ts-proto

ts-proto transforms your .proto files into strongly-typed, idiomatic TypeScript files!

Table of contents

• QuickStart
• Goals
• Example Types
• Highlights
• Current Disclaimers
• Auto-Batching / N+1 Prevention
• Usage
  • Supported options
  • Only Types
  • NestJS Support
• Building
• Assumptions
• Todo
• Typing Approach
• OneOf Handling
• Primitive Types
• Wrapper Types
• Number Types
• Current Status of Optional Values

Overview

ts-proto generates TypeScript types from protobuf schemas.

I.e. given a person.proto schema like:

message Person {
  string name = 1;
}

ts-proto will generate a person.ts file like:

interface Person {
  name: string
}

const Person = {
  encode(person): Writer { ... }
  decode(reader): Person { ... }
  toJSON(person): unknown { ... }
  fromJSON(data): Person { ... }
}

It also knows about services and will generate types for them as well, i.e.:

export interface PingService {
  ping(request: PingRequest): Promise<PingResponse>;
}

It will also generate client implementations of PingService; currently Twirp, grpc-web, and nestjs are supported.

QuickStart

• npm install ts-proto
• protoc --plugin=./node_modules/.bin/protoc-gen-ts_proto --ts_proto_out=. ./simple.proto
  • (Note that the output parameter name, ts_proto_out, is named based on the suffix of the plugin's name, i.e. "ts_proto" suffix in the --plugin=./node_modules/.bin/protoc-gen-ts_proto parameter becomes the _out prefix, per protoc's CLI conventions.)
  • On Windows, use protoc --plugin=protoc-gen-ts_proto=.\node_modules\.bin\protoc-gen-ts_proto.cmd --ts_proto_out=. ./imple.proto
  • Ensure you're using a modern protoc, i.e. the original protoc 3.0.0 doesn't support the _opt flag

This will generate *.ts source files for the given *.proto types.

If you want to package these source files into an npm package to distribute to clients, just run tsc on them as usual to generate the .js/.d.ts files, and deploy the output as a regular npm package.

Goals

• Idiomatic TypeScript/ES6 types
  • ts-proto is a clean break from either the built-in Google/Java-esque JS code of protoc or the "make .d.ts files the *.js comments" approach of protobufjs
  • (Techically the protobufjs/minimal package is used for actually reading/writing bytes.)
• TypeScript-first output
• Interfaces over classes
  • As much as possible, types are just interfaces, so you can work with messages just like regular hashes/data structures.
• Only supports codegen *.proto-to-*.ts workflow, currently no runtime reflection/loading of dynamic .proto files
• Currently ambivalent about browser support, current focus is on Node/server-side use cases

Example Types

The generated types are "just data", i.e.:

export interface Simple {
  name: string;
  age: number;
  createdAt: Date | undefined;
  child: Child | undefined;
  state: StateEnum;
  grandChildren: Child[];
  coins: number[];
}

Along with encode/decode factory methods:

export const Simple = {
  encode(message: Simple, writer: Writer = Writer.create()): Writer {
    ...
  },

  decode(reader: Reader, length?: number): Simple {
    ...
  },

  fromJSON(object: any): Simple {
    ...
  },

  fromPartial(object: DeepPartial<Simple>): Simple {
    ...
  },

  toJSON(message: Simple): unknown {
    ...
  },
};

This allows idiomatic TS/JS usage like:

const bytes = Simple.encode({ name: ..., age: ..., ... }).finish();
const simple = Simple.decode(Reader.create(bytes));
const { name, age } = simple;

Which can dramatically ease integration when converting to/from other layers without creating a class and calling the right getters/setters.

Highlights

• A poor man's attempt at "please give us back optional types"

  The canonical protobuf wrapper types, i.e. google.protobuf.StringValue, are mapped as optional values, i.e. string | undefined, which means for primitives we can kind of pretend the protobuf type system has optional types.

  (Update: ts-proto now also supports the proto3 optional keyword.)

• Timestamps are mapped as Date

• fromJSON/toJSON support the canonical Protobuf JS format (i.e. timestamps are ISO strings)

Current Disclaimers

ts-proto was originally developed for Twirp, so the clients it generates (if your *.proto files use the GRPC service constructs) assume they're talking to Twirp HTTP endpoints. There is an issue filed (#2) to support GRPC endpoints, but no work currently in progress.

