Protocol Buffers are a language-neutral, platform-neutral, extensible way of serializing structured data for use in communications protocols, data storage, and more, originally designed at Google (see).
protobuf.js is a pure JavaScript implementation with TypeScript support for node.js and the browser. It's super easy to use, blazingly fast and works out of the box with .proto files!
-
Installation
How to include protobuf.js in your project. -
Usage
A brief introduction to using the toolset. -
Examples
A few examples to get you started. -
Documentation
A list of available documentation resources. -
Command line
How to use the command line utility. -
Performance
A few internals and a benchmark on performance. -
Compatibility
Notes on compatibility regarding browsers and optional libraries. -
Building
How to build the library and its components yourself.
$> npm install protobufjs [--save --save-prefix=~]
var protobuf = require("protobufjs");
Development:
<script src="//cdn.rawgit.com/dcodeIO/protobuf.js/6.X.X/dist/protobuf.js"></script>
Production:
<script src="//cdn.rawgit.com/dcodeIO/protobuf.js/6.X.X/dist/protobuf.min.js"></script>
NOTE: Remember to replace the version tag with the exact release your project depends upon.
The protobuf
namespace will always be available globally / also supports AMD loaders.
The library supports both reflection-based and code-based use cases:
- Parsing protocol buffer definitions (.proto files) to reflection
- Loading JSON descriptors to reflection
- Generating static code without any reflection features
Where bundle size is a factor, there is a suitable distribution for each of these:
Gzipped | Downloads | How to require | Description | |
---|---|---|---|---|
full | 18.5kb | dist | require("protobufjs") |
All features. Works with everything. |
light | 15.5kb | dist/light | require("protobufjs/light") |
All features except tokenizer, parser and bundled common types. Works with JSON definitions, pure reflection and static code. |
minimal | 6.0kb+ | dist/minimal | require("protobufjs/minimal") |
Just enough to run static code. No reflection. |
In case of doubt you can just use the full library.
Each message type provides a set of methods with each method doing just one thing. This allows to avoid unnecessary operations where performance is a concern but also forces a user to perform verification explicitly where necessary - for example when dealing with user input.
Note that Message refers to any message type below.
-
Message.verify(message:
Object
):null|string
explicitly performs verification prior to encoding a plain object. Instead of throwing, it returns the error message as a string, if any.var payload = "invalid (not an object)"; var err = AwesomeMessage.verify(payload); if (err) throw Error(err);
-
Message.encode(message:
Message|Object
[, writer:Writer
]):Writer
is an automatically generated message specific encoder expecting a valid message or plain object. Note that this method does not implicitly verify the message and that it's up to the user to make sure that the data can actually be encoded properly.var buffer = AwesomeMessage.encode(message).finish();
-
Message.encodeDelimited(message:
Message|Object
[, writer:Writer
]):Writer
works likeMessage.encode
but additionally prepends the length of the message as a varint. -
Message.decode(reader:
Reader|Uint8Array
):Message
is an automatically generated message specific decoder. If required fields are missing, it throws autil.ProtocolError
with aninstance
property set to the so far decoded message. If the wire format is invalid, it throws anError
. The result is a runtime message.try { var decodedMessage = AwesomeMessage.decode(buffer); } catch (e) { if (e instanceof protobuf.util.ProtocolError) { // e.instance holds the so far decoded message with missing required fields } else { // wire format is invalid } }
-
Message.decodeDelimited(reader:
Reader|Uint8Array
):Message
works likeMessage.decode
but additionally reads the length of the message prepended as a varint. -
Message.create(properties:
Object
):Message
quickly creates a new runtime message from known to be valid properties without any conversion being performed. Where applicable, it is recommended to preferMessage.create
overMessage.fromObject
.var message = AwesomeMessage.create({ awesomeField: "AwesomeString" });
-
Message.fromObject(object:
Object
):Message
converts any plain object to a runtime message. Tries to convert whatever is specified (useMessage.verify
before if necessary).var message = AwesomeMessage.fromObject({ awesomeField: 42 }); // converts awesomeField to a string
-
Message.toObject(message:
Message
[, options:ConversionOptions
]):Object
converts a runtime message to a plain object.var object = AwesomeMessage.toObject(message, { enums: String, // enums as string names longs: String, // longs as strings (requires long.js) bytes: String, // bytes as base64 encoded strings defaults: true, // includes default values arrays: true, // populates empty arrays (repeated fields) even if defaults=false objects: true, // populates empty objects (map fields) even if defaults=false oneofs: true // includes virtual oneof fields set to the present field's name });
See also: ConversionOptions
What is a valid message?
