Source file src/encoding/gob/doc.go

     1  // Copyright 2009 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  /*
     6  Package gob manages streams of gobs - binary values exchanged between an
     7  [Encoder] (transmitter) and a [Decoder] (receiver). A typical use is transporting
     8  arguments and results of remote procedure calls (RPCs) such as those provided by
     9  [net/rpc].
    10  
    11  The implementation compiles a custom codec for each data type in the stream and
    12  is most efficient when a single [Encoder] is used to transmit a stream of values,
    13  amortizing the cost of compilation.
    14  
    15  # Basics
    16  
    17  A stream of gobs is self-describing. Each data item in the stream is preceded by
    18  a specification of its type, expressed in terms of a small set of predefined
    19  types. Pointers are not transmitted, but the things they point to are
    20  transmitted; that is, the values are flattened. Nil pointers are not permitted,
    21  as they have no value. Recursive types work fine, but
    22  recursive values (data with cycles) are problematic. This may change.
    23  
    24  To use gobs, create an [Encoder] and present it with a series of data items as
    25  values or addresses that can be dereferenced to values. The [Encoder] makes sure
    26  all type information is sent before it is needed. At the receive side, a
    27  [Decoder] retrieves values from the encoded stream and unpacks them into local
    28  variables.
    29  
    30  # Types and Values
    31  
    32  The source and destination values/types need not correspond exactly. For structs,
    33  fields (identified by name) that are in the source but absent from the receiving
    34  variable will be ignored. Fields that are in the receiving variable but missing
    35  from the transmitted type or value will be ignored in the destination. If a field
    36  with the same name is present in both, their types must be compatible. Both the
    37  receiver and transmitter will do all necessary indirection and dereferencing to
    38  convert between gobs and actual Go values. For instance, a gob type that is
    39  schematically,
    40  
    41  	struct { A, B int }
    42  
    43  can be sent from or received into any of these Go types:
    44  
    45  	struct { A, B int }	// the same
    46  	*struct { A, B int }	// extra indirection of the struct
    47  	struct { *A, **B int }	// extra indirection of the fields
    48  	struct { A, B int64 }	// different concrete value type; see below
    49  
    50  It may also be received into any of these:
    51  
    52  	struct { A, B int }	// the same
    53  	struct { B, A int }	// ordering doesn't matter; matching is by name
    54  	struct { A, B, C int }	// extra field (C) ignored
    55  	struct { B int }	// missing field (A) ignored; data will be dropped
    56  	struct { B, C int }	// missing field (A) ignored; extra field (C) ignored.
    57  
    58  Attempting to receive into these types will draw a decode error:
    59  
    60  	struct { A int; B uint }	// change of signedness for B
    61  	struct { A int; B float }	// change of type for B
    62  	struct { }			// no field names in common
    63  	struct { C, D int }		// no field names in common
    64  
    65  Integers are transmitted two ways: arbitrary precision signed integers or
    66  arbitrary precision unsigned integers. There is no int8, int16 etc.
    67  discrimination in the gob format; there are only signed and unsigned integers. As
    68  described below, the transmitter sends the value in a variable-length encoding;
    69  the receiver accepts the value and stores it in the destination variable.
    70  Floating-point numbers are always sent using IEEE 754 64-bit precision (see
    71  below).
    72  
    73  Signed integers may be received into any signed integer variable: int, int16, etc.;
    74  unsigned integers may be received into any unsigned integer variable; and floating
    75  point values may be received into any floating point variable. However,
    76  the destination variable must be able to represent the value or the decode
    77  operation will fail.
    78  
    79  Structs, arrays and slices are also supported. Structs encode and decode only
    80  exported fields. Strings and arrays of bytes are supported with a special,
    81  efficient representation (see below). When a slice is decoded, if the existing
    82  slice has capacity the slice will be extended in place; if not, a new array is
    83  allocated. Regardless, the length of the resulting slice reports the number of
    84  elements decoded.
    85  
    86  In general, if allocation is required, the decoder will allocate memory. If not,
    87  it will update the destination variables with values read from the stream. It does
    88  not initialize them first, so if the destination is a compound value such as a
    89  map, struct, or slice, the decoded values will be merged elementwise into the
    90  existing variables.
    91  
    92  Functions and channels will not be sent in a gob. Attempting to encode such a value
    93  at the top level will fail. A struct field of chan or func type is treated exactly
    94  like an unexported field and is ignored.
    95  
    96  Gob can encode a value of any type implementing the [GobEncoder] or
    97  [encoding.BinaryMarshaler] interfaces by calling the corresponding method,
    98  in that order of preference.
    99  
   100  Gob can decode a value of any type implementing the [GobDecoder] or
   101  [encoding.BinaryUnmarshaler] interfaces by calling the corresponding method,
   102  again in that order of preference.
