Source file src/encoding/json/v2/doc.go

     1  // Copyright 2020 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  //go:build goexperiment.jsonv2
     6  
     7  // Package json implements semantic processing of JSON as specified in RFC 8259.
     8  // JSON is a simple data interchange format that can represent
     9  // primitive data types such as booleans, strings, and numbers,
    10  // in addition to structured data types such as objects and arrays.
    11  //
    12  // This package (encoding/json/v2) is experimental,
    13  // and not subject to the Go 1 compatibility promise.
    14  // It only exists when building with the GOEXPERIMENT=jsonv2 environment variable set.
    15  // Most users should use [encoding/json].
    16  //
    17  // [Marshal] and [Unmarshal] encode and decode Go values
    18  // to/from JSON text contained within a []byte.
    19  // [MarshalWrite] and [UnmarshalRead] operate on JSON text
    20  // by writing to or reading from an [io.Writer] or [io.Reader].
    21  // [MarshalEncode] and [UnmarshalDecode] operate on JSON text
    22  // by encoding to or decoding from a [jsontext.Encoder] or [jsontext.Decoder].
    23  // [Options] may be passed to each of the marshal or unmarshal functions
    24  // to configure the semantic behavior of marshaling and unmarshaling
    25  // (i.e., alter how JSON data is understood as Go data and vice versa).
    26  // [jsontext.Options] may also be passed to the marshal or unmarshal functions
    27  // to configure the syntactic behavior of encoding or decoding.
    28  //
    29  // The data types of JSON are mapped to/from the data types of Go based on
    30  // the closest logical equivalent between the two type systems. For example,
    31  // a JSON boolean corresponds with a Go bool,
    32  // a JSON string corresponds with a Go string,
    33  // a JSON number corresponds with a Go int, uint or float,
    34  // a JSON array corresponds with a Go slice or array, and
    35  // a JSON object corresponds with a Go struct or map.
    36  // See the documentation on [Marshal] and [Unmarshal] for a comprehensive list
    37  // of how the JSON and Go type systems correspond.
    38  //
    39  // Arbitrary Go types can customize their JSON representation by implementing
    40  // [Marshaler], [MarshalerTo], [Unmarshaler], or [UnmarshalerFrom].
    41  // This provides authors of Go types with control over how their types are
    42  // serialized as JSON. Alternatively, users can implement functions that match
    43  // [MarshalFunc], [MarshalToFunc], [UnmarshalFunc], or [UnmarshalFromFunc]
    44  // to specify the JSON representation for arbitrary types.
    45  // This provides callers of JSON functionality with control over
    46  // how any arbitrary type is serialized as JSON.
    47  //
    48  // # JSON Representation of Go structs
    49  //
    50  // A Go struct is naturally represented as a JSON object,
    51  // where each Go struct field corresponds with a JSON object member.
    52  // When marshaling, all Go struct fields are recursively encoded in depth-first
    53  // order as JSON object members except those that are ignored or omitted.
    54  // When unmarshaling, JSON object members are recursively decoded
    55  // into the corresponding Go struct fields.
    56  // Object members that do not match any struct fields,
    57  // also known as “unknown members”, are ignored by default or rejected
    58  // if [RejectUnknownMembers] is specified.
    59  //
    60  // The representation of each struct field can be customized in the
    61  // "json" struct field tag, where the tag is a comma separated list of options.
    62  // As a special case, if the entire tag is `json:"-"`,
    63  // then the field is ignored with regard to its JSON representation.
    64  // Some options also have equivalent behavior controlled by a caller-specified [Options].
    65  // Field-specified options take precedence over caller-specified options.
    66  //
    67  // The first option is the JSON object name override for the Go struct field.
    68  // If the name is not specified, then the Go struct field name
    69  // is used as the JSON object name. JSON names containing commas or quotes,
    70  // or names identical to "" or "-", can be specified using
    71  // a single-quoted string literal, where the syntax is identical to
    72  // the Go grammar for a double-quoted string literal,
    73  // but instead uses single quotes as the delimiters.
    74  // By default, unmarshaling uses case-sensitive matching to identify
    75  // the Go struct field associated with a JSON object name.
    76  //
    77  // After the name, the following tag options are supported:
    78  //
    79  //   - omitzero: When marshaling, the "omitzero" option specifies that
    80  //     the struct field should be omitted if the field value is zero
    81  //     as determined by the "IsZero() bool" method if present,
    82  //     otherwise based on whether the field is the zero Go value.
    83  //     This option has no effect when unmarshaling.
    84  //
    85  //   - omitempty: When marshaling, the "omitempty" option specifies that
    86  //     the struct field should be omitted if the field value would have been
    87  //     encoded as a JSON null, empty string, empty object, or empty array.
    88  //     This option has no effect when unmarshaling.
    89  //
    90  //   - string: The "string" option specifies that [StringifyNumbers]
    91  //     be set when marshaling or unmarshaling a struct field value.
