encode.go (11227B)
1 // Copyright 2018 The Go Authors. All rights reserved. 2 // Use of this source code is governed by a BSD-style 3 // license that can be found in the LICENSE file. 4 5 package prototext 6 7 import ( 8 "fmt" 9 "strconv" 10 "unicode/utf8" 11 12 "google.golang.org/protobuf/encoding/protowire" 13 "google.golang.org/protobuf/internal/encoding/messageset" 14 "google.golang.org/protobuf/internal/encoding/text" 15 "google.golang.org/protobuf/internal/errors" 16 "google.golang.org/protobuf/internal/flags" 17 "google.golang.org/protobuf/internal/genid" 18 "google.golang.org/protobuf/internal/order" 19 "google.golang.org/protobuf/internal/pragma" 20 "google.golang.org/protobuf/internal/strs" 21 "google.golang.org/protobuf/proto" 22 "google.golang.org/protobuf/reflect/protoreflect" 23 "google.golang.org/protobuf/reflect/protoregistry" 24 ) 25 26 const defaultIndent = " " 27 28 // Format formats the message as a multiline string. 29 // This function is only intended for human consumption and ignores errors. 30 // Do not depend on the output being stable. It may change over time across 31 // different versions of the program. 32 func Format(m proto.Message) string { 33 return MarshalOptions{Multiline: true}.Format(m) 34 } 35 36 // Marshal writes the given proto.Message in textproto format using default 37 // options. Do not depend on the output being stable. It may change over time 38 // across different versions of the program. 39 func Marshal(m proto.Message) ([]byte, error) { 40 return MarshalOptions{}.Marshal(m) 41 } 42 43 // MarshalOptions is a configurable text format marshaler. 44 type MarshalOptions struct { 45 pragma.NoUnkeyedLiterals 46 47 // Multiline specifies whether the marshaler should format the output in 48 // indented-form with every textual element on a new line. 49 // If Indent is an empty string, then an arbitrary indent is chosen. 50 Multiline bool 51 52 // Indent specifies the set of indentation characters to use in a multiline 53 // formatted output such that every entry is preceded by Indent and 54 // terminated by a newline. If non-empty, then Multiline is treated as true. 55 // Indent can only be composed of space or tab characters. 56 Indent string 57 58 // EmitASCII specifies whether to format strings and bytes as ASCII only 59 // as opposed to using UTF-8 encoding when possible. 60 EmitASCII bool 61 62 // allowInvalidUTF8 specifies whether to permit the encoding of strings 63 // with invalid UTF-8. This is unexported as it is intended to only 64 // be specified by the Format method. 65 allowInvalidUTF8 bool 66 67 // AllowPartial allows messages that have missing required fields to marshal 68 // without returning an error. If AllowPartial is false (the default), 69 // Marshal will return error if there are any missing required fields. 70 AllowPartial bool 71 72 // EmitUnknown specifies whether to emit unknown fields in the output. 73 // If specified, the unmarshaler may be unable to parse the output. 74 // The default is to exclude unknown fields. 75 EmitUnknown bool 76 77 // Resolver is used for looking up types when expanding google.protobuf.Any 78 // messages. If nil, this defaults to using protoregistry.GlobalTypes. 79 Resolver interface { 80 protoregistry.ExtensionTypeResolver 81 protoregistry.MessageTypeResolver 82 } 83 } 84 85 // Format formats the message as a string. 86 // This method is only intended for human consumption and ignores errors. 87 // Do not depend on the output being stable. It may change over time across 88 // different versions of the program. 89 func (o MarshalOptions) Format(m proto.Message) string { 90 if m == nil || !m.ProtoReflect().IsValid() { 91 return "<nil>" // invalid syntax, but okay since this is for debugging 92 } 93 o.allowInvalidUTF8 = true 94 o.AllowPartial = true 95 o.EmitUnknown = true 96 b, _ := o.Marshal(m) 97 return string(b) 98 } 99 100 // Marshal writes the given proto.Message in textproto format using options in 101 // MarshalOptions object. Do not depend on the output being stable. It may 102 // change over time across different versions of the program. 