Customizations

Getting started with scalapb.proto#

ScalaPB's code generator provides supports many different customizations. To get access to these customizations, you need to import scalapb/scala.proto in the proto files you want to customize. You can also have the options apply to an entire proto3 package by using package-scoped options (see below).

To have scalapb/scalapb.proto available to be imported in your project, add the following SBT setting in your build.sbt:

libraryDependencies += "com.thesamet.scalapb" %% "scalapb-runtime" % scalapb.compiler.Version.scalapbVersion % "protobuf"

If you are invoking protoc manually, you will need to ensure that the files in protobuf directory are available to your project.

ScalaPB File-level Options#

ScalaPB file-level options lets you

  • specify the name of the Scala package to use (the default is using the java package name).
  • request that ScalaPB will not append the protofile name to the package name.
  • specify Scala imports so custom base traits and custom types (see below) do not require the full class name.

The file-level options are not required, unless you are interested in those customizations. If you do not want to customize the defaults, you can safely skip this section.

File-level options#

import "scalapb/scalapb.proto";
option (scalapb.options) = {
scope: FILE
package_name: "com.example.myprotos"
flat_package: true
single_file: true
java_conversions: false
import: "com.thesamet.pb.MyType"
import: "com.thesamet.other._"
preamble: "sealed trait BaseMessage"
preamble: "sealed trait CommonMessage"
lenses: true
getters: true
retain_source_code_info: false
no_default_values_in_constructor: false
preserve_unknown_fields: false
enum_value_naming: CAMEL_CASE
enum_strip_prefix: false
bytes_type: "scodec.bits.ByteVector"
};
  • scope controls whether the specified options apply only for this proto files or for the entire package. Default is FILE. See package-scoped options for more details.

  • package_name sets the Scala base package name, if this is not defined, then it falls back to the java_package option. If the java_package option is also not the found, then the package name from file's package statement is used.

  • Setting flat_package to true (default is false) makes ScalaPB not append the protofile base name to the package name. You can also apply this option globally to all files by adding it to your ScalaPB SBT Settings.

  • The single_file option makes the generator output all messages and enums to a single Scala file.

  • The java_conversions options tells ScalaPB to generate converters to the corresponding Java messages in this file. It does not automatically trigger Java source code generation for the messages. If you need to generate source code in Java, include PB.gens.java in the list of targets in sbt-protoc.

  • The preamble is a list of strings that is output at the top of the generated Scala file. This option requires single_file to be set. It is commonly used to define sealed traits that are extended using (scalapb.message).extends - see custom base traits below and this example.

  • The object_name option lets you customize the name of the generated class that contains various file-level members such as descriptors and a list of companion objects for the generated messages and enums. This is useful in case you are running into issues where the generated class name conflicts with other things in your project.

  • Setting lenses to false inhibits generation of lenses (default is true).

  • Setting getters to false inhibits generation of getters (default is true).

  • Setting retain_source_code_info to true retains information in the descriptor that can be used to retrieve source code information from the descriptor at runtime (such as comments and source code locations). This option is turned off by default to conserve source size and memory at runtime. When this option is enabled, use the location method on various descriptors to access source code information.

  • By default, all non-required fields have default values in the constructor of the generated case classes. When setting no_default_values_in_constructor to true no default values will be generated for all fields.

  • Typically, enum values appear in UPPER_CASE in proto files, and ScalaPB generates case objects with exactly the same name in Scala. If you would like ScalaPB to transform the names into CamelCase, set enum_value_naming to CAMEL_CASE.

  • It is a common practice in protobufs to prefix each enum value name with the name of the enum. For example, an enum name Size may have values named SIZE_SMALL and SIZE_LARGE. When you set enum_strip_prefix to true, ScalaPB will strip the enum's name from each value name, and they would become SMALL and LARGE. Then the name can be transformed to camel-case according to enum_value_naming. Note that the prefix that is removed is the all-caps version of the enum name followed by an underscore.

  • By default, during deserialization only known fields are retained. When setting preserve_unknown_fields to true, all generated messages in this file will preserve unknown fields. This is default behaviour in java for Proto3 messages since 3.5.0. In ScalaPB 0.10.0: the default of this field became true for consistency with Java.

