This repository contains support for various Docker images that wrap protoc
,
prototool
, grpc_cli
commands with gRPC support
in a variety of languages removing the need to install and manage these commands locally.
It relies on setting a simple volume to the docker container,
usually mapping the current directory to /defs
, and specifying the file and
language you want to generate.
- Docker images for:
protoc
withnamely/protoc
(automatically includes/usr/local/include
)- Uber's Prototool with
namely/prototool
- A custom generation script to facilitate common use-cases with
namely/protoc-all
(see below) grpc_cli
withnamely/grpc-cli
- gRPC Gateway using a custom go-based server with
namely/gen-grpc-gateway
- Google APIs included in
/opt/include/google
- Protobuf library artificats included in
/opt/include/google/protobuf
(NOTE:protoc
would only need part of the path i.e.-I /opt/include
if you import WKTs like so:
import "google/protobuf/empty.proto";
...
- Support for all C based gRPC libraries with Go and Java native libraries
If you're having trouble, see Docker troubleshooting below.
Note - throughout this document, commands for bash are prefixed with
$
and commands for PowerShell on Windows are prefixed withPS>
. It is not required to use "Windows Subsystem for Linux" (WSL) except for development work on docker-protoc itself
For protoc
, grpc_cli
and prototool
a pattern of <GRPC\_VERSION>_<CONTAINER\_VERSION>
is used for all images.
Example is namely/protoc-all:1.15_0
for gRPC version 1.15
. The latest
tag will always point to the most recent version.
Pull the container:
$ docker pull namely/protoc-all
After that, change working directory to the one that contains your .proto
definition
files.
So if you have a directory: ~/my_project/protobufs/
that has: myproto.proto
, you'd want to run this:
$ cd ~/my_project/protobufs
$ docker run -v `pwd`:/defs namely/protoc-all -f myproto.proto -l ruby #or go, csharp, etc
PS> cd ~/my_project/protobufs
PS> docker run -v ${pwd}:/defs namely/protoc-all -f myproto.proto -l ruby #or go, csharp, etc
The container automatically puts the compiled files into a gen
directory with
language-specific sub-directories. So for Golang, the files go into a directory
./gen/pb-go
; For ruby the directory is ./gen/pb-ruby
.
You can use the -o
flag to specify an output directory. This will
automatically be created. For example, add -o my-gen
to add all fileoutput to
the my-gen
directory. In this case, pb-*
subdirectories will not be created.
You can use the -d
flag to generate all proto files in a directory. You cannot
use this with the -f
option.
You can also use -i
to add extra include directories. This can be helpful to
lift protofiles up a directory when generating. As an example, say you have a
file protorepo/catalog/catalog.proto
. This will by default output to
gen/pb-go/protorepo/catalog/
because protorepo
is part of the file path
input. To remove the protorepo
you need to add an include and change the
import:
$ docker run ... namely/protoc-all -i protorepo -f catalog/catalog.proto -l go
# instead of
$ docker run ... namely/protoc-all -f protorepo/catalog/catalog.proto -l go
# which will generate files in a `protorepo` directory.
--with-rbi
to generate Ruby Sorbet type definition .rbi files
--js-out <string>
to modify the js_out=
options for node and web code generation
--grpc-web-out <string>
to modify the grpc-web_out=
options for web code generation
--grpc-out <string>
to modify the grpc_out=
options for node and web code generation. See https://www.npmjs.com/package/grpc-tools for more details.
This repo also provides a docker image namely/gen-grpc-gateway
to generate a
grpc-gateway server.
By annotating your proto (see the grpc-gateway documentation), you can generate a
server that acts as an HTTP server, and a gRPC client to your gRPC service.
Generate a gRPC Gateway docker project with
docker run -v `pwd`:/defs namely/gen-grpc-gateway -f path/to/your/proto.proto -s Service
where Service
is the name of your gRPC service defined in the proto. This will create a
folder with a simple go server.
By default, this goes in the gen/grpc-gateway
folder. You can then build the contents of this
folder into an actual runnable grpc-gateway server.
Build your gRPC Gateway server with
docker build -t my-grpc-gateway gen/grpc-gateway/
NOTE: If your service does not contain any (google.api.http)
annotations, this build will
fail with an error ...HandlerFromEndpoint is undefined
. You need to have at least one rpc
method annotated to build a gRPC Gateway.
Run this image with
docker run my-grpc-gateway --backend=grpc-service:50051
where --backend
refers to your actual gRPC server's address. The gRPC gateway
listens on port 80
for HTTP traffic.
The gateway is configured using spf13/viper, see gwy/templates/config.yaml.tmpl for configuration options.
To configure your gateway to run under a prefix, set proxy.api-prefix to that prefix. For example, if you have (google.api.http) = '/foo/bar'
, and set proxy.api-prefix
to /api/'
, your gateway will listen to requests on '/api/foo/bar'
. This can also be set with the environment variable <SERVICE>_PROXY_API-PREFIX
where <SERVICE>
is the name of the service generating the gateway.
See gwy/test.sh for an example of how to set the prefix with an environment variable.
