Fast go json parser wrapping C++ rapidjson
Latest version tested for go1.7.4 and go1.8
In package main:
import "github.com/amosmzhang/rapidjson"
...
json, err := rapidjson.NewParsedStringJson(`{"project":"rapidjson","stars":10,"use":"everywhere"}`)
...
fmt.Println(json.String())
rapidjson has two types: Container and Doc. Container is a generalized value type, and can take on any specific type (int, array, object, etc.). Doc is a specialized Container with additional parsing and memory allocation functionality. In general, create one Doc, get its Container, and work with that Container. Doc should be freed manually, but Containers associated with a Doc will be freed when Doc is freed. Key values pairs in rapidjson objects are referred to as members.
func (json *Doc) Parse(input []byte) error
func (json *Doc) ParseString(input string) error
func NewParsedJson(input []byte) (*Doc, error)
func NewParsedStringJson(input string) (*Doc, error)
func (json *Doc) HasParseError() bool
A call to HasParseError() is included at the end of each parsing func, and error is returned with details.
For outputting:
func (json *Doc) String() string
func (json *Doc) Pretty() string
func (json *Doc) Bytes() []byte
func (ct *Container) String() string
func (ct *Container) Pretty() string
func (ct *Container) Bytes() []byte
Getting a Doc's Container:
func (json *Doc) GetContainer() *Container
func (json *Doc) GetContainerNewObj() *Container
Working with Containers:
func (ct *Container) HasMember(key string) bool
func (ct *Container) GetMemberCount() (int, error)
func (ct *Container) GetMemberName(index int) string
func (ct *Container) GetMemberNames() ([]string, error)
func (ct *Container) GetMemberMap() (map[string]*Container, error)
func (ct *Container) GetMember(key string) (*Container, error)
func (ct *Container) GetPathContainer(path string) (*Container, error)
func (ct *Container) GetPathNewContainer(path string) (*Container, error)
func (ct *Container) IsEqual(other *Container) bool
Typed getters:
func (ct *Container) GetType() int
func (ct *Container) GetInt() (int, error)
func (ct *Container) GetInt64() (int64, error)
func (ct *Container) GetFloat() (float64, error)
func (ct *Container) GetBool() (bool, error)
func (ct *Container) GetString() (string, error)
func (ct *Container) GetArraySize() (int, error)
func (ct *Container) GetArrayValue(index int) *Container
func (ct *Container) GetIntArray() ([]int, error)
func (ct *Container) GetStringArray() ([]string, error)
func (ct *Container) GetArray() ([]*Container, int, error)
func (ct *Container) GetValue() (interface{}, error)
Making new Container, use root Doc to use memory allocator. Freeing Doc will free associated Containers:
func (json *Doc) NewContainer() *Container
func (json *Doc) NewContainerObj() *Container
func (json *Doc) NewContainerArray() *Container
SetValue() can be used for basic types int (all sizes), float64, bool, string, nil. Setters will overwrite previous type.
func (ct *Container) SetValue(v interface{}) error
func (ct *Container) SetContainer(item *Container)
func (ct *Container) SetContainerCopy(item *Container)
func (ct *Container) InitObj()
func (ct *Container) AddValue(key string, v interface{}) error
func (ct *Container) AddMember(key string, item *Container) error
func (ct *Container) AddMemberCopy(key string, item *Container) error
func (ct *Container) AddMemberArray(key string, items []*Container) error
func (ct *Container) AddMemberAtPath(path string, item *Container) error
func (ct *Container) AddValueAtPath(path string, v interface{}) error
func (ct *Container) SetMember(key string, item *Container) error
func (ct *Container) SetMemberCopy(key string, item *Container) error
func (ct *Container) SetMemberValue(key string, v interface{}) error
func (ct *Container) InitArray()
func (ct *Container) ArrayAppendContainer(item *Container) error
func (ct *Container) ArrayAppendCopy(item *Container) error
func (ct *Container) ArrayAppend(v interface{}) error
This set of functions duplicate functionality in some previous functions, but do not return errors so that they can be chained.
func (ct *Container) GetMemberCountOrNil() int
func (ct *Container) GetMemberNamesOrNil() []string
func (ct *Container) GetMemberMapOrNil() map[string]*Container
func (ct *Container) GetMemberOrNil(key string) *Container
func (ct *Container) GetPathContainerOrNil(path string) *Container
func (ct *Container) GetIntArrayOrNil() []int
func (ct *Container) GetStringOrNil() []string
func (ct *Container) GetArrayOrNil() []*Container
Usage example:
example, err := container.GetMemberOrNil("path1").GetMemberOrNil("path2").GetString()
func (ct *Container) RemoveMember(key string) error
func (ct *Container) ArrayClear() error
func (ct *Container) ArrayRemove(index int) error
func (ct *Container) RemoveMemberAtPath(path string) error
func (ct *Container) StripNulls(leaveEmptyArray bool) *Container
TypeNull = 0
TypeFalse = 1
TypeTrue = 2
TypeObject = 3
TypeArray = 4
TypeString = 5
TypeNumber = 6
ErrNotArray - Not an array
ErrNotObject - Not an object
ErrPathNotFound - Path not found
ErrNotInt - Not an int
ErrNotFloat - Not a float
ErrNotBool - Not a bool
ErrNotString - Not a string
ErrBadType - Bad type
ErrMemberExists - Member already exists
ErrOutOfBounds - Array index out of bounds
The following benchmark test was performed by reading 100,000 lines of JSON from a file (~1.5kb per line), and parsing each with Go's encoding/json as well as rapidjson. One binary was built using each library (all other code was identical) and both were run 5 times on a Mid 2015 Macbook Pro (2.8 GHz Intel Core i7).
RJ took a full second less to complete this operation, running in about 59% of the time it took Go's library to do the same task.
Using rapidjson: (average total 2.0018s)
$ time ./rj
100000 documents parsed
./rj 2.10s user 0.29s system 116% cpu 2.047 total
$ time ./rj
100000 documents parsed
./rj 2.05s user 0.27s system 118% cpu 1.958 total
$ time ./rj
100000 documents parsed
./rj 2.11s user 0.27s system 118% cpu 2.010 total
$ time ./rj
100000 documents parsed
./rj 2.07s user 0.27s system 117% cpu 1.985 total
$ time ./rj
100000 documents parsed
./rj 2.09s user 0.27s system 117% cpu 2.009 total
Using encoding/json: (average total 3.3674s)
$ time ./json
100000 documents parsed
./json 3.40s user 0.12s system 105% cpu 3.335 total
$ time ./json
100000 documents parsed
./json 3.46s user 0.12s system 105% cpu 3.374 total
$ time ./json
100000 documents parsed
./json 3.44s user 0.11s system 105% cpu 3.349 total
$ time ./json
100000 documents parsed
./json 3.50s user 0.12s system 105% cpu 3.412 total
$ time ./json
100000 documents parsed
./json 3.45s user 0.11s system 105% cpu 3.367 total