Struct prost_build::Config

pub struct Config { /* private fields */ }

Configuration options for Protobuf code generation.

This configuration builder can be used to set non-default code generation options.

Implementations

impl Config

pub fn new() -> Config

Creates a new code generator configuration with default options.

pub fn btree_map<I, S>(&mut self, paths: I) -> &mut Self

where I: IntoIterator<Item = S>, S: AsRef<str>,

Configure the code generator to generate Rust BTreeMap fields for Protobuf map type fields.

Arguments

paths - paths to specific fields, messages, or packages which should use a Rust BTreeMap for Protobuf map fields. Paths are specified in terms of the Protobuf type name (not the generated Rust type name). Paths with a leading . are treated as fully qualified names. Paths without a leading . are treated as relative, and are suffix matched on the fully qualified field name. If a Protobuf map field matches any of the paths, a Rust BTreeMap field is generated instead of the default HashMap.

The matching is done on the Protobuf names, before converting to Rust-friendly casing standards.

Examples

// Match a specific field in a message type.
config.btree_map(&[".my_messages.MyMessageType.my_map_field"]);

// Match all map fields in a message type.
config.btree_map(&[".my_messages.MyMessageType"]);

// Match all map fields in a package.
config.btree_map(&[".my_messages"]);

// Match all map fields. Specially useful in `no_std` contexts.
config.btree_map(&["."]);

// Match all map fields in a nested message.
config.btree_map(&[".my_messages.MyMessageType.MyNestedMessageType"]);

// Match all fields named 'my_map_field'.
config.btree_map(&["my_map_field"]);

// Match all fields named 'my_map_field' in messages named 'MyMessageType', regardless of
// package or nesting.
config.btree_map(&["MyMessageType.my_map_field"]);

// Match all fields named 'my_map_field', and all fields in the 'foo.bar' package.
config.btree_map(&["my_map_field", ".foo.bar"]);

pub fn bytes<I, S>(&mut self, paths: I) -> &mut Self

where I: IntoIterator<Item = S>, S: AsRef<str>,

Configure the code generator to generate Rust bytes::Bytes fields for Protobuf bytes type fields.

Arguments

paths - paths to specific fields, messages, or packages which should use a Rust Bytes for Protobuf bytes fields. Paths are specified in terms of the Protobuf type name (not the generated Rust type name). Paths with a leading . are treated as fully qualified names. Paths without a leading . are treated as relative, and are suffix matched on the fully qualified field name. If a Protobuf map field matches any of the paths, a Rust Bytes field is generated instead of the default Vec<u8>.

The matching is done on the Protobuf names, before converting to Rust-friendly casing standards.

Examples

// Match a specific field in a message type.
config.bytes(&[".my_messages.MyMessageType.my_bytes_field"]);

// Match all bytes fields in a message type.
config.bytes(&[".my_messages.MyMessageType"]);

// Match all bytes fields in a package.
config.bytes(&[".my_messages"]);

// Match all bytes fields. Specially useful in `no_std` contexts.
config.bytes(&["."]);

// Match all bytes fields in a nested message.
config.bytes(&[".my_messages.MyMessageType.MyNestedMessageType"]);

// Match all fields named 'my_bytes_field'.
config.bytes(&["my_bytes_field"]);

// Match all fields named 'my_bytes_field' in messages named 'MyMessageType', regardless of
// package or nesting.
config.bytes(&["MyMessageType.my_bytes_field"]);

// Match all fields named 'my_bytes_field', and all fields in the 'foo.bar' package.
config.bytes(&["my_bytes_field", ".foo.bar"]);

pub fn field_attribute<P, A>(&mut self, path: P, attribute: A) -> &mut Self

where P: AsRef<str>, A: AsRef<str>,

Add additional attribute to matched fields.

Arguments

path - a path matching any number of fields. These fields get the attribute. For details about matching fields see btree_map.

attribute - an arbitrary string that’ll be placed before each matched field. The expected usage are additional attributes, usually in concert with whole-type attributes set with type_attribute, but it is not checked and anything can be put there.