   103  
   104  # Encoding Details
   105  
   106  This section documents the encoding, details that are not important for most
   107  users. Details are presented bottom-up.
   108  
   109  An unsigned integer is sent one of two ways. If it is less than 128, it is sent
   110  as a byte with that value. Otherwise it is sent as a minimal-length big-endian
   111  (high byte first) byte stream holding the value, preceded by one byte holding the
   112  byte count, negated. Thus 0 is transmitted as (00), 7 is transmitted as (07) and
   113  256 is transmitted as (FE 01 00).
   114  
   115  A boolean is encoded within an unsigned integer: 0 for false, 1 for true.
   116  
   117  A signed integer, i, is encoded within an unsigned integer, u. Within u, bits 1
   118  upward contain the value; bit 0 says whether they should be complemented upon
   119  receipt. The encode algorithm looks like this:
   120  
   121  	var u uint
   122  	if i < 0 {
   123  		u = (^uint(i) << 1) | 1 // complement i, bit 0 is 1
   124  	} else {
   125  		u = (uint(i) << 1) // do not complement i, bit 0 is 0
   126  	}
   127  	encodeUnsigned(u)
   128  
   129  The low bit is therefore analogous to a sign bit, but making it the complement bit
   130  instead guarantees that the largest negative integer is not a special case. For
   131  example, -129=^128=(^256>>1) encodes as (FE 01 01).
   132  
   133  Floating-point numbers are always sent as a representation of a float64 value.
   134  That value is converted to a uint64 using [math.Float64bits]. The uint64 is then
   135  byte-reversed and sent as a regular unsigned integer. The byte-reversal means the
   136  exponent and high-precision part of the mantissa go first. Since the low bits are
   137  often zero, this can save encoding bytes. For instance, 17.0 is encoded in only
   138  three bytes (FE 31 40).
   139  
   140  Strings and slices of bytes are sent as an unsigned count followed by that many
   141  uninterpreted bytes of the value.
   142  
   143  All other slices and arrays are sent as an unsigned count followed by that many
   144  elements using the standard gob encoding for their type, recursively.
   145  
   146  Maps are sent as an unsigned count followed by that many key, element
   147  pairs. Empty but non-nil maps are sent, so if the receiver has not allocated
   148  one already, one will always be allocated on receipt unless the transmitted map
   149  is nil and not at the top level.
   150  
   151  In slices and arrays, as well as maps, all elements, even zero-valued elements,
   152  are transmitted, even if all the elements are zero.
   153  
   154  Structs are sent as a sequence of (field number, field value) pairs. The field
   155  value is sent using the standard gob encoding for its type, recursively. If a
   156  field has the zero value for its type (except for arrays; see above), it is omitted
   157  from the transmission. The field number is defined by the type of the encoded
   158  struct: the first field of the encoded type is field 0, the second is field 1,
   159  etc. When encoding a value, the field numbers are delta encoded for efficiency
   160  and the fields are always sent in order of increasing field number; the deltas are
   161  therefore unsigned. The initialization for the delta encoding sets the field
   162  number to -1, so an unsigned integer field 0 with value 7 is transmitted as unsigned
   163  delta = 1, unsigned value = 7 or (01 07). Finally, after all the fields have been
   164  sent a terminating mark denotes the end of the struct. That mark is a delta=0
   165  value, which has representation (00).
   166  
   167  Interface types are not checked for compatibility; all interface types are
   168  treated, for transmission, as members of a single "interface" type, analogous to
   169  int or []byte - in effect they're all treated as interface{}. Interface values
   170  are transmitted as a string identifying the concrete type being sent (a name
   171  that must be pre-defined by calling [Register]), followed by a byte count of the
   172  length of the following data (so the value can be skipped if it cannot be
   173  stored), followed by the usual encoding of concrete (dynamic) value stored in
   174  the interface value. (A nil interface value is identified by the empty string
   175  and transmits no value.) Upon receipt, the decoder verifies that the unpacked
   176  concrete item satisfies the interface of the receiving variable.
   177  
   178  If a value is passed to [Encoder.Encode] and the type is not a struct (or pointer to struct,
   179  etc.), for simplicity of processing it is represented as a struct of one field.
   180  The only visible effect of this is to encode a zero byte after the value, just as
   181  after the last field of an encoded struct, so that the decode algorithm knows when
   182  the top-level value is complete.
   183  
   184  The representation of types is described below. When a type is defined on a given
   185  connection between an [Encoder] and [Decoder], it is assigned a signed integer type
   186  id. When [Encoder.Encode](v) is called, it makes sure there is an id assigned for
   187  the type of v and all its elements and then it sends the pair (typeid, encoded-v)
   188  where typeid is the type id of the encoded type of v and encoded-v is the gob
   189  encoding of the value v.
