Skip to main content

Serialization based on ast.literal_eval

Project description

Serpent is a simple serialization library based on ast.literal_eval.

Because it only serializes literals and recreates the objects using ast.literal_eval(), the serialized data is safe to transport to other machines (over the network for instance) and de-serialize it there.

There is also a Java and a .NET (C#) implementation available. This allows for easy data transfer between the various ecosystems. You can get the full source distribution, a Java .jar file, and a .NET assembly dll. The java library can be obtained from Maven central (groupid net.razorvine artifactid serpent), and the .NET assembly can be obtained from (package Razorvine.Serpent).


  • ser_bytes = serpent.dumps(obj, indent=False, set_literals=True, module_in_classname=False): # serialize obj tree to bytes
  • obj = serpent.loads(ser_bytes) # deserialize bytes back into object tree
  • You can use ast.literal_eval yourself to deserialize, but serpent.deserialize works around a few corner cases. See source for details.

Serpent is more sophisticated than a simple repr() + literal_eval():

  • it serializes directly to bytes (utf-8 encoded), instead of a string, so it can immediately be saved to a file or sent over a socket
  • it encodes byte-types as base-64 instead of inefficient escaping notation that repr would use (this does mean you have to base-64 decode these strings manually on the receiving side to get your bytes back. You can use the serpent.tobytes utility function for this.)
  • it contains a few custom serializers for several additional Python types such as uuid, datetime, array and decimal
  • it tries to serialize unrecognised types as a dict (you can control this with __getstate__ on your own types)
  • it can create a pretty-printed (indented) output for readability purposes
  • it outputs the keys of sets and dicts in alphabetical order (when pretty-printing)
  • it works around a few quirks of ast.literal_eval() on the various Python implementations

Serpent allows comments in the serialized data (because it is just Python source code). Serpent can’t serialize object graphs (when an object refers to itself); it will then crash with a ValueError pointing out the problem.

Works with Python 2.7+ (including 3.x), IronPython 2.7+, Jython 2.7+.


  • Why not use XML? Answer: because XML.
  • Why not use JSON? Answer: because JSON is quite limited in the number of datatypes it supports, and you can’t use comments in a JSON file.
  • Why not use pickle? Answer: because pickle has security problems.
  • Why not use repr()/ast.literal_eval()? See above; serpent is a superset of this and provides more convenience. Serpent provides automatic serialization mappings for types other than the builtin primitive types. repr() can’t serialize these to literals that ast.literal_eval() understands.
  • Why not a binary format? Answer: because binary isn’t readable by humans.
  • But I don’t care about readability. Answer: doesn’t matter, ast.literal_eval() wants a literal string, so that is what we produce.
  • But I want better performance. Answer: ok, maybe you shouldn’t use serpent in this case. Find an efficient binary protocol (protobuf?)
  • Why only Python, Java and C#/.NET, but no bindings for insert-favorite-language-here? Answer: I don’t speak that language. Maybe you could port serpent yourself?
  • Where is the source? It’s on Github:
  • Can I use it everywhere? Sure, as long as you keep the copyright and disclaimer somewhere. See the LICENSE file.


# This demo script is written for Python 3.2+
# -*- coding: utf-8 -*-
from __future__ import print_function
import ast
import uuid
import datetime
import pprint
import serpent

class DemoClass:
    def __init__(self):

data = {
    "names": ["Harry", "Sally", "Peter"],
    "big": 2**200,
    "colorset": { "red", "green" },
    "id": uuid.uuid4(),
    "class": DemoClass(),
    "unicode": "€"

# serialize it
ser = serpent.dumps(data, indent=True)
open("data.serpent", "wb").write(ser)

print("Serialized form:")

# read it back
data = serpent.load(open("data.serpent", "rb"))

# you can also use ast.literal_eval if you like
ser_string = open("data.serpent", "r", encoding="utf-8").read()
data2 = ast.literal_eval(ser_string)

assert data2==data

When you run this (with python 3.2+) it prints:

Serialized form:
# serpent utf-8 python3.2
  'big': 1606938044258990275541962092341162602522202993782792835301376,
  'class': {
    '__class__': 'DemoClass',
    'b': False,
    'i': 42
  'colorset': {
  'id': 'e461378a-201d-4844-8119-7c1570d9d186',
  'names': [
  'timestamp': '2013-04-02T00:23:00.924000',
  'unicode': '€'
{'big': 1606938044258990275541962092341162602522202993782792835301376,
 'class': {'__class__': 'DemoClass', 'b': False, 'i': 42},
 'colorset': {'green', 'red'},
 'id': 'e461378a-201d-4844-8119-7c1570d9d186',
 'names': ['Harry', 'Sally', 'Peter'],
 'timestamp': '2013-04-02T00:23:00.924000',
 'unicode': '€'}

Project details

Download files

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

Filename, size & hash SHA256 hash help File type Python version Upload date
serpent-1.27-py2.py3-none-any.whl (11.4 kB) Copy SHA256 hash SHA256 Wheel py2.py3 Jul 30, 2018
serpent-1.27.tar.gz (12.8 kB) Copy SHA256 hash SHA256 Source None Jul 30, 2018

Supported by

Elastic Elastic Search Pingdom Pingdom Monitoring Google Google BigQuery Sentry Sentry Error logging AWS AWS Cloud computing DataDog DataDog Monitoring Fastly Fastly CDN DigiCert DigiCert EV certificate StatusPage StatusPage Status page