103 func (o MarshalOptions) Marshal(m proto.Message) ([]byte, error) { 104 return o.marshal(m) 105 } 106 107 // marshal is a centralized function that all marshal operations go through. 108 // For profiling purposes, avoid changing the name of this function or 109 // introducing other code paths for marshal that do not go through this. 110 func (o MarshalOptions) marshal(m proto.Message) ([]byte, error) { 111 var delims = [2]byte{'{', '}'} 112 113 if o.Multiline && o.Indent == "" { 114 o.Indent = defaultIndent 115 } 116 if o.Resolver == nil { 117 o.Resolver = protoregistry.GlobalTypes 118 } 119 120 internalEnc, err := text.NewEncoder(o.Indent, delims, o.EmitASCII) 121 if err != nil { 122 return nil, err 123 } 124 125 // Treat nil message interface as an empty message, 126 // in which case there is nothing to output. 127 if m == nil { 128 return []byte{}, nil 129 } 130 131 enc := encoder{internalEnc, o} 132 err = enc.marshalMessage(m.ProtoReflect(), false) 133 if err != nil { 134 return nil, err 135 } 136 out := enc.Bytes() 137 if len(o.Indent) > 0 && len(out) > 0 { 138 out = append(out, '\n') 139 } 140 if o.AllowPartial { 141 return out, nil 142 } 143 return out, proto.CheckInitialized(m) 144 } 145 146 type encoder struct { 147 *text.Encoder 148 opts MarshalOptions 149 } 150 151 // marshalMessage marshals the given protoreflect.Message. 152 func (e encoder) marshalMessage(m protoreflect.Message, inclDelims bool) error { 153 messageDesc := m.Descriptor() 154 if !flags.ProtoLegacy && messageset.IsMessageSet(messageDesc) { 155 return errors.New("no support for proto1 MessageSets") 156 } 157 158 if inclDelims { 159 e.StartMessage() 160 defer e.EndMessage() 161 } 162 163 // Handle Any expansion. 164 if messageDesc.FullName() == genid.Any_message_fullname { 165 if e.marshalAny(m) { 166 return nil 167 } 168 // If unable to expand, continue on to marshal Any as a regular message. 169 } 170 171 // Marshal fields. 172 var err error 173 order.RangeFields(m, order.IndexNameFieldOrder, func(fd protoreflect.FieldDescriptor, v protoreflect.Value) bool { 174 if err = e.marshalField(fd.TextName(), v, fd); err != nil { 175 return false 176 } 177 return true 178 }) 179 if err != nil { 180 return err 181 } 182 183 // Marshal unknown fields. 184 if e.opts.EmitUnknown { 185 e.marshalUnknown(m.GetUnknown()) 186 } 187 188 return nil 189 } 190 191 // marshalField marshals the given field with protoreflect.Value. 192 func (e encoder) marshalField(name string, val protoreflect.Value, fd protoreflect.FieldDescriptor) error { 193 switch { 194 case fd.IsList(): 195 return e.marshalList(name, val.List(), fd) 196 case fd.IsMap(): 197 return e.marshalMap(name, val.Map(), fd) 198 default: 199 e.WriteName(name) 200 return e.marshalSingular(val, fd) 201 } 202 } 203 204 // marshalSingular marshals the given non-repeated field value. This includes 205 // all scalar types, enums, messages, and groups. 206 func (e encoder) marshalSingular(val protoreflect.Value, fd protoreflect.FieldDescriptor) error { 207 kind := fd.Kind() 208 switch kind { 209 case protoreflect.BoolKind: 210 e.WriteBool(val.Bool()) 211 212 case protoreflect.StringKind: 213 s := val.String() 214 if !e.opts.allowInvalidUTF8 && strs.EnforceUTF8(fd) && !utf8.ValidString(s) { 215 return errors.InvalidUTF8(string(fd.FullName())) 216 } 217 e.WriteString(s) 218 219 case protoreflect.Int32Kind, protoreflect.Int64Kind, 220 protoreflect.Sint32Kind, protoreflect.Sint64Kind, 221 protoreflect.Sfixed32Kind, protoreflect.Sfixed64Kind: 222 e.WriteInt(val.Int()) 223 224 case protoreflect.Uint32Kind, protoreflect.Uint64Kind, 225 protoreflect.Fixed32Kind, protoreflect.Fixed64Kind: 226 e.WriteUint(val.Uint()) 227 228 case protoreflect.FloatKind: 229 // Encoder.WriteFloat handles the special numbers NaN and infinites. 230 e.WriteFloat(val.Float(), 32) 231 232 case protoreflect.DoubleKind: 233 // Encoder.WriteFloat handles the special numbers NaN and infinites. 