protogen, cmd/protoc-gen-go: initial commit

Package protogen provides support for writing protoc plugins.
A "plugin" in this case is a program run by protoc to generate output.

The protoc-gen-go command is a protoc plugin to generate Go code.

cmd/protoc-gen-go/golden_test.go is mostly a straight copy from
the golden test in github.com/golang/protobuf.

Change-Id: I332d0df1e4b60bb8cd926320b8721e16b99a4b71
Reviewed-on: https://go-review.googlesource.com/130175
Reviewed-by: Joe Tsai <thebrokentoaster@gmail.com>
diff --git a/.gitignore b/.gitignore
index 22d0d82..20e54cf 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1 +1,2 @@
 vendor
+cmd/protoc-gen-go/protoc-gen-go
diff --git a/cmd/protoc-gen-go/golden_test.go b/cmd/protoc-gen-go/golden_test.go
new file mode 100644
index 0000000..d0ed8ea
--- /dev/null
+++ b/cmd/protoc-gen-go/golden_test.go
@@ -0,0 +1,145 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package main
+
+import (
+	"bytes"
+	"flag"
+	"go/build"
+	"io/ioutil"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"regexp"
+	"strings"
+	"testing"
+)
+
+// Set --regenerate to regenerate the golden files.
+var regenerate = flag.Bool("regenerate", false, "regenerate golden files")
+
+// When the environment variable RUN_AS_PROTOC_GEN_GO is set, we skip running
+// tests and instead act as protoc-gen-go. This allows the test binary to
+// pass itself to protoc.
+func init() {
+	if os.Getenv("RUN_AS_PROTOC_GEN_GO") != "" {
+		main()
+		os.Exit(0)
+	}
+}
+
+func TestGolden(t *testing.T) {
+	workdir, err := ioutil.TempDir("", "proto-test")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer os.RemoveAll(workdir)
+
+	// Find all the proto files we need to compile. We assume that each directory
+	// contains the files for a single package.
+	supportTypeAliases := hasReleaseTag("go1.9")
+	packages := map[string][]string{}
+	err = filepath.Walk("testdata", func(path string, info os.FileInfo, err error) error {
+		if filepath.Base(path) == "import_public" && !supportTypeAliases {
+			// Public imports require type alias support.
+			return filepath.SkipDir
+		}
+		if !strings.HasSuffix(path, ".proto") {
+			return nil
+		}
+		dir := filepath.Dir(path)
+		packages[dir] = append(packages[dir], path)
+		return nil
+	})
+	if err != nil {
+		t.Fatal(err)
+	}
+
+	// Compile each package, using this binary as protoc-gen-go.
+	for _, sources := range packages {
+		args := []string{"-Itestdata", "--go_out=plugins=grpc,paths=source_relative:" + workdir}
+		args = append(args, sources...)
+		protoc(t, args)
+	}
+
+	// Compare each generated file to the golden version.
+	filepath.Walk(workdir, func(genPath string, info os.FileInfo, _ error) error {
+		if info.IsDir() {
+			return nil
+		}
+
+		// For each generated file, figure out the path to the corresponding
+		// golden file in the testdata directory.
+		relPath, err := filepath.Rel(workdir, genPath)
+		if err != nil {
+			t.Errorf("filepath.Rel(%q, %q): %v", workdir, genPath, err)
+			return nil
+		}
+		if filepath.SplitList(relPath)[0] == ".." {
+			t.Errorf("generated file %q is not relative to %q", genPath, workdir)
+		}
+		goldenPath := filepath.Join("testdata", relPath)
+
+		got, err := ioutil.ReadFile(genPath)
+		if err != nil {
+			t.Error(err)
+			return nil
+		}
+		if *regenerate {
+			// If --regenerate set, just rewrite the golden files.
+			err := ioutil.WriteFile(goldenPath, got, 0666)
+			if err != nil {
+				t.Error(err)
+			}
+			return nil
+		}
+
+		want, err := ioutil.ReadFile(goldenPath)
+		if err != nil {
+			t.Error(err)
+			return nil
+		}
+
+		want = fdescRE.ReplaceAll(want, nil)
+		got = fdescRE.ReplaceAll(got, nil)
+		if bytes.Equal(got, want) {
+			return nil
+		}
+
+		cmd := exec.Command("diff", "-u", goldenPath, genPath)
+		out, _ := cmd.CombinedOutput()
+		t.Errorf("golden file differs: %v\n%v", relPath, string(out))
+		return nil
+	})
+}
+
+var fdescRE = regexp.MustCompile(`(?ms)^var fileDescriptor.*}`)
+
+func protoc(t *testing.T, args []string) {
+	cmd := exec.Command("protoc", "--plugin=protoc-gen-go="+os.Args[0])
+	cmd.Args = append(cmd.Args, args...)