  • Use bytes_type to customize the Scala type used for the bytes field type. You will need to have an implicit TypeMapper[com.google.protobuf.ByteString, YourType] instance so ScalaPb can convert back and forth to the type of your choice. That implicit will be found if it is defined under YourType companion object, or on a package object that matches the generated code (or any of its parent packages).

Package-scoped options#

Note: this option is available in ScalaPB 0.8.2 and later.

Sometimes you want to have the same file-level options applied to all the proto files in your project. To accomplish that, add a package.proto file (the name does not matter) next to your proto files that looks like this:

import "scalapb/scalapb.proto";
package com.mypackage;
option (scalapb.options) = {
scope: PACKAGE
flat_package: true
};

All the options in this file will be applied to all proto files in the package com.mypackage and its sub-packages.

There is no need to explictly import this file from other protos. If you are using sbt-protoc and the file is in the proto source directory (default is src/main/protobuf) then the file will be found and the options applied. If you are invoking protoc in another way, you need to ensure that this file is passed to protoc together with the rest of the files.

If you are generating Scala code for proto files that you don't own, you can use this feature to customize code generation by creating a package.proto file for that third-party package and include it within your proto source directory.

The following rules are applied when validating package-scoped options:

  • At most one file in each package may provide package-scoped options.
  • Sub-packages may override package-scoped options provided by their parent packages. The options are merged using the Protocol Buffers mergeFrom semantics. Specifically, this implies that repeated fields such as import and preamble are concatenated.
  • Proto files get the most specific package-scoped options for the package they are in. File-level options defined in a proto file get merged with the package-level options using mergeFrom.
  • Proto files with package-scoped options must have a package statement. This is to prevent the possibility of options applied globally. Standard classes that are shipped with ScalaPB already assume certain options, so overriding options globally may lead to compilation errors.

Publishing package-scoped options#

If you are publishing a library that includes protos with package-scoped options, you need to make sure your library users source the package-scoped option proto file so the customizations are applied when they generate code.

Your users can simply import your package-scoped options from any proto file in their project to have the settings applied (a single import of the package-scoped options file would apply it globally for the code generator). However, since ScalaPB 0.10.11 and sbt-protoc 1.0.1, sbt-protoc provides a way to automate this with no need to manually import the package-scoped options file. This is accomplished by including a special attribute in the manifest of the library you publish. Add the following to your library's settings:

Compile / packageBin / packageOptions += (
Package.ManifestAttributes("ScalaPB-Options-Proto" -> "path/to/package.proto")
)

The path above is relative to the root directory of the published JAR (so src/main/protobuf is not needed). Users add your library to their projects like this:

libraryDependencies ++= Seq(
"com.example" %% "your-library" % "0.1.0",
"com.example" %% "your-library" % "0.1.0" % "protobuf"
)

The first dependency provides the precompiled class files. The second dependency makes it possible for users to import the protos in the jar file. sbt-protoc will look for the ScalaPB-Options-Proto attribute in the jar's manifest and automatically add the package scoped options file to the protoc command line.

note

Since the package-scoped options file is used as a source file in multiple projects, it should not define any types (messages, enums, services). This ensures that the package-scoped proto file does not generate any code on its own so we don't end up with duplicate class files.

Disabling package-scoped options processing#

As a consumer of third-party dependencies that come with options proto, you can disable the behavior of automatically adding the options proto to protoc by setting

Compile / PB.manifestProcessing := false

in sbt. In that case, it is your responsibility to either manually import the option protos in one of your own project source files so it gets applied, or ensure that the generator settigs used in your project are consistent with the ones used to generate the dependency. Differences in settings can lead to generated code that does not compile.

Auxiliary options#

In some situations, you may want to set some options in a proto file, but without modifying the original proto file or adding anything ScalaPB-specific to it. To accomplish that, you can define auxiliary options under package-scoped options.

