zero serialization overhead


No time to read? Go to the Quick Start or download latest release.

In for the numbers? Head over to benchmarks.

Questions? Check the FAQs.

More documentation? Go to the Documentation.

Serialization framework

The zserio serialization framework allows you to serialize data in a compact and efficient way.

The key features include

It can be retrofitted on top of almost any other serialization language or model, since it gives the developer powerful low-level access.

It features simple and compound data structures and provides advanced features for controlling at design time what writers will be able to fill in.

Although it does not have a wire format, we have added some convenience keywords lately that encapsulate some functionality. You can find more information on Zserio Invisibles.

At the moment the following languages are supported

Quick introduction

As we have stated earlier, zserio does not have any wire format. So basically: what you see is what you get (please note that zserio uses network byte order in the serialized stream, but the generated code does take care of this).

package tutorial;

struct Employee
    uint8   age;
    string  name;
    uint16  salary;
    Role    role;

enum uint8 Role
    DEVELOPER = 0,
    TEAM_LEAD = 1,
    CTO       = 2,

So if we use the schema above and serialize one employee with

the resulting byte stream looks like this:

Offset(d) 00 01 02 03 04 05 06 07 08 09 10 11 12 13

00000000  20 09 4A 6F 65 20 53 6D 69 74 68 13 88 00
Byte position value value (hex) comment
0 32 (age) 20 uint8 is of fixed size 8 bit
1 9 (string length) 09 string length is encoded in varuint64 field before actual string
2-10 Joe Smith 4A 6F 65 20 53 6D 69 74 68 UTF-8 encoded string
11-12 5000 13 88 uint16 always uses 2 bytes
13 0 00 enum is of size uint8 so it uses 1 byte

Please note that in contrast to other serialization mechanisms zserio supports as well variable integers which do not provide the full range of values but rather stick to the indicated size. Example: a varuint64 will be using max 8 bytes whilst not providing the full range of a uint64_t but a varuint will be using max 9 bytes and providing the full range of a uint64_t.

Quick Start

To be able to serialize data with zserio, you have to follow these basic steps:

  1. Download the runtimes and the zserio compiler from Github Releases
  2. Set up your development environment with the zserio runtime
  3. Write the schema definition
  4. Compile the schema and generate code
  5. Serialize/deserialize using the generated code

You can find the detailed quick start tutorial in their respective repositories:

Features overview


Documentation of the schema language can be found in the Zserio Language Overview.

Explanation of more hidden schema language features can be found in the Zserio Invisibles.

Schema language reference can be found in Quick Reference.

User Guide can be found in the Zserio Compiler User Guide.

Build instructions can be found in the Zserio Compiler Build Instructions.

C++ users can find more information in the C++ Tutorial.

Java users can find more information in the Java Tutorial.

Python users can find more information in the Python Tutorial.

Check out as well the Zserio Types Mapping for types mapping description.


Service types allow to define generic service interfaces. But note that no underlying communication library is provided by Zserio. Zserio only defines the generic interface and users are responsible for its implementation. However, Zserio provides sample implementations of several services backends:


Pubsub types allow to define generic Pub/Sub clients. Users are responsible for implementation of the generic Pub/Sub client interface provided by Zserio. However, Zserio provides sample implementations of several Pub/Sub backends:

Note that Zserio doesn't provide any Pub/Sub server. There are various implementations of servers (e.g. mosquitto) and it's the responsibility of the user's Pub/Sub client implementation to communicate with the appropriate server.