+	// We set the RUN_AS_PROTOC_GEN_GO environment variable to indicate that
+	// the subprocess should act as a proto compiler rather than a test.
+	cmd.Env = append(os.Environ(), "RUN_AS_PROTOC_GEN_GO=1")
+	out, err := cmd.CombinedOutput()
+	if len(out) > 0 || err != nil {
+		t.Log("RUNNING: ", strings.Join(cmd.Args, " "))
+	}
+	if len(out) > 0 {
+		t.Log(string(out))
+	}
+	if err != nil {
+		t.Fatalf("protoc: %v", err)
+	}
+}
+
+func hasReleaseTag(want string) bool {
+	for _, tag := range build.Default.ReleaseTags {
+		if tag == want {
+			return true
+		}
+	}
+	return false
+}
diff --git a/cmd/protoc-gen-go/main.go b/cmd/protoc-gen-go/main.go
new file mode 100644
index 0000000..9ea1420
--- /dev/null
+++ b/cmd/protoc-gen-go/main.go
@@ -0,0 +1,35 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// The protoc-gen-go binary is a protoc plugin to generate a Go protocol
+// buffer package.
+package main
+
+import (
+	"strings"
+
+	"google.golang.org/proto/protogen"
+)
+
+func main() {
+	protogen.Run(func(gen *protogen.Plugin) error {
+		for _, f := range gen.Files {
+			if !f.Generate {
+				continue
+			}
+			genFile(gen, f)
+		}
+		return nil
+	})
+}
+
+func genFile(gen *protogen.Plugin, f *protogen.File) {
+	g := gen.NewGeneratedFile(strings.TrimSuffix(f.Desc.GetName(), ".proto") + ".pb.go")
+	g.P("// Code generated by protoc-gen-go. DO NOT EDIT.")
+	g.P("// source: ", f.Desc.GetName())
+	g.P()
+	g.P("package TODO")
+
+	// TODO: Everything.
+}
diff --git a/cmd/protoc-gen-go/testdata/proto2/proto2.pb.go b/cmd/protoc-gen-go/testdata/proto2/proto2.pb.go
new file mode 100644
index 0000000..558363d
--- /dev/null
+++ b/cmd/protoc-gen-go/testdata/proto2/proto2.pb.go
@@ -0,0 +1,4 @@
+// Code generated by protoc-gen-go. DO NOT EDIT.
+// source: proto2/proto2.proto
+
+package TODO
diff --git a/cmd/protoc-gen-go/testdata/proto2/proto2.proto b/cmd/protoc-gen-go/testdata/proto2/proto2.proto
new file mode 100644
index 0000000..055bc71
--- /dev/null
+++ b/cmd/protoc-gen-go/testdata/proto2/proto2.proto
@@ -0,0 +1,12 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+syntax = "proto2";
+
+package goproto.protoc.proto2;
+
+option go_package = "google.golang.org/proto/cmd/protoc-gen-go/testdata/proto2";
+
+message Message {
+}
diff --git a/go.mod b/go.mod
index d94d903..6aa64ad 100644
--- a/go.mod
+++ b/go.mod
@@ -1,3 +1,8 @@
 module google.golang.org/proto
 
-require github.com/google/go-cmp v0.2.0
+require (
+	github.com/golang/protobuf v1.2.0
+	github.com/google/go-cmp v0.2.0
+	golang.org/x/net v0.0.0-20180821023952-922f4815f713 // indirect
+	golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f // indirect
+)
diff --git a/go.sum b/go.sum
index 5f4f636..1bfafa6 100644
--- a/go.sum
+++ b/go.sum
@@ -1,2 +1,8 @@
+github.com/golang/protobuf v1.2.0 h1:P3YflyNX/ehuJFLhxviNdFxQPkGK5cDcApsge1SqnvM=
+github.com/golang/protobuf v1.2.0/go.mod h1:6lQm79b+lXiMfvg/cZm0SGofjICqVBUtrP5yJMmIC1U=
 github.com/google/go-cmp v0.2.0 h1:+dTQ8DZQJz0Mb/HjFlkptS1FeQ4cWSnN941F8aEG4SQ=
 github.com/google/go-cmp v0.2.0/go.mod h1:oXzfMopK8JAjlY9xF4vHSVASa0yLyX7SntLO5aqRK0M=
+golang.org/x/net v0.0.0-20180821023952-922f4815f713 h1:rMJUcaDGbG+X967I4zGKCq5laYqcGKJmpB+3jhpOhPw=
+golang.org/x/net v0.0.0-20180821023952-922f4815f713/go.mod h1:mL1N/T3taQHkDXs73rZJwtUhF3w3ftmwwsq0BUmARs4=
+golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f h1:wMNYb4v58l5UBM7MYRLPG6ZhfOqbKu7X5eyFl8ZhKvA=
+golang.org/x/sync v0.0.0-20180314180146-1d60e4601c6f/go.mod h1:RxMgew5VJxzue5/jJTE5uejpjVlOe/izrB70Jof72aM=
diff --git a/protogen/names.go b/protogen/names.go
new file mode 100644
index 0000000..1976cfd
--- /dev/null
+++ b/protogen/names.go
@@ -0,0 +1,54 @@
+package protogen
+
+import (
+	"go/token"
+	"strconv"
+	"strings"
+	"unicode"
+	"unicode/utf8"
+)
+
+// A GoImportPath is the import path of a Go package. e.g., "google.golang.org/genproto/protobuf".