Note that the calls to this method are cumulative ‒ if multiple paths from multiple calls match the same field, the field gets all the corresponding attributes.

Examples

// Prost renames fields named `in` to `in_`. But if serialized through serde,
// they should as `in`.
config.field_attribute("in", "#[serde(rename = \"in\")]");

pub fn type_attribute<P, A>(&mut self, path: P, attribute: A) -> &mut Self

where P: AsRef<str>, A: AsRef<str>,

Add additional attribute to matched messages, enums and one-ofs.

Arguments

paths - a path matching any number of types. It works the same way as in btree_map, just with the field name omitted.

attribute - an arbitrary string to be placed before each matched type. The expected usage are additional attributes, but anything is allowed.

The calls to this method are cumulative. They don’t overwrite previous calls and if a type is matched by multiple calls of the method, all relevant attributes are added to it.

For things like serde it might be needed to combine with field attributes.

Examples

// Nothing around uses floats, so we can derive real `Eq` in addition to `PartialEq`.
config.type_attribute(".", "#[derive(Eq)]");
// Some messages want to be serializable with serde as well.
config.type_attribute("my_messages.MyMessageType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");
config.type_attribute("my_messages.MyMessageType.MyNestedMessageType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");

Oneof fields

The oneof fields don’t have a type name of their own inside Protobuf. Therefore, the field name can be used both with type_attribute and field_attribute ‒ the first is placed before the enum type definition, the other before the field inside corresponding message struct.

In other words, to place an attribute on the enum implementing the oneof, the match would look like my_messages.MyMessageType.oneofname.

pub fn message_attribute<P, A>(&mut self, path: P, attribute: A) -> &mut Self

where P: AsRef<str>, A: AsRef<str>,

Add additional attribute to matched messages.

Arguments

paths - a path matching any number of types. It works the same way as in btree_map, just with the field name omitted.

attribute - an arbitrary string to be placed before each matched type. The expected usage are additional attributes, but anything is allowed.

The calls to this method are cumulative. They don’t overwrite previous calls and if a type is matched by multiple calls of the method, all relevant attributes are added to it.

For things like serde it might be needed to combine with field attributes.

Examples

// Nothing around uses floats, so we can derive real `Eq` in addition to `PartialEq`.
config.message_attribute(".", "#[derive(Eq)]");
// Some messages want to be serializable with serde as well.
config.message_attribute("my_messages.MyMessageType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");
config.message_attribute("my_messages.MyMessageType.MyNestedMessageType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");

pub fn enum_attribute<P, A>(&mut self, path: P, attribute: A) -> &mut Self

where P: AsRef<str>, A: AsRef<str>,

Add additional attribute to matched enums and one-ofs.

Arguments

paths - a path matching any number of types. It works the same way as in btree_map, just with the field name omitted.

attribute - an arbitrary string to be placed before each matched type. The expected usage are additional attributes, but anything is allowed.

The calls to this method are cumulative. They don’t overwrite previous calls and if a type is matched by multiple calls of the method, all relevant attributes are added to it.

For things like serde it might be needed to combine with field attributes.

Examples

// Nothing around uses floats, so we can derive real `Eq` in addition to `PartialEq`.
config.enum_attribute(".", "#[derive(Eq)]");
// Some messages want to be serializable with serde as well.
config.enum_attribute("my_messages.MyEnumType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");
config.enum_attribute("my_messages.MyMessageType.MyNestedEnumType",
                      "#[derive(Serialize)] #[serde(rename_all = \"snake_case\")]");

Oneof fields

The oneof fields don’t have a type name of their own inside Protobuf. Therefore, the field name can be used both with enum_attribute and field_attribute ‒ the first is placed before the enum type definition, the other before the field inside corresponding message struct.

In other words, to place an attribute on the enum implementing the oneof, the match would look like my_messages.MyNestedMessageType.oneofname.

pub fn boxed<P>(&mut self, path: P) -> &mut Self

where P: AsRef<str>,

Wrap matched fields in a Box.

