rules_protobuf was initially written when the bazel protobuf ecosystem was fairly immature. Now 2 years later, this repository is showing its age. Rather than retrofit this set of rules, it's been re-written from the ground-up to work correctly with the native.proto_library rule and is available at https://github.com/stackb/rules_proto. Consequently, these rules are effectively no longer being maintained.

Please upgrade to the newer rules when appropriate and file issues if there are feature gaps that need to be filled.

Cheers, @pcj (Oct 25 2018)

https://github.com/stackb/rules_proto

Bazel skylark rules for building protocol buffers with +/- gRPC support on (osx, linux) ✨.

Bazel	rules_protobuf	gRPC

These rules sprung out of a need to have protobuf support when there was limited exposed and documented proto generation capabilities in the main bazel repository. This is a moving target. The main goals of this project are to:

1. Provide protoc, the protocol buffer compiler (v3.5.1).

2. Provide the language-specific plugins.

3. Provide the necessary libraries and dependencies for gRPC support, when possible.

4. Provide an extensible proto_language abstraction (used in conjunction with the proto_compile rule) to generate outputs for current and future custom protoc plugins not explicitly provided here.

Refer to DEPENDENCIES.md for a more detailed summary of workspace dependencies / versions.

1. Support for generation of protoc outputs via proto_compile() rule.

2. Support for generation + compilation of outputs with protobuf dependencies.

3. gRPC support.

4. Highly experimental (probably not functional yet). A work-in-progress for those interested in contributing further work.

If you have not already installed bazel on your workstation, follow the bazel instructions.

NOTE: Bazel 0.8.0 or above is required for go support.

Specify the language(s) you'd like use by loading the language-specific *_proto_repositories rule(s):

git_repository(
  name = "org_pubref_rules_protobuf",
  remote = "https://github.com/pubref/rules_protobuf",
  tag = "v0.8.2",
  #commit = "..." # alternatively, use latest commit on master
)

load("@org_pubref_rules_protobuf//java:rules.bzl", "java_proto_repositories")
java_proto_repositories()

load("@org_pubref_rules_protobuf//cpp:rules.bzl", "cpp_proto_repositories")
cpp_proto_repositories()

load("@org_pubref_rules_protobuf//go:rules.bzl", "go_proto_repositories")
go_proto_repositories()

Several languages have other rules_* dependencies that you'll need to load before the *_proto_repositories() function is invoked:

¹ Only needed for python grpc support.

To build a java-based gRPC library:

load("@org_pubref_rules_protobuf//java:rules.bzl", "java_proto_library")

java_proto_library(
  name = "protolib",
  protos = [
    "my.proto"
  ],
  with_grpc = True,
  verbose = 1, # 0=no output, 1=show protoc command, 2+ more...
)

To run the examples & tests in this repository, clone it to your workstation.

# Clone this repo
$ git clone https://github.com/pubref/rules_protobuf

# Go to examples/helloworld directory
$ cd rules_protobuf/examples/helloworld

# Run all tests
$ bazel test examples/...

# Build a server
$ bazel build cpp/server

# Run a server from the command-line
$ $(bazel info bazel-bin)/examples/helloworld/cpp/server

# Run a client
$ bazel run go/client
$ bazel run cpp/client
$ bazel run java/org/pubref/rules_protobuf/examples/helloworld/client:netty

To load alternate versions of dependencies, pass in a dict having the same overall structure of a deps.bzl file. Entries having a matching key will override those found in the file. For example, to load a different version of https://github.com/golang/protobuf, provide a different commit ID:

load("@org_pubref_rules_protobuf//go:rules.bzl", "go_proto_repositories")
go_proto_repositories(
  overrides = {
    "com_github_golang_protobuf": {
      # Override golang with a different commit
      "commit": "2c1988e8c18d14b142c0b472624f71647cf39adb",
    }
  },
)

You may already have some external dependencies already present in your workspace that rules_protobuf will attempt to load, causing a collision. To prevent rules_protobuf from loading specific external workspaces, name them in the excludes list:

go_proto_repositories(
  excludes = [
    "com_github_golang_glog",
  ]
)

To completely replace the set of dependencies that will attempt to be loaded, you can pass in a full dict object to the lang_deps attribute.

go_proto_repositories(
  lang_deps = {
    "com_github_golang_glog": {
      ...
    },
  },
)

There are several language --> language dependencies as well. For example, python_proto_repositories and ruby_proto_repositories (and more) internally call the cpp_proto_repositories rule to provide the grpc plugins. To suppress this (and have better control in your workspace), you can use the omit_cpp_repositories=True option.