   190  
   191  To define a type, the encoder chooses an unused, positive type id and sends the
   192  pair (-type id, encoded-type) where encoded-type is the gob encoding of a wireType
   193  description, constructed from these types:
   194  
   195  	type wireType struct {
   196  		ArrayT           *arrayType
   197  		SliceT           *sliceType
   198  		StructT          *structType
   199  		MapT             *mapType
   200  		GobEncoderT      *gobEncoderType
   201  		BinaryMarshalerT *gobEncoderType
   202  		TextMarshalerT   *gobEncoderType
   203  	}
   204  	type arrayType struct {
   205  		CommonType
   206  		Elem typeId
   207  		Len  int
   208  	}
   209  	type CommonType struct {
   210  		Name string // the name of the struct type
   211  		Id  int    // the id of the type, repeated so it's inside the type
   212  	}
   213  	type sliceType struct {
   214  		CommonType
   215  		Elem typeId
   216  	}
   217  	type structType struct {
   218  		CommonType
   219  		Field []fieldType // the fields of the struct.
   220  	}
   221  	type fieldType struct {
   222  		Name string // the name of the field.
   223  		Id   int    // the type id of the field, which must be already defined
   224  	}
   225  	type mapType struct {
   226  		CommonType
   227  		Key  typeId
   228  		Elem typeId
   229  	}
   230  	type gobEncoderType struct {
   231  		CommonType
   232  	}
   233  
   234  If there are nested type ids, the types for all inner type ids must be defined
   235  before the top-level type id is used to describe an encoded-v.
   236  
   237  For simplicity in setup, the connection is defined to understand these types a
   238  priori, as well as the basic gob types int, uint, etc. Their ids are:
   239  
   240  	bool        1
   241  	int         2
   242  	uint        3
   243  	float       4
   244  	[]byte      5
   245  	string      6
   246  	complex     7
   247  	interface   8
   248  	// gap for reserved ids.
   249  	WireType    16
   250  	ArrayType   17
   251  	CommonType  18
   252  	SliceType   19
   253  	StructType  20
   254  	FieldType   21
   255  	// 22 is slice of fieldType.
   256  	MapType     23
   257  
   258  Finally, each message created by a call to Encode is preceded by an encoded
   259  unsigned integer count of the number of bytes remaining in the message. After
   260  the initial type name, interface values are wrapped the same way; in effect, the
   261  interface value acts like a recursive invocation of Encode.
   262  
   263  In summary, a gob stream looks like
   264  
   265  	(byteCount (-type id, encoding of a wireType)* (type id, encoding of a value))*
   266  
   267  where * signifies zero or more repetitions and the type id of a value must
   268  be predefined or be defined before the value in the stream.
   269  
   270  Compatibility: Any future changes to the package will endeavor to maintain
   271  compatibility with streams encoded using previous versions. That is, any released
   272  version of this package should be able to decode data written with any previously
   273  released version, subject to issues such as security fixes. See the Go compatibility
   274  document for background: https://golang.org/doc/go1compat
   275  
   276  See "Gobs of data" for a design discussion of the gob wire format:
   277  https://blog.golang.org/gobs-of-data
   278  
   279  # Security
   280  
   281  This package is not designed to be hardened against adversarial inputs, and is
   282  outside the scope of https://go.dev/security/policy. In particular, the [Decoder]
   283  does only basic sanity checking on decoded input sizes, and its limits are not
   284  configurable. Care should be taken when decoding gob data from untrusted
   285  sources, which may consume significant resources.
   286  */
   287  package gob
   288  
   289  /*
   290  Grammar:
   291  
   292  Tokens starting with a lower case letter are terminals; int(n)
   293  and uint(n) represent the signed/unsigned encodings of the value n.
   294  
   295  GobStream:
   296  	DelimitedMessage*
   297  DelimitedMessage:
   298  	uint(lengthOfMessage) Message
   299  Message:
   300  	TypeSequence TypedValue
   301  TypeSequence
   302  	(TypeDefinition DelimitedTypeDefinition*)?