234 e.WriteFloat(val.Float(), 64) 235 236 case protoreflect.BytesKind: 237 e.WriteString(string(val.Bytes())) 238 239 case protoreflect.EnumKind: 240 num := val.Enum() 241 if desc := fd.Enum().Values().ByNumber(num); desc != nil { 242 e.WriteLiteral(string(desc.Name())) 243 } else { 244 // Use numeric value if there is no enum description. 245 e.WriteInt(int64(num)) 246 } 247 248 case protoreflect.MessageKind, protoreflect.GroupKind: 249 return e.marshalMessage(val.Message(), true) 250 251 default: 252 panic(fmt.Sprintf("%v has unknown kind: %v", fd.FullName(), kind)) 253 } 254 return nil 255 } 256 257 // marshalList marshals the given protoreflect.List as multiple name-value fields. 258 func (e encoder) marshalList(name string, list protoreflect.List, fd protoreflect.FieldDescriptor) error { 259 size := list.Len() 260 for i := 0; i < size; i++ { 261 e.WriteName(name) 262 if err := e.marshalSingular(list.Get(i), fd); err != nil { 263 return err 264 } 265 } 266 return nil 267 } 268 269 // marshalMap marshals the given protoreflect.Map as multiple name-value fields. 270 func (e encoder) marshalMap(name string, mmap protoreflect.Map, fd protoreflect.FieldDescriptor) error { 271 var err error 272 order.RangeEntries(mmap, order.GenericKeyOrder, func(key protoreflect.MapKey, val protoreflect.Value) bool { 273 e.WriteName(name) 274 e.StartMessage() 275 defer e.EndMessage() 276 277 e.WriteName(string(genid.MapEntry_Key_field_name)) 278 err = e.marshalSingular(key.Value(), fd.MapKey()) 279 if err != nil { 280 return false 281 } 282 283 e.WriteName(string(genid.MapEntry_Value_field_name)) 284 err = e.marshalSingular(val, fd.MapValue()) 285 if err != nil { 286 return false 287 } 288 return true 289 }) 290 return err 291 } 292 293 // marshalUnknown parses the given []byte and marshals fields out. 294 // This function assumes proper encoding in the given []byte. 295 func (e encoder) marshalUnknown(b []byte) { 296 const dec = 10 297 const hex = 16 298 for len(b) > 0 { 299 num, wtype, n := protowire.ConsumeTag(b) 300 b = b[n:] 301 e.WriteName(strconv.FormatInt(int64(num), dec)) 302 303 switch wtype { 304 case protowire.VarintType: 305 var v uint64 306 v, n = protowire.ConsumeVarint(b) 307 e.WriteUint(v) 308 case protowire.Fixed32Type: 309 var v uint32 310 v, n = protowire.ConsumeFixed32(b) 311 e.WriteLiteral("0x" + strconv.FormatUint(uint64(v), hex)) 312 case protowire.Fixed64Type: 313 var v uint64 314 v, n = protowire.ConsumeFixed64(b) 315 e.WriteLiteral("0x" + strconv.FormatUint(v, hex)) 316 case protowire.BytesType: 317 var v []byte 318 v, n = protowire.ConsumeBytes(b) 319 e.WriteString(string(v)) 320 case protowire.StartGroupType: 321 e.StartMessage() 322 var v []byte 323 v, n = protowire.ConsumeGroup(num, b) 324 e.marshalUnknown(v) 325 e.EndMessage() 326 default: 327 panic(fmt.Sprintf("prototext: error parsing unknown field wire type: %v", wtype)) 328 } 329 330 b = b[n:] 331 } 332 } 333 334 // marshalAny marshals the given google.protobuf.Any message in expanded form. 335 // It returns true if it was able to marshal, else false. 336 func (e encoder) marshalAny(any protoreflect.Message) bool { 337 // Construct the embedded message. 338 fds := any.Descriptor().Fields() 339 fdType := fds.ByNumber(genid.Any_TypeUrl_field_number) 340 typeURL := any.Get(fdType).String() 341 mt, err := e.opts.Resolver.FindMessageByURL(typeURL) 342 if err != nil { 343 return false 344 } 345 m := mt.New().Interface() 346 347 // Unmarshal bytes into embedded message. 348 fdValue := fds.ByNumber(genid.Any_Value_field_number) 349 value := any.Get(fdValue) 350 err = proto.UnmarshalOptions{ 351 AllowPartial: true, 352 Resolver: e.opts.Resolver, 353 }.Unmarshal(value.Bytes(), m) 354 if err != nil { 355 return false 356 } 357 358 // Get current encoder position. If marshaling fails, reset encoder output 359 // back to this position. 360 pos := e.Snapshot() 361 362 // Field name is the proto field name enclosed in []. 363 e.WriteName("[" + typeURL + "]") 364 err = e.marshalMessage(m.ProtoReflect(), true) 365 if err != nil { 366 e.Reset(pos) 367 return false 368 } 369 return true 370 }