That said, the message/interface types that ts-proto generates are not coupled to Twirp and should be fully usable in other Protobuf environments (either GRPC-based or even just reading protobuf files from disk/etc.). The client-related output can also be disabled (see the Usage section).

ts-proto also does not currently have any infrastructure to help implement the server-side of a GRPC (either Twirp or pure GRPC) service, i.e. built-in Express bindings or something like that. However, again, the types/interfaces that ts-proto generates for your messages and services are still generally very helpful in setting up your own server-side implementations.

Auto-Batching / N+1 Prevention

(Note: this is currently only supported by the Twirp clients.)

If you're using ts-proto's clients to call backend micro-services, similar to the N+1 problem in SQL applications, it is easy for micro-service clients to (when serving an individual request) inadvertantly trigger multiple separate RPC calls for "get book 1", "get book 2", "get book 3", that should really be batched into a single "get books [1, 2, 3]" (assuming the backend supports a batch-oriented RPC method).

ts-proto can help with this, and essentially auto-batch your individual "get book" calls into batched "get books" calls.

For ts-proto to do this, you need to implement your service's RPC methods with the batching convention of:

• A method name of Batch<OperationName>
• The Batch<OperationName> input type has a single repeated field (i.e. repeated string ids = 1)
• The Batch<OperationName> output type has either a:
  • A single repeated field (i.e. repeated Foo foos = 1) where the output order is the same as the input ids order, or
  • A map of the input to an output (i.e. map<string, Entity> entities = 1;)

When ts-proto recognizes methods of this pattern, it will automatically create a "non-batch" version of <OperationName> for the client, i.e. client.Get<OperationName>, that takes a single id and returns a single result.

This provides the client code with the illusion that it can make individual Get<OperationName> calls (which is generally preferrable/easier when implementing the client's business logic), but the actual implementation that ts-proto provides will end up making Batch<OperationName> calls to the backend service.

You also need to enable the useContext=true build-time parameter, which gives all client methods a Go-style ctx parameter, with a getDataLoaders method that lets ts-proto cache/resolve request-scoped DataLoaders, which provide the fundamental auto-batch detection/flushing behavior.

See the batching.proto file and related tests for examples/more details.

But the net effect is that ts-proto can provide SQL-/ORM-style N+1 prevention for clients calls, which can be critical especially in high-volume / highly-parallel implementations like GraphQL front-end gateways calling backend micro-services.

Usage

ts-proto is a protoc plugin, so you run it by (either directly in your project, or more likely in your mono-repo schema pipeline, i.e. like Ibotta or Namely):

• Add ts-proto to your package.json
• Run npm install to download it
• Invoke protoc with a plugin parameter like:

protoc --plugin=node_modules/ts-proto/protoc-gen-ts_proto ./batching.proto -I.

Supported options

• With --ts_proto_opt=context=true, the services will have a Go-style ctx parameter, which is useful for tracing/logging/etc. if you're not using node's async_hooks api due to performance reasons.

• With --ts_proto_opt=forceLong=long, all 64 bit numbers will be parsed as instances of Long (using the long library).

  Alternatively, if you pass --ts_proto_opt=forceLong=string, all 64 bit numbers will be outputted as strings.

• With --ts_proto_opt=env=node or browser or both, ts-proto will make environment-specific assumptions in your output. This defaults to both, which makes no environment-specific assumptions.

  Using node changes the types of bytes from Uint8Array to Buffer for easier integration with the node ecosystem which generally uses Buffer.

  Currently browser doesn't have any specific behavior other than being "not node". It probably will soon/at some point.

• With --ts_proto_opt=useOptionals=true, non-scalar fields are declared as optional TypeScript properties, e.g. field?: Message instead of field: Message | undefined.

  ts-proto defaults to useOptionals=false, e.g. field: Message | undefined, because it is the "most-safe" for use cases like:

  interface SomeMessage {
    firstName: string | undefined;
    lastName: string | undefined;
  }
  
  const data = { firstName: 'a', lastTypo: 'b' };
  
  // This will compile even though `lastTypo` means that `lastName` is not assigned
  const message: SomeMessage = {
    ...data,
  };

  However, the type-safety of useOptionals=false is admittedly tedious if you have many inherently-unused fields, so you can use useOptionals=true if that trade-off makes sense for your project.