    92  //     This causes numeric types to be encoded as a JSON number
    93  //     within a JSON string, and to be decoded from a JSON string
    94  //     containing the JSON number without any surrounding whitespace.
    95  //     This extra level of encoding is often necessary since
    96  //     many JSON parsers cannot precisely represent 64-bit integers.
    97  //
    98  //   - case: When unmarshaling, the "case" option specifies how
    99  //     JSON object names are matched with the JSON name for Go struct fields.
   100  //     The option is a key-value pair specified as "case:value" where
   101  //     the value must either be 'ignore' or 'strict'.
   102  //     The 'ignore' value specifies that matching is case-insensitive
   103  //     where dashes and underscores are also ignored. If multiple fields match,
   104  //     the first declared field in breadth-first order takes precedence.
   105  //     The 'strict' value specifies that matching is case-sensitive.
   106  //     This takes precedence over the [MatchCaseInsensitiveNames] option.
   107  //
   108  //   - inline: The "inline" option specifies that
   109  //     the JSON representable content of this field type is to be promoted
   110  //     as if they were specified in the parent struct.
   111  //     It is the JSON equivalent of Go struct embedding.
   112  //     A Go embedded field is implicitly inlined unless an explicit JSON name
   113  //     is specified. The inlined field must be a Go struct
   114  //     (that does not implement any JSON methods), [jsontext.Value],
   115  //     map[~string]T, or an unnamed pointer to such types. When marshaling,
   116  //     inlined fields from a pointer type are omitted if it is nil.
   117  //     Inlined fields of type [jsontext.Value] and map[~string]T are called
   118  //     “inlined fallbacks” as they can represent all possible
   119  //     JSON object members not directly handled by the parent struct.
   120  //     Only one inlined fallback field may be specified in a struct,
   121  //     while many non-fallback fields may be specified. This option
   122  //     must not be specified with any other option (including the JSON name).
   123  //
   124  //   - unknown: The "unknown" option is a specialized variant
   125  //     of the inlined fallback to indicate that this Go struct field
   126  //     contains any number of unknown JSON object members. The field type must
   127  //     be a [jsontext.Value], map[~string]T, or an unnamed pointer to such types.
   128  //     If [DiscardUnknownMembers] is specified when marshaling,
   129  //     the contents of this field are ignored.
   130  //     If [RejectUnknownMembers] is specified when unmarshaling,
   131  //     any unknown object members are rejected regardless of whether
   132  //     an inlined fallback with the "unknown" option exists. This option
   133  //     must not be specified with any other option (including the JSON name).
   134  //
   135  //   - format: The "format" option specifies a format flag
   136  //     used to specialize the formatting of the field value.
   137  //     The option is a key-value pair specified as "format:value" where
   138  //     the value must be either a literal consisting of letters and numbers
   139  //     (e.g., "format:RFC3339") or a single-quoted string literal
   140  //     (e.g., "format:'2006-01-02'"). The interpretation of the format flag
   141  //     is determined by the struct field type.
   142  //
   143  // The "omitzero" and "omitempty" options are mostly semantically identical.
   144  // The former is defined in terms of the Go type system,
   145  // while the latter in terms of the JSON type system.
   146  // Consequently they behave differently in some circumstances.
   147  // For example, only a nil slice or map is omitted under "omitzero", while
   148  // an empty slice or map is omitted under "omitempty" regardless of nilness.
   149  // The "omitzero" option is useful for types with a well-defined zero value
   150  // (e.g., [net/netip.Addr]) or have an IsZero method (e.g., [time.Time.IsZero]).
   151  //
   152  // Every Go struct corresponds to a list of JSON representable fields
   153  // which is constructed by performing a breadth-first search over
   154  // all struct fields (excluding unexported or ignored fields),
   155  // where the search recursively descends into inlined structs.
   156  // The set of non-inlined fields in a struct must have unique JSON names.
   157  // If multiple fields all have the same JSON name, then the one
   158  // at shallowest depth takes precedence and the other fields at deeper depths
   159  // are excluded from the list of JSON representable fields.
   160  // If multiple fields at the shallowest depth have the same JSON name,
   161  // but exactly one is explicitly tagged with a JSON name,
   162  // then that field takes precedence and all others are excluded from the list.
   163  // This is analogous to Go visibility rules for struct field selection
   164  // with embedded struct types.
   165  //
   166  // Marshaling or unmarshaling a non-empty struct
   167  // without any JSON representable fields results in a [SemanticError].
   168  // Unexported fields must not have any `json` tags except for `json:"-"`.
   169  //
   170  // # Security Considerations
   171  //
   172  // JSON is frequently used as a data interchange format to communicate
   173  // between different systems, possibly implemented in different languages.
   174  // For interoperability and security reasons, it is important that
   175  // all implementations agree upon the semantic meaning of the data.
   176  //
   177  // [For example, suppose we have two micro-services.]
   178  // The first service is responsible for authenticating a JSON request,
   179  // while the second service is responsible for executing the request
   180  // (having assumed that the prior service authenticated the request).