A valid message is an object not missing any required fields and exclusively using JS types for its fields that are understood by the wire format writer.
- Calling
Message.verify
with a valid message returnsnull
and otherwise the error as a string. - Calling
Message.create
orMessage.encode
with any object assumes valid types. - Calling
Message.fromObject
with any object naively converts all values to the optimal JS type.
Type | Expected JS type (create) | Naive conversion (fromObject) |
---|---|---|
int32 uint32 sint32 fixed32 sfixed32 |
Number (32 bit integer) |
`value |
int64 uint64 sint64 fixed64 sfixed64 |
Long -like (optimal)Number (53 bit integer) |
Long.fromValue(value) parseInt(value, 10) without long.js |
float double |
Number |
Number(value) |
bool | Boolean |
Boolean(value) |
string | String |
String(value) |
bytes | Uint8Array (optimal)Buffer (optimal)Array.<Number> (8 bit integers)String (base64) |
base64.decode(value) if a StringObject with non-zero .length is kept |
enum | Number (32 bit integer) |
Looks up the numeric id if a string |
message | Valid message | Message.fromObject(value) |
- Explicit
undefined
andnull
are considered as not set when optional. - Repeated fields are
Array.<T>
. - Map fields are
Object.<string,T>
with the key being the string representation of the respective value or an 8 characters long binary hash string for Long-likes. String
refers to both objects and values whileNumber
refers to values only.
It is possible to load existing .proto files using the full library, which parses and compiles the definitions to ready to use (reflection-based) message classes:
// awesome.proto
package awesomepackage;
syntax = "proto3";
message AwesomeMessage {
string awesome_field = 1; // becomes awesomeField
}
protobuf.load("awesome.proto", function(err, root) {
if (err)
throw err;
// Obtain a message type
var AwesomeMessage = root.lookupType("awesomepackage.AwesomeMessage");
// Exemplary payload
var payload = { awesomeField: "AwesomeString" };
// Verify the payload if necessary (i.e. when possibly incomplete or invalid)
var errMsg = AwesomeMessage.verify(payload);
if (errMsg)
throw Error(errMsg);
// Create a new message
var message = AwesomeMessage.fromObject(payload); // or use .create if payload is already known to be valid
// Encode a message to an Uint8Array (browser) or Buffer (node)
var buffer = AwesomeMessage.encode(message).finish();
// ... do something with buffer
// Decode an Uint8Array (browser) or Buffer (node) to a message
var message = AwesomeMessage.decode(buffer);
// ... do something with message
// If your application uses length-delimited buffers, there is also encodeDelimited and decodeDelimited.
// Maybe convert the message back to a plain object
var object = AwesomeMessage.toObject(message, {
longs: String,
enums: String,
bytes: String,
// see ConversionOptions
});
});
Additionally, promise syntax can be used by omitting the callback, if preferred:
protobuf.load("awesome.proto")
.then(function(root) {
...
});
The library utilizes a JSON format that is equivalent to a .proto definition (see also: Command line usage).
The following is identical to the .proto definition seen above, but it can also be used with just the light library because it doesn't require the parser:
// awesome.json
{
"nested": {
"AwesomeMessage": {
"fields": {
"awesomeField": {
"type": "string",
"id": 1
}
}
}
}
}
A JSON descriptor can either be loaded the usual way:
protobuf.load("awesome.json", function(err, root) {
if (err) throw err;
// Continue at "Obtain a message type" above
});
Or you can load it inline:
var jsonDescriptor = require("./awesome.json"); // exemplary for node
var root = protobuf.Root.fromJSON(jsonDescriptor);
// Continue at "Obtain a message type" above
Both the full and the light library include full reflection support. You could, for example, define the .proto definitions seen in the examples above using just reflection:
...
var Root = protobuf.Root,
Type = protobuf.Type,
Field = protobuf.Field;
var AwesomeMessage = new Type("AwesomeMessage").add(new Field("awesomeField", 1, "string"));
var root = new Root().define("awesomepackage").add(AwesomeMessage);
// Continue at "Create a new message" above
...
