Polyglot is a grpc client which can talk to any grpc server. In order to make a call, the following are required:
- A compiled Polyglot binary,
- the .proto files for the service or grpc reflection enabled on the remote server,
- and a request proto instance in text format.
In particular, it is not necessary to generate grpc classes for the service or to compile the protos into the Polyglot binary.
- Supports unary, client streaming, server streaming, and bidi streaming rpcs.
- Runs on Windows, Mac and Linux.
- Parses proto files at runtime to discover services. Supports pretty-printing discovered services.
- Can discover services by reflection if the remote server has reflection enabled
- Supports authentication via oauth.
- Accepts request protos through stdin and can output responses to stdout to allow chaining.
- Supports plain text connections as well as TLS.
- Supports passing custom grpc metadata over the command line.
- Supports all protobuf well-known-types, including fields of type "Any".
All you need to run Polyglot is a Java runtime. Binaries for Mac, Linux, and Windows are available from the releases page.
The "Hello World" of using Polyglot is to make an rpc call. This can be done using call
command as follows:
$ echo <json-request> | java -jar polyglot.jar \
call \
--endpoint=<host>:<port> \
--full_method=<some.package.Service/doSomething> \
--proto_discovery_root=<path>
For stream requests double newlines \n\n
are used to separate your json requests as follows:
$ echo '<json-request-1> \n\n <json-request-2> ... \n\n <json-request-n>' | java -jar polyglot.jar \
call \
--endpoint=<host>:<port> \
--full_method=<some.package.Service/doSomething> \
--proto_discovery_root=<path>
For more invocation examples, see the examples directory.
If the remote server has reflection enabled, there is no need to pass the proto files to Polyglot. The example invocation above then becomes:
$ echo <json-request> | java -jar polyglot.jar \
call \
--endpoint=<host>:<port> \
--full_method=<some.package.Service/doSomething>
Bu default, Polyglot always tries to use reflection before compiling local protos. Reflection can be turned off explicitly by setting the flag --use_reflection=false
.
Some of the features of Polyglot (such as Oauth, see below) require some configuration. Moreover, that sort of configuration tends to remain identical across multiple Polyglot runs. In order to improve usability, Polyglot supports loading a configuration set from a file at runtime. This configuration set can contain multiple named Configuration
objects (schema defined here). An example configuration could look like this:
{
"configurations": [
{
"name": "production",
"call_config": {
"use_tls": "true",
"oauth_config": {
"refresh_token_credentials": {
"token_endpoint_url": "https://auth.example.com/token",
"client": {
"id": "example_id",
"secret": "example_secret"
},
"refresh_token_path": "/path/to/refresh/token"
}
}
},
"proto_config": {
"proto_discovery_root": "/home/dave/protos",
"include_paths": [
"/home/dave/lib"
]
}
},
{
"name": "staging",
"call_config": {
"oauth_config": {
"refresh_token_credentials": {
"token_endpoint_url": "https://auth-staging.example.com/token"
}
}
},
"proto_config": {
"proto_discovery_root": "/home/dave/staging/protos",
"include_paths": [
"/home/dave/staging/lib"
]
}
}
]
}
By default, Polyglot tries to find a config file at $HOME/.polyglot/config.pb.json
, but this can be overridden with the --config_set_path
flag. By default, Polyglot uses the first configuration in the set, but this can be overridden with the --config_name
flag.
The general philosophy is for the configuration to drive Polyglot's behavior and for command line flags to allow selectively overriding parts of the configuration. For a full list of what can be configured, please see config.proto
.
Polyglot uses statically linked boringssl libraries under the hood and doesn't require the host machine to have any specific libraries. Whether or not the client uses TLS to talk to the server can be controlled using the --use_tls
flag or the corresponding configuration entry.
Polyglot can also do client certificate authentication with the --tls_client_cert_path
and --tls_client_key_path
flags. If the hostname on the server does not match the endpoint (e.g. connecting
to localhost
, but the server thinks it's foo.example.com
), --tls_client_override_authority=foo.example.com
can be used.
Polyglot has built-in support for authentication of requests using OAuth tokens in two ways:
- Loading an access token from disk and attaching it to the request.
- Loading a refresh token from disk, exchanging it for an access token, and attaching the access token to the request.
In order to use this feature, Polyglot needs an OauthConfiguration
inside its Configuration
. For details on how to populate the OauthConfiguration
, please see the documentation of the fields in config.proto
.
Polyglot supports printing a list of all the discovered services using the list_services
command. This command can be invoked as follows:
$ java -jar polyglot.jar \
list_services \
--proto_discovery_root=<path> \
The printed services can be filtered using --service_filter=<service_name>
or --method_filter=<method_name>
, and the --with_message
flag can be used to also print the exact format of the requests.
It is possible to add custom grpc metadata to calls made using Polyglot by setting the --metadata=key1:value1,key2:value2
flag.
Polyglot uses the slf4j
logging framework with the org.slf4j.impl.SimpleLogger
implementation. It also redirects grpc's JUL logs to slf4j
. To change the default log level, specify the org.slf4j.simpleLogger.defaultLogLevel
JVM argument, e.g., by passing the argument -Dorg.slf4j.simpleLogger.defaultLogLevel=debug
.
In order to build Polyglot from source, you will need:
$ bazel build src/main/java/me/dinowernli/grpc/polyglot
After calling this, you should have a fresh binary at:
./bazel-bin/src/main/java/me/dinowernli/grpc/polyglot
If you would like to build a deployable (fat) jar, run:
$ bazel build src/main/java/me/dinowernli/grpc/polyglot:polyglot_deploy.jar
Example invocations can be found in the examples directory. In order to run a simple rpc call, invoke run-server.sh
followed by (in a different terminal) call-command-example.sh
.
$ bazel test //src/...