Arguments

path - a path matching any number of fields. These fields get the attribute. For details about matching fields see btree_map.

Examples

config.boxed(".my_messages.MyMessageType.my_field");

pub fn service_generator( &mut self, service_generator: Box<dyn ServiceGenerator>, ) -> &mut Self

Configures the code generator to use the provided service generator.

pub fn compile_well_known_types(&mut self) -> &mut Self

Configures the code generator to not use the prost_types crate for Protobuf well-known types, and instead generate Protobuf well-known types from their .proto definitions.

pub fn disable_comments<I, S>(&mut self, paths: I) -> &mut Self

where I: IntoIterator<Item = S>, S: AsRef<str>,

Configures the code generator to omit documentation comments on generated Protobuf types.

Example

Occasionally .proto files contain code blocks which are not valid Rust. To avoid doctest failures, annotate the invalid code blocks with an ignore or no_run attribute, or disable doctests for the crate with a Cargo.toml entry. If neither of these options are possible, then omit comments on generated code during doctest builds:

let mut config = prost_build::Config::new();
config.disable_comments(&["."]);
config.compile_protos(&["src/frontend.proto", "src/backend.proto"], &["src"])?;

As with other options which take a set of paths, comments can be disabled on a per-package or per-symbol basis.

pub fn skip_debug<I, S>(&mut self, paths: I) -> &mut Self

where I: IntoIterator<Item = S>, S: AsRef<str>,

Skips generating impl Debug for types

pub fn extern_path<P1, P2>( &mut self, proto_path: P1, rust_path: P2, ) -> &mut Self

where P1: Into<String>, P2: Into<String>,

Declare an externally provided Protobuf package or type.

extern_path allows prost types in external crates to be referenced in generated code.

When prost compiles a .proto which includes an import of another .proto, it will automatically recursively compile the imported file as well. extern_path can be used to instead substitute types from an external crate.

Example

As an example, consider a crate, uuid, with a prost-generated Uuid type:

// uuid.proto

syntax = "proto3";
package uuid;

message Uuid {
    string uuid_str = 1;
}

The uuid crate implements some traits for Uuid, and publicly exports it:

// lib.rs in the uuid crate

include!(concat!(env!("OUT_DIR"), "/uuid.rs"));

pub trait DoSomething {
    fn do_it(&self);
}

impl DoSomething for Uuid {
    fn do_it(&self) {
        println!("Done");
    }
}

A separate crate, my_application, uses prost to generate message types which reference Uuid:

// my_application.proto

syntax = "proto3";
package my_application;

import "uuid.proto";

message MyMessage {
    uuid.Uuid message_id = 1;
    string some_payload = 2;
}

Additionally, my_application depends on the trait impls provided by the uuid crate:

// `main.rs` of `my_application`

use uuid::{DoSomething, Uuid};

include!(concat!(env!("OUT_DIR"), "/my_application.rs"));

pub fn process_message(msg: MyMessage) {
    if let Some(uuid) = msg.message_id {
        uuid.do_it();
    }
}

Without configuring uuid as an external path in my_application’s build.rs, prost would compile a completely separate version of the Uuid type, and process_message would fail to compile. However, if my_application configures uuid as an extern path with a call to .extern_path(".uuid", "::uuid"), prost will use the external type instead of compiling a new version of Uuid. Note that the configuration could also be specified as .extern_path(".uuid.Uuid", "::uuid::Uuid") if only the Uuid type were externally provided, and not the whole uuid package.

Usage

extern_path takes a fully-qualified Protobuf path, and the corresponding Rust path that it will be substituted with in generated code. The Protobuf path can refer to a package or a type, and the Rust path should correspondingly refer to a Rust module or type.