+type GoImportPath string
+
+func (p GoImportPath) String() string { return strconv.Quote(string(p)) }
+
+// A GoPackageName is the name of a Go package. e.g., "protobuf".
+type GoPackageName string
+
+// cleanPacakgeName converts a string to a valid Go package name.
+func cleanPackageName(name string) GoPackageName {
+	name = strings.Map(badToUnderscore, name)
+	// Identifier must not be keyword: insert _.
+	if token.Lookup(name).IsKeyword() {
+		name = "_" + name
+	}
+	// Identifier must not begin with digit: insert _.
+	if r, _ := utf8.DecodeRuneInString(name); unicode.IsDigit(r) {
+		name = "_" + name
+	}
+	return GoPackageName(name)
+}
+
+// badToUnderscore is the mapping function used to generate Go names from package names,
+// which can be dotted in the input .proto file.  It replaces non-identifier characters such as
+// dot or dash with underscore.
+func badToUnderscore(r rune) rune {
+	if unicode.IsLetter(r) || unicode.IsDigit(r) || r == '_' {
+		return r
+	}
+	return '_'
+}
+
+// baseName returns the last path element of the name, with the last dotted suffix removed.
+func baseName(name string) string {
+	// First, find the last element
+	if i := strings.LastIndex(name, "/"); i >= 0 {
+		name = name[i+1:]
+	}
+	// Now drop the suffix
+	if i := strings.LastIndex(name, "."); i >= 0 {
+		name = name[:i]
+	}
+	return name
+}
diff --git a/protogen/protogen.go b/protogen/protogen.go
new file mode 100644
index 0000000..b10edad
--- /dev/null
+++ b/protogen/protogen.go
@@ -0,0 +1,224 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Package protogen provides support for writing protoc plugins.
+//
+// Plugins for protoc, the Protocol Buffers Compiler, are programs which read
+// a CodeGeneratorRequest protocol buffer from standard input and write a
+// CodeGeneratorResponse protocol buffer to standard output. This package
+// provides support for writing plugins which generate Go code.
+package protogen
+
+import (
+	"bytes"
+	"fmt"
+	"io/ioutil"
+	"os"
+	"path/filepath"
+	"strings"
+
+	"github.com/golang/protobuf/proto"
+	descpb "github.com/golang/protobuf/protoc-gen-go/descriptor"
+	pluginpb "github.com/golang/protobuf/protoc-gen-go/plugin"
+)
+
+// Run executes a function as a protoc plugin.
+//
+// It reads a CodeGeneratorRequest message from os.Stdin, invokes the plugin
+// function, and writes a CodeGeneratorResponse message to os.Stdout.
+//
+// If a failure occurs while reading or writing, Run prints an error to
+// os.Stderr and calls os.Exit(1).
+func Run(f func(*Plugin) error) {
+	if err := run(f); err != nil {
+		fmt.Fprintf(os.Stderr, "%s: %v\n", filepath.Base(os.Args[0]), err)
+		os.Exit(1)
+	}
+}
+
+func run(f func(*Plugin) error) error {
+	in, err := ioutil.ReadAll(os.Stdin)
+	if err != nil {
+		return err
+	}
+	req := &pluginpb.CodeGeneratorRequest{}
+	if err := proto.Unmarshal(in, req); err != nil {
+		return err
+	}
+	gen, err := New(req)
+	if err != nil {
+		return err
+	}
+	if err := f(gen); err != nil {
+		// Errors from the plugin function are reported by setting the
+		// error field in the CodeGeneratorResponse.