For example, if you are given this proto file:

syntax = "proto3";
package a.b.c;
message Foo {
string hello = 1;
}

You can add a file package.proto with the following content:

syntax = "proto3";
package a.b.c;
import "scalapb/scalapb.proto";
option (scalapb.options) = {
scope: PACKAGE
aux_message_options: [
{
target: "a.b.c.Foo"
options: {
extends: "com.myexample.SomeTrait"
}
}
]
aux_field_options: [
{
target: "a.b.c.Foo.hello"
options: {
scala_name: "goodbye"
}
}
]
};

The list aux_message_options contains options targeted at different messages define under the same proto package of the package-scoped options. The target name needs to be fully-qualified message name in the protobuf namespace. Similar to aux_message_options, we also have aux_enum_options, aux_enum_value_options and aux_field_options. See example usage here.

Primitive wrappers#

In proto 3, unlike proto 2, primitives are not wrapped in an option by default. The standard technique to obtain an optional primitive is to wrap it inside a message (since messages are provided inside an Option). Google provides standard wrappers to the primitive types in wrappers.proto.

primitive_wrappers is enabled by default for ScalaPB>=0.6.0. Whenever one of the standard wrappers is used, it will be mapped to Option[X] where X is a primitive type. For example:

syntax = "proto3";
import "google/protobuf/wrappers.proto";
message MyMessage {
google.protobuf.Int32Value my_int32 = 5;
}

would generate

case class MyMessage(myInt32: Option[Int]) extends ...

To disable primitive wrappers in a file:

import "scalapb/scalapb.proto";
option (scalapb.options) = {
no_primitive_wrappers: true
};

In versions of ScalaPB prior to 0.6.0, primitive wrappers had to be turned on manually in each file:

import "scalapb/scalapb.proto";
option (scalapb.options) = {
primitive_wrappers: true
};

Custom base traits for messages#

Note: this option is available in ScalaPB 0.6.1 and later.

ScalaPBs allows you to specify custom base traits to a generated case class. This is useful when you have a few messages that share common fields and you would like to be able to access those fields through a single trait.

Example:

import "scalapb/scalapb.proto";
message CustomerWithPhone {
option (scalapb.message).extends = "com.thesamet.pb.BaseCustomer";
optional string customer_id = 1;
optional string name = 2;
optional string phone = 3;
}

In your code, define the base trait DomainEvent and include any subset of the fields:

package com.thesamet.pb
trait BaseCustomer {
def customerId: Option[String]
def name: Option[String]
}

You can specify any number of base traits for a message.

It is also possible to make the generated companion classes extend a class or trait, by using the companion_extends option. For example:

message MyMessage {
option (scalapb.message).extends = "MySuperClass";
option (scalapb.message).companion_extends = "MySuperCompanionClass";
int32 n = 1;
}

Will generate a case class that extends MySuperClass, and the companion object will extend MySuperCompanionClass.

Custom base traits for sealed oneofs#

Note: this option is available in ScalaPB 0.9.0 and later.

Use the following option to define one or more base traits for a generated SealedOneof:

message MyEither {
option (scalapb.message).sealed_oneof_extends = "MyBaseTrait";
oneof sealed_value {
Left left = 1;
Right right = 2;
}
}

Custom base traits for enums#

In a similar fashion to custom base traits for messages, it is possible to define custom base traits for enum types, for the companion objects of enum types and even for specific values.

For example:

syntax = "proto2";
package enum_example;
import "scalapb/scalapb.proto";
enum MyEnum {
option (scalapb.enum_options).extends = "example.EnumOptions.EnumBase";
option (scalapb.enum_options).companion_extends = "example.EnumOptions.EnumCompanionBase";
Unknown = 0;
V1 = 1 [(scalapb.enum_value).extends = "example.EnumOptions.ValueMixin"];
V2 = 2;
}

The generated code will look something like this:

sealed trait MyEnum extends GeneratedEnum
with example.EnumOptions.EnumBase {
/* ... */
}
object MyEnum extends GeneratedEnumCompanion[MyEnum]
with example.EnumOptions.EnumCompanionBase {
case object Unknown extends MyEnum { /* ... */ }
case object V1 extends MyEnum
with example.EnumOptions.ValueMixin { /* ... */ }
case object V2 extends MyEnum { /* ... */ }
/* ... */
}