  Eventually if TypesCript supports Exact Types, that should allow ts-proto to switch to useOptionals=true as the default/only behavior, have the generated Message.encode/Message.toPartial/etc. methods accept Exact<T> versions of the message types, and the result would be both safe + succinct.

  Also see the comment in this issue which explains some of the nuance behind making all fields optional (currently useOptionals only makes message fields optional), specifically that a message created with const message: Message = { ...key not set... } vs. const message = Message.decode(...key not set...) would look different to clients.

  (Also note that each message's Message.fromPartial(...) static methods are specifically meant to address this, because it allows you to create a message with all keys optional, but still applies the usual protobuf default-value-on-missing-key logic, so that code that reads the message get more consistent behavior.

• With --ts_proto_opt=oneof=unions, oneof fields will be generated as ADTs.

  See the "OneOf Handling" section.

• With --ts_proto_opt=lowerCaseServiceMethods=true, the method names of service methods will be lowered/camel-case, i.e. service.findFoo instead of service.FindFoo.

• With --ts_proto_opt=outputEncodeMethods=false, the Message.encode and Message.decode methods for working with protobuf-encoded/binary data will not be output.

  This is useful if you want "only types".

• With --ts_proto_opt=outputJsonMethods=false, the Message.fromJSON and Message.toJSON methods for working with JSON-coded data will not be output.

  This is also useful if you want "only types".

• With --ts_proto_opt=outputClientImpl=false, the client implementations, i.e. FooServiceClientImpl, that implement the client-side (in Twirp, see next option for grpc-web) RPC interfaces will not be output.

• With --ts_proto_opt=outputClientImpl=grpc-web, the client implementations, i.e. FooServiceClientImpl, will use the @improbable-eng/grpc-web library at runtime to send grpc messages to a grpc-web backend.

  (Note that this only uses the grpc-web runtime, you don't need to use any of their generated code, i.e. the ts-proto output replaces their ts-protoc-gen output.)

  You'll need to add the @improbable-eng/grpc-web and a transport to your project's package.json; see the integration/grpc-web directory for a working example.

• With --ts_proto_opt=returnObservable=true, the return type of service methods will be Observable<T> instead of Promise<T>.

• With--ts_proto_opt=addGrpcMetadata=true, the last argument of service methods will accept the grpc Metadata type, which contains additional information with the call (i.e. access tokens/etc.).

  (Requires nestJs=true.)

• With--ts_proto_opt=addNestjsRestParameter=true, the last argument of service methods will be an rest parameter with type any. This way you can use custom decorators you could normally use in nestjs.

  (Requires nestJs=true.)

• With --ts_proto_opt=nestJs=true, the defaults will change to generate NestJS protobuf friendly types & service interfaces that can be used in both the client-side and server-side of NestJS protobuf implementations. See the nestjs readme for more information and implementation examples.

  Specifically outputEncodeMethods, outputJsonMethods, and outputClientImpl will all be false, and lowerCaseServiceMethods will be true.

  Note that addGrpcMetadata, addNestjsRestParameter and returnObservable will still be false.

Only Types

If you're looking for ts-proto to generate only types for your Protobuf types then passing all three of outputEncodeMethods, outputJsonMethods, and outputClientImpl as false is probably what you want, i.e.:

--ts_proto_opt=outputEncodeMethods=false,outputJsonMethods=false,outputClientImpl=false.

NestJS Support

We have a great way of working together with nestjs. ts-proto generates interfaces and decorators for you controller, client. For more information see the nestjs readme.

Sponsors

Kudos to our sponsors:

• ngrok funded ts-proto's initial grpc-web support.

If you need ts-proto customizations or priority support for your company, you can ping me at via email.

Building

ts-proto does not use pbjs at runtime, but we do use it in the ts-proto build process (to bootstrap the types used to parse the incoming protobuf metadata types, as well as for the test suite to ensure the ts-proto implementations match the ts-proto).