The gateway will turn any HTTP headers that it receives into gRPC metadata. Any
permanent HTTP headers
will be prefixed with grpcgateway-
in the metadata, so that your server receives both
the HTTP client to gateway headers, as well as the gateway to gRPC server headers.
Any headers starting with Grpc-
will be prefixed with an X-
, this is because grpc-
is a reserved metadata prefix.
All other headers will be converted to metadata as is.
You can configure CORS for your gateway through the configuration. This will allow your gateway to receive requests from different origins.
There are four values:
-
cors.allow-origin
: Value to set for Access-Control-Allow-Origin header. -
cors.allow-credentials
: Value to set for Access-Control-Allow-Credentials header. -
cors.allow-methods
: Value to set for Access-Control-Allow-Methods header. -
cors.allow-headers
: Value to set for Access-Control-Allow-Headers header.For CORS, you will want to configure your
cors.allow-methods
to be the HTTP verbs set in your proto (i.e.GET
,PUT
, etc.), as well asOPTIONS
, so that your service can handle the preflight request.If you are not using CORS, you can leave these configuration values at their default, and your gateway will not accept CORS requests.
-
grpc.max-call-recv-msg-size
: Sets the maximum message size in bytes the client can receive. -
grpc.max-call-send-msg-size
: Sets the maximum message size in bytes the client can send.
You can configure additional headers to be sent in the HTTP response.
Set environment variable with prefix <SERVICE>_RESPONSE-HEADERS_
(e.g SOMESERVICE_RESPONSE-HEADERS_SOME-HEADER-KEY
).
You can also set headers in the your configuration file (e.g response-headers.some-header-key
)
By default, gen-grpc-gateway
will use a marshaler/unmarshaler based on jsonpb. You can change this behavior by setting gateway.use-jsonpb-v2-marshaler: true
, which will use protojson - a newer version which is more aligned with proto <=> json mapping.
By default, gen-grpc-gateway
will return proto names as they are in the proto messages. You can change this behavior by setting gateway.use-json-names: true
and the gateway will use camelCase JSON names.
By default, gen-grpc-gateway
will not emit unpopulated fields. You can change this behavior by setting gateway.emit-unpopulated: true
and the gateway will populate these fields with default values.
By default, gen-grpc-gateway
will discard unknown fields from requests. You can change this behavior by setting gateway.keep-unknown: true
and the gateway will keep these fields in the requests.
The gateway project used spf13/viper for configuration. The generated gateway code includes a config file that can be overridden with cli flags or environment variables. For environment variable overrides use a <SERVICE>_
prefix, upcase the setting, and replace .
with _
.
This repo also contains a Dockerfile for building a grpc_cli
.
Run it with
docker run -v `pwd`:/defs --rm -it namely/grpc-cli call docker.for.mac.localhost:50051 \\
LinkShortener.ResolveShortLink "short_link:'asdf'" --protofiles=link_shortener.proto
You can pass multiple files to --protofiles
by separating them with commas, for example
--protofiles=link_shortener.proto,foo/bar/baz.proto,biz.proto
. All of the protofiles
must be relative to pwd
, since pwd
is mounted into the container.
See the grpc_cli documentation for more information. You may find it useful to bind this to an alias:
alias grpc_cli='docker run -v `pwd`:/defs --rm -it namely/grpc-cli'
Note the use of single quotes in the alias, which avoids expanding the pwd
parameter when the alias
is created.
Now you can call it with
grpc_cli call docker.for.mac.localhost:50051 LinkShortener.ResolveShortLink "short_link:'asdf'" --protofiles=link_shortener.proto
Thank you for considering a contribution to namely/docker-protoc!
If you'd like to make an enhancement, or add a container for another language compiler, you will need to run one of the build scripts in this repo. You will also need to be running Mac, Linux, or WSL 2, and have Docker installed. From the repository root, run this command to build all the known containers:
$ make build
Note the version tag in Docker's console output - this image tag is required to run the tests using the container with your changes.
You can change some environment variables relevant to the build by setting them as prefixes to the make command. For example, this would build the containers using Node.js 15 and gRPC 1.35. See some interesting variables in variables.sh and entrypoint.sh.
$ NODE_VERSION=15 GRPC_VERSION=1.35 make build
To run the tests, identify your image tag from the build step and run make test
as below:
$ CONTAINER=namely/protoc-all:VVV make test
(VVV
is your version from the tag in the console output when running make build
). Running this will
demonstrate that your new image can successfully build containers for each language.
Open a PR and ping one of the Namely employees who have worked on this repo recently. We will take a look as soon as we can. Thank you!!
Namely employees can merge PRs and the latest version will be pushed up via CI. It is also possible to do this manually by running this:
$ make push
This will build and push the containers to the Namely registry located on DockerHub. You must be authorized to push to this repo.
Docker must be configured to use Linux containers.
If on Windows, you must have your C:
drive shared with Docker. Open the Docker settings (right-click Docker icon in notification area) and pick the Shared Drives tab. Ensure C:
is listed and the box is checked. If you are still experiencing trouble, click "Reset credentials..." on that tab and re-enter your local Windows username and password.