// Declare the `uuid` Protobuf package and all nested packages and types as externally
// provided by the `uuid` crate.
config.extern_path(".uuid", "::uuid");

// Declare the `foo.bar.baz` Protobuf package and all nested packages and types as
// externally provided by the `foo_bar_baz` crate.
config.extern_path(".foo.bar.baz", "::foo_bar_baz");

// Declare the `uuid.Uuid` Protobuf type (and all nested types) as externally provided
// by the `uuid` crate's `Uuid` type.
config.extern_path(".uuid.Uuid", "::uuid::Uuid");

pub fn file_descriptor_set_path<P>(&mut self, path: P) -> &mut Self

where P: Into<PathBuf>,

When set, the FileDescriptorSet generated by protoc is written to the provided filesystem path.

This option can be used in conjunction with the include_bytes! macro and the types in the prost-types crate for implementing reflection capabilities, among other things.

Example

In build.rs:

config.file_descriptor_set_path(
    PathBuf::from(env::var("OUT_DIR").expect("OUT_DIR environment variable not set"))
        .join("file_descriptor_set.bin"));

In lib.rs:

let file_descriptor_set_bytes = include_bytes!(concat!(env!("OUT_DIR"), "/file_descriptor_set.bin"));
let file_descriptor_set = prost_types::FileDescriptorSet::decode(&file_descriptor_set_bytes[..]).unwrap();

pub fn skip_protoc_run(&mut self) -> &mut Self

In combination with with file_descriptor_set_path, this can be used to provide a file descriptor set as an input file, rather than having prost-build generate the file by calling protoc.

In build.rs:

config.file_descriptor_set_path("path/from/build/system")
    .skip_protoc_run()
    .compile_protos(&["src/items.proto"], &["src/"]);

pub fn retain_enum_prefix(&mut self) -> &mut Self

Configures the code generator to not strip the enum name from variant names.

Protobuf enum definitions commonly include the enum name as a prefix of every variant name. This style is non-idiomatic in Rust, so by default prost strips the enum name prefix from variants which include it. Configuring this option prevents prost from stripping the prefix.

pub fn out_dir<P>(&mut self, path: P) -> &mut Self

where P: Into<PathBuf>,

Configures the output directory where generated Rust files will be written.

If unset, defaults to the OUT_DIR environment variable. OUT_DIR is set by Cargo when executing build scripts, so out_dir typically does not need to be configured.

pub fn default_package_filename<S>(&mut self, filename: S) -> &mut Self

where S: Into<String>,

Configures what filename protobufs with no package definition are written to. The filename will be appended with the .rs extension.

pub fn enable_type_names(&mut self) -> &mut Self

Configures the code generator to include type names.

Message types will implement Name trait, which provides type and package name. This is needed for encoding messages as Any type.

pub fn type_name_domain<I, S, D>(&mut self, paths: I, domain: D) -> &mut Self

where I: IntoIterator<Item = S>, S: AsRef<str>, D: AsRef<str>,

Specify domain names to use with message type URLs.

Domains

paths - a path matching any number of types. It works the same way as in btree_map, just with the field name omitted.

domain - an arbitrary string to be used as a prefix for type URLs.

Examples

// Full type URL of the message `google.profile.Person`,
// will be `type.googleapis.com/google.profile.Person`.
config.type_name_domain(&["."], "type.googleapis.com");

pub fn prost_path<S>(&mut self, path: S) -> &mut Self

where S: Into<String>,

Configures the path that’s used for deriving Message for generated messages. This is mainly useful for generating crates that wish to re-export prost. Defaults to ::prost::Message if not specified.

pub fn protoc_arg<S>(&mut self, arg: S) -> &mut Self

where S: AsRef<OsStr>,

Add an argument to the protoc protobuf compilation invocation.

Example build.rs

fn main() -> Result<()> {
  let mut prost_build = prost_build::Config::new();
  // Enable a protoc experimental feature.
  prost_build.protoc_arg("--experimental_allow_proto3_optional");
  prost_build.compile_protos(&["src/frontend.proto", "src/backend.proto"], &["src"])?;
  Ok(())
}

pub fn protoc_executable<S>(&mut self, executable: S) -> &mut Self

where S: Into<PathBuf>,

Set the path to protoc executable to be used by prost-build

Use the provided path to find protoc. This can either be a file name which is searched for in the PATH or an aboslute path to use a specific executable.

