Awesome Open Source
Awesome Open Source

Colfer Build Status

Colfer is a binary serialization format optimized for speed and size.

The project's compiler colf(1) generates source code from schema definitions to marshal and unmarshall data structures.

This is free and unencumbered software released into the public domain. The format is inspired by Protocol Buffers.

Language Support

  • C, ISO/IEC 9899:2011 compliant a.k.a. C11, C++ compatible
  • Go, a.k.a. golang
  • Java, Android compatible
  • JavaScript, a.k.a. ECMAScript, NodeJS compatible


  • Simple and straightforward in use
  • No dependencies other than the core library
  • Both faster and smaller than the competition
  • Robust against malicious input
  • Maximum of 127 fields per data structure
  • No support for enumerations
  • Framed; suitable for concatenation/streaming


  • Rust and Python support
  • Protocol revision


Download a prebuilt compiler or run go get -u to make one yourself. Homebrew users can also brew install colfer.

The command prints its own manual when invoked without arguments.

	colf — compile Colfer schemas

	colf [-h]
	colf [-vf] [-b directory] [-p package] \
		[-s expression] [-l expression] C [file ...]
	colf [-vf] [-b directory] [-p package] [-t files] \
		[-s expression] [-l expression] Go [file ...]
	colf [-vf] [-b directory] [-p package] [-t files] \
		[-x class] [-i interfaces] [-c file] \
		[-s expression] [-l expression] Java [file ...]
	colf [-vf] [-b directory] [-p package] \
		[-s expression] [-l expression] JavaScript [file ...]

	The output is source code for either C, Go, Java or JavaScript.

	For each operand that names a file of a type other than
	directory, colf reads the content as schema input. For each
	named directory, colf reads all files with a .colf extension
	within that directory. If no operands are given, the contents of
	the current directory are used.

	A package definition may be spread over several schema files.
	The directory hierarchy of the input is not relevant to the
	generated code.

  -b directory
    	Use a base directory for the generated code. (default ".")
  -c file
    	Insert a code snippet from a file.
  -f	Normalize the format of all schema input on the fly.
  -h	Prints the manual to standard error.
  -i interfaces
    	Make all generated classes implement one or more interfaces.
    	Use commas as a list separator.
  -l expression
    	Set the default upper limit for the number of elements in a
    	list. The expression is applied to the target language under
    	the name ColferListMax. (default "64 * 1024")
  -p package
    	Compile to a package prefix.
  -s expression
    	Set the default upper limit for serial byte sizes. The
    	expression is applied to the target language under the name
    	ColferSizeMax. (default "16 * 1024 * 1024")
  -t files
    	Supply custom tags with one or more files. Use commas as a list
    	separator. See the TAGS section for details.
  -v	Enable verbose reporting to standard error.
  -x class
    	Make all generated classes extend a super class.

	Tags, a.k.a. annotations, are source code additions for structs
	and/or fields. Input for the compiler can be specified with the
	-f option. The data format is line-oriented.

		<line> :≡ <qual> <space> <code> ;
		<qual> :≡ <package> '.' <dest> ;
		<dest> :≡ <struct> | <struct> '.' <field> ;

	Lines starting with a '#' are ignored (as comments). Java output
	can take multiple tag lines for the same struct or field. Each
	code line is applied in order of appearance.

	The command exits 0 on success, 1 on error and 2 when invoked
	without arguments.

	Compile ./io.colf with compact limits as C:

		colf -b src -s 2048 -l 96 C io.colf

	Compile ./*.colf with a common parent as Java:

		colf -p com.example.model -x Java

	Report bugs at <>.

	Text validation is not part of the marshalling and unmarshalling
	process. C and Go just pass any malformed UTF-8 characters. Java
	and JavaScript replace unmappable content with the '?' character
	(ASCII 63).

	protoc(1), flatc(1)

It is recommended to commit the generated source code into the respective version control to preserve build consistency and minimise the need for compiler installations. Alternatively, you may use the Maven plugin.



Data structures are defined in .colf files. The format is quite self-explanatory.

// Package demo offers a demonstration.
// These comment lines will end up in the generated code.
package demo

// Course is the grounds where the game of golf is played.
type course struct {
	ID    uint64
	name  text
	holes []hole
	image binary
	tags  []text

type hole struct {
	// Lat is the latitude of the cup.
	lat float64
	// Lon is the longitude of the cup.
	lon float64
	// Par is the difficulty index.
	par uint8
	// Water marks the presence of water.
	water bool
	// Sand marks the presence of sand.
	sand bool

See what the generated code looks like in C, Go, Java or JavaScript.

The following table shows how Colfer data types are applied per language.

Colfer C Go Java JavaScript
bool char bool boolean Boolean
uint8 uint8_t uint8 byte † Number
uint16 uint16_t uint16 short † Number
uint32 uint32_t uint32 int † Number
uint64 uint64_t uint64 long † Number ‡
int32 int32_t int32 int Number
int64 int64_t int64 long Number ‡
float32 float float32 float Number
float64 double float64 double Number
timestamp timespec time.Time †† time.Instant Date + Number
text const char* + size_t string String String
binary uint8_t* + size_t []byte byte[] Uint8Array
list * + size_t slice array Array
  • † signed representation of unsigned data, i.e. may overflow to negative.
  • ‡ range limited to [1 - 2⁵³, 2⁵³ - 1]
  • †† timezone not preserved

Lists may contain floating points, text, binaries or data structures.


Colfer is suited for untrusted data sources such as network I/O or bulk streams. Marshalling and unmarshalling comes with built-in size protection to ensure predictable memory consumption. The format prevents memory bombs by design.

The marshaller may not produce malformed output, regardless of the data input. In no event may the unmarshaller read outside the boundaries of a serial. Fuzz testing did not reveal any volnurabilities yet. Computing power is welcome.


Name changes do not affect the serialization format. Deprecated fields should be renamed to clearly discourage their use. For backwards compatibility new fields must be added to the end of colfer structs. Thus the number of fields can be seen as the schema version.


Colfer aims to be the fastest and the smallest format without compromising on reliability. See the benchmark wiki for a comparison. Suboptimal performance is treated like a bug.

Get A Weekly Email With Trending Projects For These Topics
No Spam. Unsubscribe easily at any time.
Javascript (1,118,155
Java (403,872
C (189,870
Golang (167,817
Go (167,817
Format (16,225
Code Generation (1,629
Code Generator (1,464
Binary Data (726
Serialization Format (18
Serializable Objects (9