Custom types#

You can customize the Scala type of any field. One use-case for this is when you would like to use type-safe wrappers around primitive values to enforce unit correctness. For example, instead of using a raw integer for time fields, you can wrap them in a Seconds class.

import "scalapb/scalapb.proto";
message Connection {
optional int32 timeout = 1 [(scalapb.field).type = "mydomain.Seconds"];
}

We would like to write code like this:

val c = Connection().update(_.timeout := Seconds(5))

How will ScalaPB know how to convert from the original type (Integer) to the custom type Seconds? For each custom type you need to define an implicit TypeMapper that will tell ScalaPB how to convert between the custom type and the base Scala type. A good place to define this implicit is in the companion class for your custom type, since the Scala compiler will look for a typemapper there by default. If your typemapper is defined elsewhere, you will need to import it manually by using the import file-level option.

package mydomain
case class Seconds(v: Int) extends AnyVal
object Seconds {
implicit val typeMapper = TypeMapper(Seconds.apply)(_.v)
}

TypeMapper takes two function parameters. The first converts from the original type to the custom type. The second function converts from the custom type to the original type.

In addition to primitive values, you can customize enums and messages as well.

For more examples, see:

If you have a TypeMapper that maps a generated type into a type you don't own (such as String, or a third-party class) then you don't have access to the companion object to define the typemapper in. Instead, you can place the typemapper in one of the parent package objects of the generated code. For example, if you want to map an enum to a string, and the message containing it goes into the a.b.c package, you can define the type mapper like this:

// src/main/scala/a/b/c/package.scala:
package a.b
package object c {
implicit val segmentType =
TypeMapper[SegmentType, String](_.name)(SegmentType.fromName(_).get)
}

Message-level custom type and boxing#

In the previous section you saw how to customize the type generated for a specific field. ScalaPB also lets you specify a custom type at the message level. When type is set at the message level, that type is used for all the fields that use that message. This eliminates the need to specify type on each field of this type.

// duration.proto
syntax = "proto3";
package mytypes;
import "scalapb/scalapb.proto";
message Duration {
option (scalapb.message).type = "mytypes.MyDurationClass";
int32 seconds = 1;
}

In a Scala file define an implicit mapper:

import scalapb.TypeMapper
import mytypes.duration.Duration
case class MyDurationClass(seconds: Int)
object MyDurationClass {
implicit val tm = TypeMapper[Duration, MyDurationClass] {
d: Duration => MyDurationClass(d.seconds) } {
m: MyDurationClass => Duration(m.seconds)
}
}

Now, each time you reference Duration in a proto file, the generated field in Scala code will be of type MyDuration:

syntax = "proto3";
package mytypes;
import "scalapb/scalapb.proto";
import "duration.proto";
message Usage {
Duration dd = 1; // will become dd: Option[MyDuration]
repeated Duration ds = 2; // will become ds: Seq[MyDuration]
// You can eliminate the boxing of an optional field in an Option by using
// no_box
Duration dd_nobox = 3 [(scalapb.field).no_box = true]; // will become ddNoBox: MyDuration
}

If you do not want any instance of your message to be boxed (regardless if it has a custom type), you can set no_box at the message-level:

// duration_nobox.proto
syntax = "proto3";
package mytypes;
import "scalapb/scalapb.proto";
message Duration {
option (scalapb.message).type = "mytypes.MyDurationType";
option (scalapb.message).no_box = true; // do not wrap in Option
int32 seconds = 1;
}

Then when this message is used, it will not be wrapped in an Option. If no_box is specified at the field level, it overrides the value specified at the message level.

Custom types on maps#

Since version 0.6.0 it is possible to customize the key and value types of maps. Like the custom types described above you will need to have a TypeMapper for the custom type.

Example:

message CustomMaps {
// Will generate Map[String, com.thesamet.pb.Years]
map<string, int32> string_to_year = 1 [
(scalapb.field).value_type = "com.thesamet.pb.Years"];
// Will generate Map[PersonId, Int]
map<string, int32> person_to_int = 2 [
(scalapb.field).key_type = "com.thesamet.pb.PersonId"];
// Will generate Map[PersonId, com.thesamet.pb.Years]
map<string, int32> person_to_year = 3 [
(scalapb.field).key_type = "com.thesamet.pb.PersonId",
(scalapb.field).value_type = "com.thesamet.pb.Years"];
}

Example: see CustomMaps in maps.proto

You can also customize the collection type used for a map. See the next section for details.