   303  DelimitedTypeDefinition:
   304  	uint(lengthOfTypeDefinition) TypeDefinition
   305  TypedValue:
   306  	int(typeId) Value
   307  TypeDefinition:
   308  	int(-typeId) encodingOfWireType
   309  Value:
   310  	SingletonValue | StructValue
   311  SingletonValue:
   312  	uint(0) FieldValue
   313  FieldValue:
   314  	builtinValue | ArrayValue | MapValue | SliceValue | StructValue | InterfaceValue
   315  InterfaceValue:
   316  	NilInterfaceValue | NonNilInterfaceValue
   317  NilInterfaceValue:
   318  	uint(0)
   319  NonNilInterfaceValue:
   320  	ConcreteTypeName TypeSequence InterfaceContents
   321  ConcreteTypeName:
   322  	uint(lengthOfName) [already read=n] name
   323  InterfaceContents:
   324  	int(concreteTypeId) DelimitedValue
   325  DelimitedValue:
   326  	uint(length) Value
   327  ArrayValue:
   328  	uint(n) FieldValue*n [n elements]
   329  MapValue:
   330  	uint(n) (FieldValue FieldValue)*n  [n (key, value) pairs]
   331  SliceValue:
   332  	uint(n) FieldValue*n [n elements]
   333  StructValue:
   334  	(uint(fieldDelta) FieldValue)*
   335  */
   336  
   337  /*
   338  For implementers and the curious, here is an encoded example. Given
   339  	type Point struct {X, Y int}
   340  and the value
   341  	p := Point{22, 33}
   342  the bytes transmitted that encode p will be:
   343  	1f ff 81 03 01 01 05 50 6f 69 6e 74 01 ff 82 00
   344  	01 02 01 01 58 01 04 00 01 01 59 01 04 00 00 00
   345  	07 ff 82 01 2c 01 42 00
   346  They are determined as follows.
   347  
   348  Since this is the first transmission of type Point, the type descriptor
   349  for Point itself must be sent before the value. This is the first type
   350  we've sent on this Encoder, so it has type id 65 (0 through 64 are
   351  reserved).
   352  
   353  	1f	// This item (a type descriptor) is 31 bytes long.
   354  	ff 81	// The negative of the id for the type we're defining, -65.
   355  		// This is one byte (indicated by FF = -1) followed by
   356  		// ^-65<<1 | 1. The low 1 bit signals to complement the
   357  		// rest upon receipt.
   358  
   359  	// Now we send a type descriptor, which is itself a struct (wireType).
   360  	// The type of wireType itself is known (it's built in, as is the type of
   361  	// all its components), so we just need to send a *value* of type wireType
   362  	// that represents type "Point".
   363  	// Here starts the encoding of that value.
   364  	// Set the field number implicitly to -1; this is done at the beginning
   365  	// of every struct, including nested structs.
   366  	03	// Add 3 to field number; now 2 (wireType.structType; this is a struct).
   367  		// structType starts with an embedded CommonType, which appears
   368  		// as a regular structure here too.
   369  	01	// add 1 to field number (now 0); start of embedded CommonType.
   370  	01	// add 1 to field number (now 0, the name of the type)
   371  	05	// string is (unsigned) 5 bytes long
   372  	50 6f 69 6e 74	// wireType.structType.CommonType.name = "Point"
   373  	01	// add 1 to field number (now 1, the id of the type)
   374  	ff 82	// wireType.structType.CommonType._id = 65
   375  	00	// end of embedded wiretype.structType.CommonType struct
   376  	01	// add 1 to field number (now 1, the field array in wireType.structType)
   377  	02	// There are two fields in the type (len(structType.field))
   378  	01	// Start of first field structure; add 1 to get field number 0: field[0].name
   379  	01	// 1 byte
   380  	58	// structType.field[0].name = "X"
   381  	01	// Add 1 to get field number 1: field[0].id
   382  	04	// structType.field[0].typeId is 2 (signed int).
   383  	00	// End of structType.field[0]; start structType.field[1]; set field number to -1.
   384  	01	// Add 1 to get field number 0: field[1].name
   385  	01	// 1 byte
   386  	59	// structType.field[1].name = "Y"
   387  	01	// Add 1 to get field number 1: field[1].id
   388  	04	// struct.Type.field[1].typeId is 2 (signed int).
   389  	00	// End of structType.field[1]; end of structType.field.
   390  	00	// end of wireType.structType structure
   391  	00	// end of wireType structure
   392  
   393  Now we can send the Point value. Again the field number resets to -1:
   394  
   395  	07	// this value is 7 bytes long
   396  	ff 82	// the type number, 65 (1 byte (-FF) followed by 65<<1)
   397  	01	// add one to field number, yielding field 0
   398  	2c	// encoding of signed "22" (0x2c = 44 = 22<<1); Point.x = 22
   399  	01	// add one to field number, yielding field 1
   400  	42	// encoding of signed "33" (0x42 = 66 = 33<<1); Point.y = 33
   401  	00	// end of structure
   402  
   403  The type encoding is long and fairly intricate but we send it only once.
   404  If p is transmitted a second time, the type is already known so the
   405  output will be just:
   406  
   407  	07 ff 82 01 2c 01 42 00
   408  
   409  A single non-struct value at top level is transmitted like a field with
   410  delta tag 0. For instance, a signed integer with value 3 presented as
   411  the argument to Encode will emit:
   412  
   413  	03 04 00 06
   414  
   415  Which represents:
   416  
   417  	03	// this value is 3 bytes long
   418  	04	// the type number, 2, represents an integer
   419  	00	// tag delta 0
   420  	06	// value 3
   421  
   422  */
   423  

View as plain text