+		//
+		// In contrast, errors that indicate a problem in protoc
+		// itself (unparsable input, I/O errors, etc.) are reported
+		// to stderr.
+		gen.Error(err)
+	}
+	resp := gen.Response()
+	out, err := proto.Marshal(resp)
+	if err != nil {
+		return err
+	}
+	if _, err := os.Stdout.Write(out); err != nil {
+		return err
+	}
+	return nil
+}
+
+// A Plugin is a protoc plugin invocation.
+type Plugin struct {
+	// Request is the CodeGeneratorRequest provided by protoc.
+	Request *pluginpb.CodeGeneratorRequest
+
+	// Files is the set of files to generate and everything they import.
+	// Files appear in topological order, so each file appears before any
+	// file that imports it.
+	Files       []*File
+	filesByName map[string]*File
+
+	packageImportPath string // Go import path of the package we're generating code for.
+
+	genFiles []*GeneratedFile
+	err      error
+}
+
+// New returns a new Plugin.
+func New(req *pluginpb.CodeGeneratorRequest) (*Plugin, error) {
+	gen := &Plugin{
+		Request:     req,
+		filesByName: make(map[string]*File),
+	}
+
+	// TODO: Figure out how to pass parameters to the generator.
+	for _, param := range strings.Split(req.GetParameter(), ",") {
+		var value string
+		if i := strings.Index(param, "="); i >= 0 {
+			value = param[i+1:]
+			param = param[0:i]
+		}
+		switch param {
+		case "":
+			// Ignore.
+		case "import_prefix":
+			// TODO
+		case "import_path":
+			gen.packageImportPath = value
+		case "paths":
+			// TODO
+		case "plugins":
+			// TODO
+		case "annotate_code":
+			// TODO
+		default:
+			if param[0] != 'M' {
+				return nil, fmt.Errorf("unknown parameter %q", param)
+			}
+			// TODO
+		}
+	}
+
+	for _, fdesc := range gen.Request.ProtoFile {
+		f := newFile(gen, fdesc)
+		name := f.Desc.GetName()
+		if gen.filesByName[name] != nil {
+			return nil, fmt.Errorf("duplicate file name: %q", name)
+		}
+		gen.Files = append(gen.Files, f)
+		gen.filesByName[name] = f
+	}
+	for _, name := range gen.Request.FileToGenerate {
+		f, ok := gen.FileByName(name)
+		if !ok {
+			return nil, fmt.Errorf("no descriptor for generated file: %v", name)
+		}
+		f.Generate = true
+	}
+	return gen, nil
+}
+
+// Error records an error in code generation. The generator will report the
+// error back to protoc and will not produce output.
+func (gen *Plugin) Error(err error) {
+	if gen.err == nil {
+		gen.err = err
+	}
+}
+
+// Response returns the generator output.
+func (gen *Plugin) Response() *pluginpb.CodeGeneratorResponse {
+	resp := &pluginpb.CodeGeneratorResponse{}
+	if gen.err != nil {
+		resp.Error = proto.String(gen.err.Error())
+		return resp
+	}
+	for _, gf := range gen.genFiles {
+		resp.File = append(resp.File, &pluginpb.CodeGeneratorResponse_File{
+			Name:    proto.String(gf.path),
+			Content: proto.String(string(gf.Content())),
+		})
+	}
+	return resp
+}
+
+// FileByName returns the file with the given name.
+func (gen *Plugin) FileByName(name string) (f *File, ok bool) {
+	f, ok = gen.filesByName[name]
+	return f, ok
+}
+
+// A File is a .proto source file.
+type File struct {
+	// TODO: Replace with protoreflect.FileDescriptor.
+	Desc *descpb.FileDescriptorProto
+
+	// Generate is true if the generator should generate code for this file.
+	Generate bool
+}
+
+func newFile(gen *Plugin, p *descpb.FileDescriptorProto) *File {
+	return &File{
+		Desc: p,
+	}
+}
+
+// A GeneratedFile is a generated file.
+type GeneratedFile struct {
+	path string
+	buf  bytes.Buffer
+}
+
+// NewGeneratedFile creates a new generated file with the given path.
+func (gen *Plugin) NewGeneratedFile(path string) *GeneratedFile {
+	g := &GeneratedFile{
+		path: path,
+	}
+	gen.genFiles = append(gen.genFiles, g)
+	return g
+}
+
+// P prints a line to the generated output. It converts each parameter to a
+// string following the same rules as fmt.Print. It never inserts spaces
+// between parameters.
+//
+// TODO: .meta file annotations.
+func (g *GeneratedFile) P(v ...interface{}) {
+	for _, x := range v {
+		fmt.Fprint(&g.buf, x)
+	}
+	fmt.Fprintln(&g.buf)
+}
+
+// Write implements io.Writer.