Detailed information on the reflection structure is available within the documentation.
You can also extend runtime message classes with your own custom functionality and even register your own constructor with a reflected message type:
...
// Define your own constructor
function AwesomeMessage(properties) {
// custom initialization code
...
}
// Register your constructor with its reflected type (*)
root.lookupType("awesomepackage.AwesomeMessage").ctor = AwesomeMessage;
// Define your custom functionality
AwesomeMessage.customStaticMethod = function() { ... };
AwesomeMessage.prototype.customInstanceMethod = function() { ... };
// Continue at "Create a new message" above
(*) Besides referencing its reflected type through AwesomeMessage.$type
and AwesomeMesage#$type
, the respective custom class is automatically populated with:
AwesomeMessage.create
AwesomeMessage.encode
andAwesomeMessage.encodeDelimited
AwesomeMessage.decode
andAwesomeMessage.decodeDelimited
AwesomeMessage.verify
AwesomeMessage.fromObject
,AwesomeMessage.toObject
,AwesomeMessage#toObject
andAwesomeMessage#toJSON
Afterwards, decoded messages of this type are instanceof AwesomeMessage
.
Alternatively, you can also just reuse and extend the internal constructor if custom initialization code is not required:
...
// Reuse the internal constructor
var AwesomeMessage = root.lookupType("awesomepackage.AwesomeMessage").ctor;
// Define your custom functionality
AwesomeMessage.customStaticMethod = function() { ... };
AwesomeMessage.prototype.customInstanceMethod = function() { ... };
// Continue at "Create a new message" above
The library also supports services but it doesn't make any assumptions about the actual transport channel. Instead, a user must provide a suitable RPC implementation, which is an asynchronous function that takes the reflected service method, the binary request and a node-style callback as its parameters:
function rpcImpl(method, requestData, callback) {
// perform the request using an HTTP request or a WebSocket for example
var responseData = ...;
// and call the callback with the binary response afterwards:
callback(null, responseData);
}
Example:
// greeter.proto
syntax = "proto3";
service Greeter {
rpc SayHello (HelloRequest) returns (HelloReply) {}
}
message HelloRequest {
string name = 1;
}
message HelloReply {
string message = 1;
}
...
var Greeter = root.lookup("Greeter");
var greeter = Greeter.create(/* see above */ rpcImpl, /* request delimited? */ false, /* response delimited? */ false);
greeter.sayHello({ name: 'you' }, function(err, response) {
console.log('Greeting:', response.message);
});
Services also support promises:
greeter.sayHello({ name: 'you' })
.then(function(response) {
console.log('Greeting:', response.message);
});
There is also an example for streaming RPC.
The library ships with its own type definitions and modern editors like Visual Studio Code should automatically detect and use them for code completion when following this pattern:
// node.js
import * as protobuf from "protobufjs";
import * as Long from "long"; // optional
// browser only (alternatively)
import * as protobuf from "./node_modules/protobufjs/index.js";
import * as Long from "./node_modules/long/dist/long.js"; // optional
protobuf.load("awesome.proto", function(err, root) {
if (err)
throw err;
// example code
var AwesomeMessage = root.lookupType("AwesomeMessage");
var message = AwesomeMessage.create({ awesomeField: "hello" });
var buffer = AwesomeMessage.encode(message).finish();
...
});
To achieve the same with static code generated by pbjs, there is the pbts command line utility to generate type definitions from static code as well.
Let's say you generated your static code to bundle.js
and its type definitions to bundle.d.ts
, then you can do:
import * as root from "./bundle.js";
// example code
var AwesomeMessage = root.AwesomeMessage;
var message = AwesomeMessage.create({ awesomeField: "hello" });
var buffer = AwesomeMessage.encode(message).finish();
...
- API Documentation
- CHANGELOG
- Frequently asked questions on our wiki
- Questions and answers on StackOverflow
The pbjs
command line utility can be used to bundle and translate between .proto and .json files. It also generates static code.
Consolidates imports and converts between file formats.
-t, --target Specifies the target format. Also accepts a path to require a custom target.
json JSON representation
json-module JSON representation as a module
proto2 Protocol Buffers, Version 2
proto3 Protocol Buffers, Version 3
static Static code without reflection
static-module Static code without reflection as a module
-p, --path Adds a directory to the include path.
