eprotoc

Erlang proto file parser and code generator.

goals

The project has several goals:

1. Avoid record syntax

• Protocol buffer messages are expected to be used by potentially many modules
• Records aren't easily shared across modules

2. Preserve the flexibility of protocol buffers

• Protocol buffers are meant to be backwards compatibile and the underlying data structure should not break this
• We are using keylists for now but this may change as benchmarks are performed (orddicts are also an option).

3. Avoid header files

• Header files make it harder to reason about code and break encapsulation intrinsic to protocol buffers
• Prefer instead module-based organization of messages and enums

In addition, the project introduces no new dependencies (everything is raw erlang).

usage

To build the project, you need rebar in your $PATH, after which you can just run make or rebar compile.

The project ships with a simple convenience script for generating code from .proto files. To do so run:

eprotoc /path/to/file.proto /path/to/src/dir

This will generate a number of modules to the path you specify. For example:

package Foo;

message Bar {
    required uint32 zap = 1;
}

will output the module foo__bar.erl with the functions

fun g_zap/1,
fun s_zap/2,
fun encode/1,
fun decode/1

The first two functions are a getter and setter for the zap field and the last two functions are an encoder and decoder for a bar message list. To create a new message of type Bar, we would just use an empty list since lists are the internal structure of the message:

NewBar = foo__bar:s_zap([], 100),
foo__bar:encode(NewBar).

This creates an iolist representing a Bar message with zap set to 100 that can then be sent over the wire.

See the test folder for examples of enum usage, the repeated field, nested messages, imported messages, and more.

known caveats

• missing tests on a number of data types and cases of more extreme nesting

roadmap

1. Add more tests