Custom collection types#

By default, ScalaPB compiles repeated fields into a Seq[T]. When a message is parsed from bytes, the default implementation instantiates a Vector[T], which is a subtype of Seq[T]. You can instruct ScalaPB to use a different collection type for one field by specifying the collection_type option. You can also specify a collection_type for the entire proto file by specifying a collection_type at the file-level.

If both are defined then the field-level setting wins.

Similar to collection_type, we have map_type for map types. By default, ScalaPB generates scala.collection.immutable.Map for maps, and you can customize it at the field level, or file-level by specifying a map_type option.

map_type was introduced in ScalaPB 0.8.5.

import "scalapb/scalapb.proto";
option (scalapb.options) = {
collection_type: "Set"
};
message CollTest {
// Will generate Set[Int] due to file-level option.
repeated int32 rep1 = 1;
// Will generate an Array[String]
repeated string rep2 = 2 [(scalapb.field).collection_type="Array"];
// Will generate Seq[collection.immutable.Seq]
repeated bool rep3 = 3 [
(scalapb.field).collection_type="collection.immutable.Seq"];
map<int32, string> my_map [
(scalapb.field).map_type="collection.mutable.Map"];
]
}

Note on mutable collection: ScalaPB assumes that all data is immutable. For example, the result of serializedSize is cached in a private field. When choosing mutable collections, you must be careful not to mutate any collection after it has been passed to any message, or you might get some surprising results!

Note: using Array is not supported along with Java conversions.

Note: Most Scala collections can be used with this feature. If you are trying to implement your own collection type, it may be useful to check MyVector and MyMap, the simplest custom collection that is compatible with ScalaPB:

Custom names#

Sometimes it may be useful to manually specify the name of a field in the generated code. For example, if you have a field named hash_code, then the camel-case version of it would be hashCode. Since that name would conflict with the hashCode() method we inherit from Java, ScalaPB issues an error. You can tell ScalaPB to use an alternative name by using the scala_name option:

optional string hash_code = 1 [(scalapb.field).scala_name = "myHashCode"];

It is also possible to customize the Scala name of an enum value:

enum MyEnum {
DEFAULT = 0;
FOO = 1 [(scalapb.enum_value).scala_name = "Bar"];
}

The same customization can be applied to oneof fields:

oneof notify {
option (scalapb.oneof).scala_name = "myNotify";
string foo = 1;
int32 bar = 2;
}

Adding annotations#

Since ScalaPB 0.6.3, you can add annotations to the generated case classes like this:

message BarMessage {
option (scalapb.message).annotations = "@mypackage.CustomAnnotation";
option (scalapb.message).annotations = "@mypackage.CustomAnnotation1";
option (scalapb.message).annotations = "@mypackage.CustomAnnotation2";
}

In ScalaPB 0.7.0, you can add annotations to the companion object of a message and to individual fields:

message BarMessage {
option (scalapb.message).companion_annotations = "@mypackage.AnotherAnnotation2";
optional string x = 1 [
(scalapb.field).annotations = '@deprecated("Will be gone", "1.0")'
];
}

In ScalaPB 0.10.9, you can also add annotations to the auto generated unknownFields field:

message BarMessage {
option (scalapb.message).unknown_field_annotations = "@annotation1";
}

In ScalaPB 0.11.4, you can also add annotations to the enum values and the Unrecognized case class:

enum BarEnum {
option (scalapb.enum_options) = "@annotation"
}
enum BarEnum {
// every value will have the annotation added.
option (scalapb.enum_options).base_annotations = "@annotation1";
// only known values (case objects) will have the annotation added.
option (scalapb.enum_options).recognized_annotations = "@annotation2";
// only the unrecognized case class will have the annotation added.
option (scalapb.enum_options).unrecognized_annotations = "@annotation3";
// only this value (case object) will have the annotation added.
BarValue = 1 [(scalapb.enum_value).annotations = "@annotation4"];
}