Skip to main content

Mutable mappings persisted in the filesystem, encrypted if you want.

Project description

File system mappings

This started as "copying the code" from the talk Build powerful, new data structures with Python's abstract base classes by Raymond Hettinger since I couldn't find it anywhere.

After that I started modifying the code and creating new utilities.

Install with:

pip install fsdict

Dictionary implementation that saves everything to files in disk.


When a value is going it be written to the disk a new salt is created. Then this salt is passed to the hashing function. The values are encrypted using PBKDF2HMAC. The salt is generated for each entry. It is saved in the filename separated by a quadruple underscore. Don't use quadruple underscore in your variable names when setting/getting the dictionary, it will break, you can use double and triple underscores. It can be fixed adding a couple of if statements inside, but so far I have never had to create a key/value name with 4*_ inside.

The salt is 64 bytes long and it's created using secrets.token_urlsafe. The salt is created as a string and then encoded to bytes so that it's easier to implement a static salt that you can keep as an environment variable.


It is a single self-contained 200 lines-of-code file. The only dependency is cryptography. Install it with:

pip install cryptography

If you don't want to add another dependecy to your project, just copy and pasting the contents of the file to your project. When copying the file you may extend it adding more functionality to it. That functionality may be useful to other users so please consider openning a pull request to add it to the current project.


You need to provide a password when creating the dictionary object or set an environment variable:

Creating an environment variable (recommended):

export PASS=mypass

When creating the dictionary:

from fsdict.flightcryptofiledict import FileDict

d = FileDict("newname", password="secretpassword")

Using different data types

If the dictionary values need to be something different from strings a custom encoder/decoder can be passed. The only condition is that those functions should return a bytes object. Examples:

>>> from fsdict.flightcryptofiledict import FileDict
>>> enc_filedict = FileDict("cryptest", password="password")

>>> import pickle
>>> import math

>>> enc_filedict.encoder = lambda x: pickle.dumps(x)
>>> enc_filedict.decoder = lambda x: pickle.loads(x)

>>> enc_filedict["myfunc"] = math.cos
>>> enc_filedict["myfunc"]
<function math.cos(x, /)>

>>> enc_filedict["myfunc"](23)

>>> import json
>>> enc_filedict.encoder = lambda x: json.dumps(x).encode()
>>> enc_filedict.decoder = lambda x: json.loads(x.decode())

>>> enc_filedict["mydict"] = {"a": 12}
>>> enc_filedict["mydict"]
{'a': 12}

>>> enc_filedict["mydict"]["a"]

Compressing to file

Create a tarfile with the same name as the dictionary.


You can use a different filename. The **kwargs passed to the function will be passed to the tarfile function.

enc_filedict.compress(filename="compressed", compresslevel=7)

The command above will generate the file: compressed.tar.gz

More or less the same as before but the salt is static. The salt is parsed as a base64 encoded string. It will be less secure but faster. You can pass the salt as a string when creating a new dictionary. If not, it will try to get it from the environment variables.

Uses an SQLite database instead of the filesystem. This module doest NOT implement the encoding/decoding capabilities. Right now it works as a key-value storage for string-like objects. This is planned for the 0.6.0 release.

Same as, but uses an sqlite database instead of the file system. It only needs a password and generates a different salt for each item. The salt is stored in a table column and the data in another one.

Notes for SQLite

Bith sqldict and flightcryptosqldict include an option called fast. By default it's set to True. This makes SQLite use WAL mode plus a few other optimizations to increase performance.


The performance method for each dictionary is calculated like this (adapted to each case):

In [1]: from fsdict.cryptosqldict import SQLDict as d

In [2]: dd = d("perf_test", password="mypass")

In [3]: from string import ascii_lowercase as letters

In [4]: from random import choice

In [5]: def randstr(n):
   ...:     return "".join([choice(letters) for _ in range(n)])

In [6]: %%timeit
   ...: dd[randstr(10)] = randstr(100)

The SQLite results were done without the speedup optimizations mentioned in the notes above.

RESULTS: 67.4 ms ± 2.6 ms per loop (mean ± std. dev. of 7 runs, 1 loop each) 1.26 ms ± 261 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) 578 µs ± 40.3 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) 809 µs ± 35.2 µs per loop (mean ± std. dev. of 7 runs, 1000 loops each) 68.9 ms ± 1.87 ms per loop (mean ± std. dev. of 7 runs, 1 loop each)


Ricardo Ander-Egg Aguilar – @ricardoanderegg

Distributed under the MIT license. See LICENSE for more information.


  1. Fork it (
  2. Create your feature branch (git checkout -b feature/fooBar)
  3. Commit your changes (git commit -am 'Add some fooBar')
  4. Push to the branch (git push origin feature/fooBar)
  5. Create a new Pull Request


  • 0.6
    • Make fsdict.sqldict store values as blobs instead of text
    • Add doctests
    • Add fsdict.cryptosqldict
    • Better PASS/SALT handling when passing environment variables
  • 0.5.3
    • Add check_same_thread option to class initialization
    • Better naming for the modules
    • Update README with missing information

Project details

Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

fsdict-0.8.1.tar.gz (12.0 kB view hashes)

Uploaded Source

Built Distribution

fsdict-0.8.1-py3-none-any.whl (14.4 kB view hashes)

Uploaded Python 3

Supported by

AWS AWS Cloud computing and Security Sponsor Datadog Datadog Monitoring Fastly Fastly CDN Google Google Download Analytics Microsoft Microsoft PSF Sponsor Pingdom Pingdom Monitoring Sentry Sentry Error logging StatusPage StatusPage Status page