each value is encoded according to the following scheme: builtin types (except for the bytes type) are json-encoded, while user-defined classes are pickled.

To keep logic simple, no automatic compaction is performed in the background: unused space must be reclaimed explicitly through the reclaim function.

from viperdb import ViperDB

db = ViperDB('./db')
# db can be used as a simple dictionary
db['hello'] = 'ViperDB!'
assert db['hello'] == 'ViperDB'

del db['hello']
assert 'hello' not in db

db.reclaim() # call this method periodically to free unused space.
db.close() # flush any pending write and close the database.

ViperDB is a very recent project. However, it is actively maintained to ensure high quality. If you find any bug, or have some suggestion, feel free to contribute by opening a new issue or making a pull request.