After running yarn install, run ./pbjs.sh to create the bootstrap types, and ./integration/pbjs.sh to create the integration test types. These pbjs-generated files are not currently checked in.

After this the tests should pass.

After making changes to ts-proto, you can run cd integration and ./codegen.sh to re-generate the test case *.ts output files that are in each integration/<test-case>/ directory.

The test suite's proto files (i.e. simple.proto, batching.proto, etc.) currently have serialized/.bin copies checked into git (i.e. simple.bin, batching.bin, etc.), so that the test suite can run without having to invoke the protoc build chain. I.e. if you change the simple.proto/etc. files, you'll need to run ./integration/update-bins.sh, which does require having the protoc executable available.

Assumptions

• TS/ES6 module name is the proto package

Todo

• Support the string-based encoding of duration in fromJSON/toJSON
• Support the json_name annotation
• Make oneof=unions the default behavior in 2.0
• Probably change forceLong default in 2.0, should default to forceLong=long

OneOf Handling

By default, oneof fields are modeled "flatly" in the message, i.e. oneof either_field { string field_a; string field_b } means that the message will have field_a: string | undefined; field_b: string | undefined.

With this output, you'll have to check both if object.field_a and if object.field_b, and if you set one, you'll have to remember to unset the other.

We recommend using the oneof=unions option, which will change the output to be an Abstract Data Type/ADT like:

interface YourMessage {
  eitherField: { $case: 'field_a'; field_a: string } | { $case: 'field_b'; field_b: string };
}

As this will automatically enforce only one of field_a or field_b "being set" at a time, because the values are stored in the eitherField field that can only have a single value at a time.

In ts-proto's currently-unscheduled 2.x release, oneof=unions will become the default behavior.

Primitive Types

Protobuf has the somewhat annoying behavior that primitives types cannot differentiate between set-to-defalut-value and unset.

I.e. if you have a string name = 1, and set object.name = '', Protobuf will skip sending the tagged name field over the wire, because its understood that readers on the other end will, when they see name is not included in the payload, return empty string.

ts-proto models this behavior, of "unset" values being the primitive's default. (Technically by setting up an object prototype that knows the default values of the message's primitive fields.)

If you want fields where you can model set/unset, see Wrapper Types.

Wrapper Types

In core Protobuf, unset primitive fields become their respective default values (so you loose ability to distinguish "unset" from "default").

However, unset message fields stay null.

This allows a cute hack where you can model a logical string | unset by creating a field that is technically a message (i.e. so it can stay null for the unset case), but the message only has a single string field (i.e for storing the value in the set case).

Protobuf has already "blessed" this pattern with several built-in types, i.e. google.protobuf.StringValue, google.protobuf.Int32Value, etc.

ts-proto understands these wrapper types and "re-idiomizes" them by generating a google.protobuf.StringValue name = 1 field as a name: string | undefined, and hides the StringValue implementation detail from your code (i.e. during encode/decode of the name field on the wire to external consumers, it's still read/written as a StringValue message field).

This makes dealing with string | unset in your code much nicer, albeit it's unfortunate that, in Protobuf core, this is not as simple as marking a string name = 1 field as optional, i.e. you have to "dirty" your proto files a bit by knowing to use the StringValue convention.

Number Types

Numbers are by default assumed to be plain JavaScript numbers.

This is fine for Protobuf types like int32 and float, but 64-bit types like int64 can't be 100% represented by JavaScript's number type, because int64 can have larger/smaller values than number.

ts-proto's default configuration (which is forceLong=number) is to still use number for 64-bit fields, and then throw an error if a value (at runtime) is larger than Number.MAX_SAFE_INTEGER.

If you expect to use 64-bit / higher-than-MAX_SAFE_INTEGER values, then you can use the ts-proto forceLong option, which uses the long npm package to support the entire range of 64-bit values.

The protobuf number types map to JavaScript types based on the forceLong config option:

Where (*) indicates they might throw an error at runtime.

Current Status of Optional Values

• Required primitives: use as-is, i.e. string name = 1.
• Optional primitives: use wrapper types, i.e. StringValue name = 1.
• Required messages: not available
• Optional primitives: use as-is, i.e. SubMessage message = 1.