Use the proto_deps attribute to name proto rule dependencies. Use of proto_deps implies you're using imports, so read on...

In all cases, these rules will include a --proto_path=. (-I.) argument. This is functionally equivalent to --proto_path=$(bazel info execution_root). Therefore, when the protoc tool is invoked, it will 'see' whatever directory structure exists at the bazel execution root for your workspace. To better learn what this looks like, cd $(bazel info execution_root) and look around. In general, it contains all your sourcefiles as they appear in your workspace with an additional external/WORKSPACE_NAME directory for all dependencies used.

This has implications for import statements in your protobuf sourcefiles, if you use them. The two cases to consider are imports within your workspace (referred to here as 'internal' imports), and imports of other protobuf files in an external workspace (external imports).

Internal imports should require no additional parameters if your import statements follow the same directory structure of your workspace. For example, the examples/helloworld/proto/helloworld.proto file imports the examples/proto/common.proto file. Since this matches the workspace directory structure, protoc can find it, and no additional arguments to a cc_proto_library are required for protoc code generation step.

Obviously, importing a file does not mean that code will be generated for it. Therefore, use of the imports attribute implies that the generated files for the imported message or service already exist somewhere that can be used as a dependency some other library rule (such as srcs for java_library).

Rather than using imports, it often make more sense to declare a dependency on another proto_library rule via the proto_deps attribute. This makes the import available to the calling rule and performs the code generation step. For example, the cc_proto_library rule in examples/helloworld/proto:cpp names the //examples/proto:cpp's cc_proto_library rule in its proto_deps attribute to accomplish both code generation and compilation of object files for the proto chain.

The same logic applies to external imports. The two questions to ask yourself when setting up your rules are:

[Question 1]: Can protoc "see" the imported file? In order to satisfy this requirement, pass in the full path of the required file(s) relative to the execution root where protoc will be run. For example, the well-known descriptor.proto could be made visible to protoc via:

java_proto_library(
  name = 'fooprotos',
  protos = 'foo.proto`,
  imports = [
      "external/com_google_protobuf/src/",
    ],
    inputs = [
      "@com_google_protobuf//:well_known_protos",
    ],
)

This would be imported as import "google/protobuf/descriptor.proto" given that the file @com_google_protobuf/src/google/protobuf/descriptor.proto is in the package google.protobuf.

[Question 2]: Can the cc_proto_library rule "see" the generated protobuf files? (in this case descriptor.pb.{h,cc}. Just because the file was imported does not imply that protoc will generate outputs for it, so somewhere in the cc_library rule dependency chain these files must be present. This could be via another cc_proto_library rule defined elswhere, or a some other filegroup or label list. If the source is another cc_proto_library rule, specify that in the proto_deps attribute to the calling cc_proto_library rule. Otherwise, pass a label that includes the (pregenerated) protobuf files to the deps attribute, just as you would any typical cc_library rule.

Important note about sandboxing: simply stating the path where protoc should look for imports (via the imports attribute) is not enough to work with the bazel sandbox. Bazel is very particular about needing to know exactly which inputs are required for a rule, and exactly what output files it generates. If an input is not declared, it will not be exposed in the sandbox. Therefore, we have to provide both the import path and a label-generating rule in the inputs attribute that names the files we want available in the sandbox (given here by :well_known_protos).

If you are having problems, put verbose={1,2,3} in your build rule and/or disable sandboxing with --spawn_strategy=standalone.

Contributions welcome; please create Issues or GitHub pull requests.

• Much thanks to all contributors and the members of the bazel, protobuf, and gRPC teams.