-o, --out Saves to a file instead of writing to stdout.
Module targets only:
-w, --wrap Specifies the wrapper to use. Also accepts a path to require a custom wrapper.
default Default wrapper supporting both CommonJS and AMD
commonjs CommonJS wrapper
amd AMD wrapper
es6 ES6 wrapper (implies --es6)
-r, --root Specifies an alternative protobuf.roots name.
-l, --lint Linter configuration. Defaults to protobuf.js-compatible rules:
eslint-disable block-scoped-var, no-redeclare, no-control-regex, no-prototype-builtins
--es6 Enables ES6 syntax (const/let instead of var)
Proto sources only:
--keep-case Keeps field casing instead of converting to camel case.
Static targets only:
--no-create Does not generate create functions used for reflection compatibility.
--no-encode Does not generate encode functions.
--no-decode Does not generate decode functions.
--no-verify Does not generate verify functions.
--no-convert Does not generate convert functions like from/toObject
--no-delimited Does not generate delimited encode/decode functions.
--no-beautify Does not beautify generated code.
--no-comments Does not output any JSDoc comments.
usage: pbjs [options] file1.proto file2.json ... (or) other | pbjs [options] -
For production environments it is recommended to bundle all your .proto files to a single .json file, which minimizes the number of network requests and avoids any parser overhead (hint: works with just the light library):
$> pbjs -t json file1.proto file2.proto > bundle.json
Now, either include this file in your final bundle:
var root = protobuf.Root.fromJSON(require("./bundle.json"));
or load it the usual way:
protobuf.load("bundle.json", function(err, root) {
...
});
The pbjs
utility is also capable of generating static code (hint: works with just the minimal library). For example
$> pbjs -t static-module -w commonjs -o compiled.js file1.proto file2.proto
will generate static code for definitions within file1.proto
and file2.proto
to a CommonJS module compiled.js
.
ProTip! Documenting your .proto files with /** ... */
-blocks or (trailing) /// ...
lines translates to generated static code.
Likewise, the pbts
command line utility can be used to generate TypeScript definitions from pbjs
-generated static modules.
Generates TypeScript definitions from annotated JavaScript files.
-o, --out Saves to a file instead of writing to stdout.
-g, --global Name of the global object in browser environments, if any.
--no-comments Does not output any JSDoc comments.
Internal flags:
-n, --name Wraps everything in a module of the specified name.
-m, --main Whether building the main library without any imports.
usage: pbts [options] file1.js file2.js ... (or) other | pbts [options] -
Picking up on the example above, the following not just generates static code to a CommonJS module compiled.js
but also its respective TypeScript definitions to compiled.d.ts
:
$> pbjs -t static-module -w commonjs -o compiled.js file1.proto file2.proto
$> pbts -o compiled.d.ts compiled.js
Additionally, TypeScript definitions of static modules are compatible with their reflection-based counterparts (i.e. as exported by JSON modules), as long as the following conditions are met:
- Instead of using
new SomeMessage(...)
, always useSomeMessage.create(...)
because reflection objects do not provide a constructor. - Types, services and enums must start with an uppercase letter to become available as properties of the reflected types as well (i.e. to be able to use
MyMessage.MyEnum
instead ofroot.lookup("MyMessage.MyEnum")
).
For example, the following generates a JSON module bundle.js
and a bundle.d.ts
, but no static code:
$> pbjs -t json-module -w commonjs -o bundle.js file1.proto file2.proto
$> pbjs -t static-module file1.proto file2.proto | pbts -o bundle.d.ts -
While using .proto files directly requires the full library respectively pure reflection/JSON the light library, pretty much all code but the relatively short descriptors is shared.
Static code, on the other hand, requires just the minimal library, but generates additional, albeit editable, source code without any reflection features.
There is no significant difference performance-wise as the code generated statically is pretty much the same as generated at runtime and both are largely interchangeable as seen in the previous section.
Both utilities can be used programmatically by providing command line arguments and a callback to their respective main
functions:
var pbjs = require("protobufjs/cli/pbjs");
pbjs.main([ "--target", "json-module", "path/to/myproto.proto" ], function(err, output) {
if (err)
throw err;
// do something with output
});
The package includes a benchmark that tries to compare performance to native JSON as far as this is possible. On an i7-2600K running node 6.9.1 it yields:
benchmarking encoding performance ...