+func (g *GeneratedFile) Write(p []byte) (n int, err error) {
+	return g.buf.Write(p)
+}
+
+// Content returns the contents of the generated file.
+func (g *GeneratedFile) Content() []byte {
+	return g.buf.Bytes()
+}
diff --git a/protogen/protogen_test.go b/protogen/protogen_test.go
new file mode 100644
index 0000000..1d23cc0
--- /dev/null
+++ b/protogen/protogen_test.go
@@ -0,0 +1,94 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+package protogen
+
+import (
+	"io/ioutil"
+	"os"
+	"os/exec"
+	"path/filepath"
+	"strings"
+	"testing"
+
+	"github.com/golang/protobuf/proto"
+	pluginpb "github.com/golang/protobuf/protoc-gen-go/plugin"
+)
+
+func TestFiles(t *testing.T) {
+	gen, err := New(makeRequest(t, "testdata/go_package/no_go_package_import.proto"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	for _, test := range []struct {
+		path         string
+		wantGenerate bool
+	}{
+		{
+			path:         "go_package/no_go_package_import.proto",
+			wantGenerate: true,
+		},
+		{
+			path:         "go_package/no_go_package.proto",
+			wantGenerate: false,
+		},
+	} {
+		f, ok := gen.FileByName(test.path)
+		if !ok {
+			t.Errorf("%q: not found by gen.FileByName", test.path)
+			continue
+		}
+		if f.Generate != test.wantGenerate {
+			t.Errorf("%q: Generate=%v, want %v", test.path, f.Generate, test.wantGenerate)
+		}
+	}
+}
+
+// makeRequest returns a CodeGeneratorRequest for the given protoc inputs.
+//
+// It does this by running protoc with the current binary as the protoc-gen-go
+// plugin. This "plugin" produces a single file, named 'request', which contains
+// the code generator request.
+func makeRequest(t *testing.T, args ...string) *pluginpb.CodeGeneratorRequest {
+	workdir, err := ioutil.TempDir("", "test")
+	if err != nil {
+		t.Fatal(err)
+	}
+	defer os.RemoveAll(workdir)
+
+	cmd := exec.Command("protoc", "--plugin=protoc-gen-go="+os.Args[0])
+	cmd.Args = append(cmd.Args, "--go_out="+workdir, "-Itestdata")
+	cmd.Args = append(cmd.Args, args...)
+	cmd.Env = append(os.Environ(), "RUN_AS_PROTOC_PLUGIN=1")
+	out, err := cmd.CombinedOutput()
+	if len(out) > 0 || err != nil {
+		t.Log("RUNNING: ", strings.Join(cmd.Args, " "))
+	}
+	if len(out) > 0 {
+		t.Log(string(out))
+	}
+	if err != nil {
+		t.Fatalf("protoc: %v", err)
+	}
+
+	b, err := ioutil.ReadFile(filepath.Join(workdir, "request"))
+	if err != nil {
+		t.Fatal(err)
+	}
+	req := &pluginpb.CodeGeneratorRequest{}
+	if err := proto.UnmarshalText(string(b), req); err != nil {
+		t.Fatal(err)
+	}
+	return req
+}
+
+func init() {
+	if os.Getenv("RUN_AS_PROTOC_PLUGIN") != "" {
+		Run(func(p *Plugin) error {
+			g := p.NewGeneratedFile("request")
+			return proto.MarshalText(g, p.Request)
+		})
+		os.Exit(0)
+	}
+}
diff --git a/protogen/testdata/go_package/no_go_package.proto b/protogen/testdata/go_package/no_go_package.proto
new file mode 100644
index 0000000..d71884c
--- /dev/null
+++ b/protogen/testdata/go_package/no_go_package.proto
@@ -0,0 +1,9 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Proto source file with no go_package option.
+
+syntax = "proto3";
+package goproto.testdata;
+message M {}
diff --git a/protogen/testdata/go_package/no_go_package_import.proto b/protogen/testdata/go_package/no_go_package_import.proto
new file mode 100644
index 0000000..bb1a73f
--- /dev/null
+++ b/protogen/testdata/go_package/no_go_package_import.proto
@@ -0,0 +1,13 @@
+// Copyright 2018 The Go Authors. All rights reserved.
+// Use of this source code is governed by a BSD-style
+// license that can be found in the LICENSE file.
+
+// Import of a proto source file with no go_package option.
+
+syntax = "proto3";
+package goproto.testdata;
+import "go_package/no_go_package.proto";
+message M1 {
+  M Field = 1;
+}
+