A parser/packer combinator library for .net, where "Jars" (a serializer paired with a deserializer) are created by combining and augment primitive jars into complex jars. The resulting combinators are optimized and compiled at runtime.
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.



PickleJar is a library for describing simple binary formats, and using that description to parse and pack values.

You parse and pack values (a.k.a. 'pickling') using a 'Jar'. PickleJar includes basic jars, like Jar.Int32LittleEndian and Jar.Utf8, as well as methods to augment and combine jars, like Jar.RepeatNTimes and Jar.NullTerminated. Code for the combined and augmented jars is generated, optimized, and compiled at runtime (in cases where this has been implemented).

See this blog post for more discussion.

============ Current Work

I am currently working on:

  • Adding more optimization cases (e.g. inlining bits of RepeatNTimes, packing by blitting).
  • Adding support for building anonymous type parsers, or using a dictionary instead of a type.
  • Testing error cases.

============ Installation

NuGet Package

===== Usage

All functionality for making jars is present on the static Strilanc.PickleJar.Jar class. The quickest way to see what's available is to "dot around" Jar with intellisense, or browse through the NuDoq documentation.

Example #1: Pickling a string

// if the string is UTF8-encoded and null-terminated:
var stringJar1 = Jar.Utf8.NullTerminated();

// if the string is ASCII-encoded and prefixed by the byte length of the encoded characters:
var stringJar2 = Jar.Ascii.DataSizePrefixed(Jar.Int32LittleEndian, includePrefixInSize: false);

// packing:
// returns new byte[] { 116,101,115,116,226,156,147,0 }
byte[] data = stringJar1.Pack("test✓");

// parsing:
// returns value="aca", consumed=7
ParsedValue<string> = stringJar2.Parse(new byte[] {3,0,0,0, 97,99,97,255,255,255});

Example #2: Parsing a custom type

Parsing two contiguous floats into a point class.

// The type (the library is smart enough to realize it must pass X and Y to the constructor instead of setting them):
public sealed class Point {
    public readonly float X;
    public readonly float Y;
    public Point(float x, float y) {
        X = x;
        Y = y;

// The jar, keying the component jars by name and then building the custom type jar:
var pointJar = new Jar.Builder {
    {"x", Jar.Float32LittleEndian},
    {"y", Jar.Float32LittleEndian}

// parsing:
// returns value={X=0.0f, Y=1.0f}, consumed=8
var p = pointJar.Parse(new byte[]{0,0,0,0, 0,0,128,63});

Example #3: Using StructLayoutAttribute to allow memcpy optimizations

// (the layout attribute forces a particular memory representation, which the library will notice and exploit)
[StructLayoutAttribute(LayoutKind.Sequential, Pack = 1)]
public struct Point3 {
    public int X;
    public int Y;
    public int Z;

var bulkPointParser = new Jar.Builder {
    // notice that the fields come in the same order as they are declared
    {"x", Jar.Int32LittleEndian},
    {"y", Jar.Int32LittleEndian},
    {"z", Jar.Int32LittleEndian}

// because an array of Point3 has the same representation as the serialized data, a memcpy is  valid parser
// the library notices this, and makes an unsafe parser that beats the pants off safe hand-written C#

var singlePointData = new byte[] {1,0,0,0,2,0,0,0,3,0,0,0};
var tenMillion = 10000000;
var bulkPointData = Enumerable.Repeat(singlePointData, tenMillion).SelectMany(e => e).ToArray();
var points = bulkPointParser.Parse(bulkPointData).Value; // parses at about 1GB/s