Type.encode to buffer x 547,361 ops/sec ±0.27% (94 runs sampled)
JSON.stringify to string x 310,848 ops/sec ±0.73% (92 runs sampled)
JSON.stringify to buffer x 173,608 ops/sec ±1.51% (86 runs sampled)
Type.encode to buffer was fastest
JSON.stringify to string was 43.5% slower
JSON.stringify to buffer was 68.7% slower
benchmarking decoding performance ...
Type.decode from buffer x 1,294,378 ops/sec ±0.86% (90 runs sampled)
JSON.parse from string x 291,944 ops/sec ±0.72% (92 runs sampled)
JSON.parse from buffer x 256,325 ops/sec ±1.50% (90 runs sampled)
Type.decode from buffer was fastest
JSON.parse from string was 77.4% slower
JSON.parse from buffer was 80.3% slower
benchmarking combined performance ...
Type to/from buffer x 254,126 ops/sec ±1.13% (91 runs sampled)
JSON to/from string x 122,896 ops/sec ±1.29% (90 runs sampled)
JSON to/from buffer x 88,005 ops/sec ±0.87% (89 runs sampled)
Type to/from buffer was fastest
JSON to/from string was 51.7% slower
JSON to/from buffer was 65.3% slower
benchmarking verifying performance ...
Type.verify x 6,246,765 ops/sec ±2.00% (87 runs sampled)
benchmarking message from object performance ...
Type.fromObject x 2,892,973 ops/sec ±0.70% (92 runs sampled)
benchmarking message to object performance ...
Type.toObject x 3,601,738 ops/sec ±0.72% (93 runs sampled)
Note that JSON is a native binding nowadays and as such is about as fast as it possibly can get. So, how can protobuf.js be faster?
- The benchmark is somewhat flawed.
- Reader and writer interfaces configure themselves according to the environment to eliminate redundant conditionals.
- Node-specific reader and writer subclasses benefit from node's buffer binding.
- Reflection has built-in code generation that builds type-specific encoders, decoders and verifiers at runtime.
- Encoders and decoders do not implicitly call
verify
on messages to avoid unnecessary overhead where messages are already known to be valid. It's up to the user to callverify
where necessary. - Quite a bit of V8-specific profiling is accountable for everything else.
You can also run the benchmark ...
$> npm run bench
and the profiler yourself (the latter requires a recent version of node):
$> npm run prof <encode|decode|encode-browser|decode-browser> [iterations=10000000]
Note that as of this writing, the benchmark suite performs significantly slower on node 7.2.0 compared to 6.9.1 because moths.
- Because the internals of this package do not rely on
google/protobuf/descriptor.proto
, options are parsed and presented literally. - If typed arrays are not supported by the environment, plain arrays will be used instead.
- Support for pre-ES5 environments (except IE8) can be achieved by using a polyfill.
- Support for Content Security Policy-restricted environments (like Chrome extensions without unsafe-eval) can be achieved by generating and using static code instead.
- If you need a proper way to work with 64 bit values (uint64, int64 etc.), you can install long.js alongside this library. All 64 bit numbers will then be returned as a
Long
instance instead of a possibly unsafe JavaScript number (see).
To build the library or its components yourself, clone it from GitHub and install the development dependencies:
$> git clone https://github.com/dcodeIO/protobuf.js.git
$> cd protobuf.js
$> npm install
Building the respective development and production versions with their respective source maps to dist/
:
$> npm run build
Building the documentation to docs/
:
$> npm run docs
Building the TypeScript definition to index.d.ts
:
$> npm run types
By default, protobuf.js integrates into your browserify build-process without requiring any optional modules. Hence:
-
If you need int64 support, explicitly require the
long
module somewhere in your project as it will be excluded otherwise. This assumes that a globalrequire
function is present that protobuf.js can call to obtain the long module.If there is no global
require
function present after bundling, it's also possible to assign the long module programmatically:var Long = ...; protobuf.util.Long = Long; protobuf.configure();
-
If you have any special requirements, there is the bundler for reference.
License: BSD 3-Clause License