vtprotobuf, the Vitess Protocol Buffers compiler

This repository provides the protoc-gen-go-vtproto plug-in for protoc, which is used by Vitess to generate optimized marshall & unmarshal code.

The code generated by this compiler is based on the optimized code generated by gogo/protobuf, although this package is not a fork of the original gogo compiler, as it has been implemented to support the new ProtoBuf APIv2 packages.

Available features

vtprotobuf is implemented as a helper plug-in that must be run alongside the upstream protoc-gen-go generator, as it generates fully-compatible auxiliary code to speed up (de)serialization of Protocol Buffer messages.

The following features can be generated:

• size: generates a func (p *YourProto) SizeVT() int helper that behaves identically to calling proto.Size(p) on the message, except the size calculation is fully unrolled and does not use reflection. This helper function can be used directly, and it'll also be used by the marshal codegen to ensure the destination buffer is properly sized before ProtoBuf objects are marshalled to it.

• marshal: generates the following helper methods

  • func (p *YourProto) MarshalVT() ([]byte, error): this function behaves identically to calling proto.Marshal(p), except the actual marshalling has been fully unrolled and does not use reflection or allocate memory. This function simply allocates a properly sized buffer by calling SizeVT on the message and then uses MarshalToSizedBufferVT to marshal to it.

  • func (p *YourProto) MarshalToVT(data []byte) (int, error): this function can be used to marshal a message to an existing buffer. The buffer must be large enough to hold the marshalled message, otherwise this function will panic. It returns the number of bytes marshalled. This function is useful e.g. when using memory pooling to re-use serialization buffers.

  • func (p *YourProto) MarshalToSizedBufferVT(data []byte) (int, error): this function behaves like MarshalTo but expects that the input buffer has the exact size required to hold the message, otherwise it will panic.

• unmarshal: generates a func (p *YourProto) UnmarshalVT(data []byte) that behaves similarly to calling proto.Unmarshal(data, p) on the message, except the unmarshalling is performed by unrolled codegen without using reflection and allocating as little memory as possible. If the receiver p is not fully zeroed-out, the unmarshal call will actually behave like proto.Merge(data, p). This is because the proto.Unmarshal in the ProtoBuf API is implemented by resetting the destionation message and then calling proto.Merge on it. To ensure proper Unmarshal semantics, ensure you've called proto.Reset on your message before calling UnmarshalVT, or that your message has been newly allocated.

• pool: generates the following helper methods

  • func (p *YourProto) ResetVT(): this function behaves similarly to proto.Reset(p), except it keeps as much memory as possible available on the message, so that further calls to UnmarshalVT on the same message will need to allocate less memory. This an API meant to be used with memory pools and does not need to be used directly.

  • func (p *YourProto) ReturnToVTPool(): this function returns message p to a local memory pool so it can be reused later. It clears the object properly with ResetVT before storing it on the pool. This method should only be used on messages that were obtained from a memory pool by calling YourProtoFromVTPool. Using p after calling this method will lead to undefined behavior.

  • func YourProtoFromVTPool() *YourProto: this function returns a YourProto message from a local memory pool, or allocates a new one if the pool is currently empty. The returned message is always empty and ready to be used (e.g. by calling UnmarshalVT on it). Once the message has been processed, it must be returned to the memory pool by calling ReturnToVTPool() on it. Returning the message to the pool is not mandatory (it does not leak memory), but if you don't return it, that defeats the whole point of memory pooling.

Usage

1. Install protoc-gen-go-vtproto:

go install github.com/planetscale/vtprotobuf/cmd/protoc-gen-go-vtproto

2. Ensure your project is already using the ProtoBuf v2 API (i.e. google.golang.org/protobuf). The vtprotobuf compiler is not compatible with APIv1 generated code.

3. Update your protoc generator to use the new plug-in. Example from Vitess:

for name in $(PROTO_SRC_NAMES); do \
    $(VTROOT)/bin/protoc \
    --go_out=. --plugin protoc-gen-go="${GOBIN}/protoc-gen-go" \
    --go-grpc_out=. --plugin protoc-gen-go-grpc="${GOBIN}/protoc-gen-go-grpc" \
    --go-vtproto_out=. --plugin protoc-gen-go-vtproto="${GOBIN}/protoc-gen-go-vtproto" \
    --go-vtproto_opt=features=marshal+unmarshal+size \
    proto/$${name}.proto; \
done

Note that the vtproto compiler runs like an auxiliary plug-in to the protoc-gen-go in APIv2, just like the new GRPC compiler plug-in, protoc-gen-go-grpc. You need to run it alongside the upstream generator, not as a replacement.

4. (Optional) Pass the features that you want to generate as --go-vtproto_opt. If no features are given, all the codegen steps will be performed.

5. Compile the .proto files in your project. You should see _vtproto.pb.go files next to the .pb.go and _grpc.pb.go files that were already being generated.

6. (Optional) Switch your RPC framework to use the optimized helpers (see following sections)

Using the optimized code with RPC frameworks

The protoc-gen-go-vtproto compiler does not overwrite any of the default marshalling or unmarshalling code for your ProtoBuf objects. Instead, it generates helper methods that can be called explicitly to opt-in to faster (de)serialization.

vtprotobuf with GRPC

To use vtprotobuf with the new versions of GRPC, you need to register the codec provided by the github.com/planetscale/vtprotobuf/codec/grpc package.

package servenv

import (
    "github.com/planetscale/vtprotobuf/codec/grpc"
	"google.golang.org/grpc/encoding"
	_ "google.golang.org/grpc/encoding/proto"
)

func init() {
	encoding.RegisterCodec(grpc.Codec{})
}

Note that we perform a blank import _ "google.golang.org/grpc/encoding/proto" of the default proto coded that ships with GRPC to ensure it's being replaced by us afterwards. The provided Codec will serialize & deserialize all ProtoBuf messages using the optimized codegen.

Mixing ProtoBuf implementations with GRPC

If you're running a complex GRPC service, you may need to support serializing ProtoBuf messages from different sources, including from external packages that will not have optimized vtprotobuf marshalling code. This is perfectly doable by implementing a custom codec in your own project that serializes messages based on their type. The Vitess project implements a custom codec to support ProtoBuf messages from Vitess itself and those generated by the etcd API -- you can use it as a reference.

Twirp

Twirp does not support customizing the Marshalling/Unmarshalling codec by default. In order to support vtprotobuf, you'll need to perform a search & replace on the generated Twirp files after running protoc. Here's an example:

for twirp in $${dir}/*.twirp.go; \
do \
  echo 'Updating' $${twirp}; \
  sed -i '' -e 's/respBytes, err := proto.Marshal(respContent)/respBytes, err := respContent.MarshalVT()/g' $${twirp}; \
  sed -i '' -e 's/if err = proto.Unmarshal(buf, reqContent); err != nil {/if err = reqContent.UnmarshalVT(buf); err != nil {/g' $${twirp}; \
done; \

DRPC

To use vtprotobuf as a DRPC encoding, simply pass github.com/planetscale/vtprotobuf/codec/drpc as the protolib flag in your protoc-gen-go-drpc invocation.

Example:

protoc --go_out=. --go-vtproto_out=. --go-drpc_out=. --go-drpc_opt=protolib=github.com/planetscale/vtprotobuf/codec/drpc