Example build.rs

fn main() -> Result<()> {
  let mut prost_build = prost_build::Config::new();
  prost_build.protoc_executable("protoc-27.1");
  prost_build.compile_protos(&["src/frontend.proto", "src/backend.proto"], &["src"])?;
  Ok(())
}

pub fn include_file<P>(&mut self, path: P) -> &mut Self

where P: Into<PathBuf>,

Configures the optional module filename for easy inclusion of all generated Rust files

If set, generates a file (inside the OUT_DIR or out_dir() as appropriate) which contains a set of pub mod XXX statements combining to load all Rust files generated. This can allow for a shortcut where multiple related proto files have been compiled together resulting in a semi-complex set of includes.

Turning a need for:

pub mod Foo {
    pub mod Bar {
        include!(concat!(env!("OUT_DIR"), "/foo.bar.rs"));
    }
    pub mod Baz {
        include!(concat!(env!("OUT_DIR"), "/foo.baz.rs"));
    }
}

Into the simpler:

include!(concat!(env!("OUT_DIR"), "/_includes.rs"));

pub fn format(&mut self, enabled: bool) -> &mut Self

Configures the code generator to format the output code via prettyplease.

By default, this is enabled but if the format feature is not enabled this does nothing.

pub fn compile_fds(&mut self, fds: FileDescriptorSet) -> Result<()>

Compile a FileDescriptorSet into Rust files during a Cargo build with additional code generator configuration options.

This method is like compile_protos function except it does not invoke protoc and instead requires the user to supply a FileDescriptorSet.

Example build.rs

fn main() -> std::io::Result<()> {
  let file_descriptor_set = fds();

  prost_build::Config::new()
    .compile_fds(file_descriptor_set)
}

pub fn load_fds( &mut self, protos: &[impl AsRef<Path>], includes: &[impl AsRef<Path>], ) -> Result<FileDescriptorSet>

Loads .proto files as a FileDescriptorSet. This allows inspection of the descriptors before calling Config::compile_fds. This could be used to change Config attributes after introspecting what is actually present in the .proto files.

Example build.rs

fn main() -> std::io::Result<()> {
  let mut config = Config::new();
  let file_descriptor_set = config.load_fds(&["src/frontend.proto", "src/backend.proto"], &["src"])?;

  // Add custom attributes to messages that are service inputs or outputs.
  for file in &file_descriptor_set.file {
      for service in &file.service {
          for method in &service.method {
              if let Some(input) = &method.input_type {
                  config.message_attribute(input, "#[derive(custom_proto::Input)]");
              }
              if let Some(output) = &method.output_type {
                  config.message_attribute(output, "#[derive(custom_proto::Output)]");
              }
          }
      }
  }

  config.compile_fds(file_descriptor_set)
}

pub fn compile_protos( &mut self, protos: &[impl AsRef<Path>], includes: &[impl AsRef<Path>], ) -> Result<()>

Compile .proto files into Rust files during a Cargo build with additional code generator configuration options.

This method is like the prost_build::compile_protos function, with the added ability to specify non-default code generation options. See that function for more information about the arguments and generated outputs.

The protos and includes arguments are ignored if skip_protoc_run is specified.

Example build.rs

fn main() -> Result<()> {
  let mut prost_build = prost_build::Config::new();
  prost_build.btree_map(&["."]);
  prost_build.compile_protos(&["src/frontend.proto", "src/backend.proto"], &["src"])?;
  Ok(())
}

pub fn generate( &mut self, requests: Vec<(Module, FileDescriptorProto)>, ) -> Result<HashMap<Module, String>>

Processes a set of modules and file descriptors, returning a map of modules to generated code contents.

This is generally used when control over the output should not be managed by Prost, such as in a flow for a protoc code generating plugin. When compiling as part of a build.rs file, instead use Self::compile_protos().

Trait Implementations

impl Debug for Config

fn fmt(&self, fmt: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for Config

fn default() -> Config

Returns the “default value” for a type.