   181  // If an attacker were able to maliciously craft a JSON request such that
   182  // both services believe that the same request is from different users,
   183  // it could bypass the authenticator with valid credentials for one user,
   184  // but maliciously perform an action on behalf of a different user.
   185  //
   186  // According to RFC 8259, there unfortunately exist many JSON texts
   187  // that are syntactically valid but semantically ambiguous.
   188  // For example, the standard does not define how to interpret duplicate
   189  // names within an object.
   190  //
   191  // The v1 [encoding/json] and [encoding/json/v2] packages
   192  // interpret some inputs in different ways. In particular:
   193  //
   194  //   - The standard specifies that JSON must be encoded using UTF-8.
   195  //     By default, v1 replaces invalid bytes of UTF-8 in JSON strings
   196  //     with the Unicode replacement character,
   197  //     while v2 rejects inputs with invalid UTF-8.
   198  //     To change the default, specify the [jsontext.AllowInvalidUTF8] option.
   199  //     The replacement of invalid UTF-8 is a form of data corruption
   200  //     that alters the precise meaning of strings.
   201  //
   202  //   - The standard does not specify a particular behavior when
   203  //     duplicate names are encountered within a JSON object,
   204  //     which means that different implementations may behave differently.
   205  //     By default, v1 allows for the presence of duplicate names,
   206  //     while v2 rejects duplicate names.
   207  //     To change the default, specify the [jsontext.AllowDuplicateNames] option.
   208  //     If allowed, object members are processed in the order they are observed,
   209  //     meaning that later values will replace or be merged into prior values,
   210  //     depending on the Go value type.
   211  //
   212  //   - The standard defines a JSON object as an unordered collection of name/value pairs.
   213  //     While ordering can be observed through the underlying [jsontext] API,
   214  //     both v1 and v2 generally avoid exposing the ordering.
   215  //     No application should semantically depend on the order of object members.
   216  //     Allowing duplicate names is a vector through which ordering of members
   217  //     can accidentally be observed and depended upon.
   218  //
   219  //   - The standard suggests that JSON object names are typically compared
   220  //     based on equality of the sequence of Unicode code points,
   221  //     which implies that comparing names is often case-sensitive.
   222  //     When unmarshaling a JSON object into a Go struct,
   223  //     by default, v1 uses a (loose) case-insensitive match on the name,
   224  //     while v2 uses a (strict) case-sensitive match on the name.
   225  //     To change the default, specify the [MatchCaseInsensitiveNames] option.
   226  //     The use of case-insensitive matching provides another vector through
   227  //     which duplicate names can occur. Allowing case-insensitive matching
   228  //     means that v1 or v2 might interpret JSON objects differently from most
   229  //     other JSON implementations (which typically use a case-sensitive match).
   230  //
   231  //   - The standard does not specify a particular behavior when
   232  //     an unknown name in a JSON object is encountered.
   233  //     When unmarshaling a JSON object into a Go struct, by default
   234  //     both v1 and v2 ignore unknown names and their corresponding values.
   235  //     To change the default, specify the [RejectUnknownMembers] option.
   236  //
   237  //   - The standard suggests that implementations may use a float64
   238  //     to represent a JSON number. Consequently, large JSON integers
   239  //     may lose precision when stored as a floating-point type.
   240  //     Both v1 and v2 correctly preserve precision when marshaling and
   241  //     unmarshaling a concrete integer type. However, even if v1 and v2
   242  //     preserve precision for concrete types, other JSON implementations
   243  //     may not be able to preserve precision for outputs produced by v1 or v2.
   244  //     The `string` tag option can be used to specify that an integer type
   245  //     is to be quoted within a JSON string to avoid loss of precision.
   246  //     Furthermore, v1 and v2 may still lose precision when unmarshaling
   247  //     into an any interface value, where unmarshal uses a float64
   248  //     by default to represent a JSON number.
   249  //     To change the default, specify the [WithUnmarshalers] option
   250  //     with a custom unmarshaler that pre-populates the interface value
   251  //     with a concrete Go type that can preserve precision.
   252  //
   253  // RFC 8785 specifies a canonical form for any JSON text,
   254  // which explicitly defines specific behaviors that RFC 8259 leaves undefined.
   255  // In theory, if a text can successfully [jsontext.Value.Canonicalize]
   256  // without changing the semantic meaning of the data, then it provides a
   257  // greater degree of confidence that the data is more secure and interoperable.
   258  //
   259  // The v2 API generally chooses more secure defaults than v1,
   260  // but care should still be taken with large integers or unknown members.
   261  //
   262  // [For example, suppose we have two micro-services.]: https://www.youtube.com/watch?v=avilmOcHKHE&t=1057s
   263  package json
   264  
   265  // requireKeyedLiterals can be embedded in a struct to require keyed literals.
   266  type requireKeyedLiterals struct{}
   267  
   268  // nonComparable can be embedded in a struct to prevent comparability.
   269  type nonComparable [0]func()
   270  

View as plain text