ProtoDSL

A Kotlin Multiplatform, code-first, reflection-free Protobuf DSL. Define messages by declaring PbSchema singletons with PbType fields, then serialize/deserialize without reflection or codegen.

The framework is previously used in Acidify.

The goal is a compact, readable DSL that favors expressiveness over maximal performance.

Basic usage

Define a message model:

object TestMessage : PbSchema() {
    val int32Field = PbInt32[1]
    val boolField = PbBoolean[2]
    val optionalField = PbOptional[PbInt32[3]]
}

Construct, serialize, and deserialize:

val message = TestMessage {
    it[int32Field] = 1
    it[boolField] = true
    it[optionalField] = null
}

val encoded = message.toByteArray()
val decoded = TestMessage(encoded)

val decodedInt32Field = decoded.get { int32Field }
// Equivalent to: decoded[TestMessage.int32Field]

Message construction and field access

• Use object : PbSchema to define models and avoid reflection limits on non-JVM targets.
• Call TestMessage { ... } (via operator fun invoke) to create a PbObject<TestMessage>.
• Use it[field] = value to set fields; use get { field } to read fields.

Field types

PbType handles conversion between values and low-level DataTokens. Common types:

• Scalars: PbInt32, PbInt64, PbBoolean, PbFixed32, PbFixed64
• String/bytes: PbString, PbBytes
• Optional: PbOptional[PbType]
• Repeated: PbRepeatedInt32, PbRepeatedInt64, PbRepeatedString, PbRepeatedBytes
• Nested messages: PbMessage(fieldNumber, schema), PbRepeatedMessage(fieldNumber, schema)

Example:

object Nested : PbSchema() {
    val id = PbInt32[1]
}

object Parent : PbSchema() {
    val nested = PbMessage(1, Nested)
    val repeatedNested = PbRepeatedMessage(2, Nested)
}

Performance note

This project does not aim for extreme performance. The focus is on DSL ergonomics and cross-platform usability; the overhead is acceptable for typical use cases.

See also

• Related article (Chinese) that explains the design and implementation of this project