/gRPCClient.jl

A Julia gRPC Client

Primary LanguageJuliaOtherNOASSERTION

gRPCClient.jl

A Julia gRPC Client.

Build Status codecov.io

Generating gRPC Service Client

gRPC services are declared in .proto files. Use gRPCClient.generate to generate client code from specification files.

gRPC code generation uses protoc and the ProtoBuf.jl package. To be able to generate gRPC client code, ProtoBuf package must be installed along with gRPCClient.

The protoc file must have service generation turned on for at least one of C++, python or Java, e.g. one of:

option cc_generic_services = true;
option py_generic_services = true;
option java_generic_services = true;

The Julia code generated can be improper if the package name declared in the proto specification has .. Set a suitable package name without ..

julia> using Pkg

julia> Pkg.add("ProtoBuf")
...
  Installed ProtoBuf ──── v0.11.0
Downloading artifact: protoc
...
julia> Pkg.add("gRPCClient")
...
julia> # or Pkg.develop(PackageSpec(url="https://github.com/JuliaComputing/gRPCClient.jl"))

julia> using gRPCClient

julia> gRPCClient.generate("route_guide.proto")
┌ Info: Generating gRPC client
│   proto = "RouteguideClients/route_guide.proto"
└   outdir = "RouteguideClients"
┌ Info: Detected
│   package = "routeguide"
└   service = "RouteGuide"
┌ Info: Generated
└   outdir = "RouteguideClients"

The generated code can either be published as a package or included and used as a module.

julia> using gRPCClient

julia> include("RouteguideClients/RouteguideClients.jl");

julia> using .RouteguideClients

julia> import .RouteguideClients: Point, Feature, GetFeature

julia> Base.show(io::IO, location::Point) =
    print(io, string("[", location.latitude, ", ", location.longitude, "]"))

julia> Base.show(io::IO, feature::Feature) =
    print(io, string(feature.name, " - ", feature.location))

julia> client = RouteGuideBlockingClient("https://server:10000/");

julia> point = Point(; latitude=409146138, longitude=-746188906); # request param

julia> feature, status_future = GetFeature(client, point);

julia> gRPCCheck(status_future) # check status of request
true

julia> feature # this is the API return value
Berkshire Valley Management Area Trail, Jefferson, NJ, USA - [409146138, -746188906]

The generated module is named after the package declared in the proto file. And for each service, a pair of clients are generated in the form of <service_name>Client and <service_name>BlockingClient.

The service methods generated for <service_name>Client are identical to the ones generated for <service_name>BlockingClient, except that they spawn off the actual call into a task and accept a callback method that is invoked with the results. The <service_name>BlockingClient may however be more intuitive to use.

Each service method returns (or calls back with, in the case of non-blocking clients) two values:

  • The result, which can be a Julia struct or a Channel for streaming output.
  • And, the gRPC status.

The gRPCCheck method checks the status for success or failure. Note that for methods with streams as input or output, the gRPC status will not be ready until the method completes. So the status check and stream use must be done in separate tasks. E.g.:

@sync begin
   in_channel = Channel{RouteguideClients.RouteNote}(1)
   @async begin
      # send inputs
      for input in inputs
         put!(in_channel, input)
      end
      close(in_channel)
   end
   out_channel, status_future = RouteguideClients.RouteChat(client, in_channel)
   @async begin
      # consume outputs
      for output in out_channel
         # use output
      end
   end
   @async begin
      gRPCCheck(status_future)
   end
end

APIs and Implementation Details

The generated gRPC client (RouteGuideBlockingClient in the example above) uses a gRPC controller and channel behind the scenes to communicate with the server.

gRPCController

A gRPCController contains settings to control the behavior of gRPC requests. Each gRPC client holds an instance of the controller created using keyword arguments passed to its constructor.

gRPCController(;
    [ maxage::Int = 0, ]
    [ keepalive::Int64 = 60, ]
    [ negotiation::Symbol = :http2_prior_knowledge, ]
    [ revocation::Bool = true, ]
    [ request_timeout::Real = Inf, ]
    [ connect_timeout::Real = 0, ]
    [ max_message_length = DEFAULT_MAX_MESSAGE_LENGTH, ]
    [ max_recv_message_length = 0, ]
    [ max_send_message_length = 0, ]
    [ enable_shared_locks = false, ]
    [ verbose::Bool = false, ]
)
  • maxage: maximum age (seconds) of a connection beyond which it will not be reused (default 180 seconds, same as setting this to 0).
  • keepalive: interval (seconds) in which to send TCP keepalive messages on the connection (default 60 seconds).
  • negotiation: how to negotiate HTTP2, can be one of :http2_prior_knowledge (no negotiation, the default), :http2_tls (http2 upgrade but only over tls), or :http2 (http2 upgrade)
  • revocation: whether to check for certificate recovation (default is true)
  • request_timeout: request timeout (seconds)
  • connect_timeout: connect timeout (seconds) (default is 300 seconds, same as setting this to 0)
  • max_message_length: maximum message length (default is 4MB)
  • max_recv_message_length: maximum message length to receive (default is max_message_length, same as setting this to 0)
  • max_send_message_length: maximum message length to send (default is max_message_length, same as setting this to 0)
  • enable_shared_locks: whether to enable locks for using gRPCClient across tasks/threads concurrently (experimental, default is false)
  • verbose: whether to print out verbose communication logs (default false)

gRPCChannel

gRPCChannel(baseurl::String)

gRPCChannel represents a connection to a specific service endpoint (service baseurl) of a gRPC server.

A channel also usually has a single network connection backing it and multiple streams of requests can flow through it at any time. The number of streams that can be multiplexed is negotiated between the client and the server.

gRPCStatus

gRPCStatus represents the status of a request. It has the following fields:

  • success: whether the request was completed successfully.
  • grpc_status: the grpc status code returned
  • message: any error message if request was not successful

gRPCCheck

gRPCCheck(status; throw_error::Bool=true)

Method to check the response of a gRPC request and raise a gRPCException if it has failed. If throw_error is set to false, returns true or false indicating success instead.

gRPCException

Every gRPC request returns the result and a future representing the status of the gRPC request. Use the gRPCCheck method on the status future to check the request status and throw a gRPCException if it is not successful.

The abstract gRPCException type has the following concrete implementations:

  • gRPCMessageTooLargeException
  • gRPCServiceCallException

gRPCMessageTooLargeException

A gRPMessageTooLargeException exception is thrown when a message is encountered that has a size greater than the limit configured. Specifically, max_recv_message_length while receiving and max_send_message_length while sending.

A gRPMessageTooLargeException has the following members:

  • limit: the limit value that was exceeded
  • encountered: the amount of data that was actually received or sent before this error was triggered. Note that this may not correspond to the full size of the data, as error may be thrown before actually materializing the complete data.

gRPCServiceCallException

A gRPCServiceCallException is thrown if a gRPC request is not successful. It has the following members:

  • grpc_status: grpc status code for this request
  • message: any error message if request was not successful

TODO

  • Concurrent use of gRPCClient is still somewhat flaky.

Credits

This package was originally developed at Julia Computing