Fizz is a wrapper for Gin based on gadgeto/tonic.
It generates wrapping gin-compatible handlers that do all the repetitive work and wrap the call to your handlers. It can also generates an *almost* complete OpenAPI 3 specification of your API.
To create a Fizz instance, you can pass an existing Gin engine to fizz.NewFromEngine
, or use fizz.New
that will use a new default Gin engine.
engine := gin.Default()
engine.Use(...) // register global middlewares
f := fizz.NewFromEngine(engine)
A Fizz instance implements the http.HandlerFunc
interface, which means it can be used as the base handler of your HTTP server.
srv := &http.Server{
Addr: ":4242",
Handler: f,
}
srv.ListenAndServe()
Fizz abstracts the GET
, POST
, PUT
, PATCH
, DELETE
, OPTIONS
, HEAD
and TRACE
methods of a Gin engine. These functions accept a variadic list of handlers as the last parameter, but since Fizz relies on tonic to retrieve the informations required to generate the OpenAPI specification of the operation, only one of the handlers registered MUST be wrapped with Tonic.
In the following example, the BarHandler
is a simple middleware that will be executed before the FooHandler
, but the generator will use the input/output type of the FooHandler
to generate the specification of the operation.
func BarHandler(c *gin.Context) { ... }
func FooHandler(*gin.Context, *Foo) (*Bar, error) { ... }
fizz := fizz.New()
fizz.GET("/foo/bar", nil, BarHandler, tonic.Handler(FooHandler, 200))
However, registering only standard handlers that follow the gin.HandlerFunc
signature is accepted, but the OpenAPI generator will ignore the operation and it won't appear in the specification.
To enrich an operation, you can pass a list of optional OperationOption
functions as the second parameters of the GET
, POST
, PUT
, PATCH
, DELETE
, OPTIONS
and HEAD
methods.
// Set the default response description.
// A default status text will be created from the code if it is omitted.
fizz.StatusDescription(desc string)
// Set the summary of the operation.
fizz.Summary(summary string)
fizz.Summaryf(format string, a ...interface{})
// Set the description of the operation.
fizz.Description(desc string)
fizz.Descriptionf(format string, a ...interface{})
// Override the ID of the operation.
// Must be a unique string used to identify the operation among
// all operations described in the API.
fizz.ID(id string)
// Mark the operation as deprecated.
fizz.Deprecated(deprecated bool)
// Add an additional response to the operation.
// model and header may be `nil`.
fizz.Response(statusCode, desc string, model interface{}, headers []*ResponseHeader)
// Add an additional header to the default response.
// Model can be of any type, and may also be `nil`,
// in which case the string type will be used as default.
fizz.Header(name, desc string, model interface{})
// Override the binding model of the operation.
fizz.InputModel(model interface{})
NOTES:
fizz.InputModel
allows to override the operation input regardless of how the handler implementation really binds the request parameters. It is the developer responsibility to ensure that the binding matches the OpenAPI specification.- The fist argument of the
fizz.Reponse
method which represents an HTTP status code is of type string because the spec accept the valuedefault
. See the Responses Object documentation for more informations.
To help you declare additional headers, predefined variables for Go primitives types that you can use as the third argument of the fizz.Header
method are available.
Integer int32
Long int64
Float float32
Double float64
String string
Byte []byte
Binary []byte
Boolean bool
DateTime time.Time
Exactly like you would do with Gin, you can create a group of routes using the method Group
. Unlike Gin own method, Fizz's one takes two other optional arguments, name
and description
. These parameters will be used to create a tag in the OpenAPI specification that will be applied to all the routes added to the group.
grp := f.Group("/subpath", "MyGroup", "Group description", middlewares...)
If the name
parameter is empty, the tag won't be created and it won't be used.
Subgroups of subgroups can be created to an infinite depth, according yo your needs.
foo := f.Group("/foo", "Foo", "Foo group")
// all routes registered on group bar will have
// a relative path starting with /foo/bar
bar := f.Group("/bar", "Bar", "Bar group")
// /foo/bar/{barID}
bar.GET("/:barID", nil, tonic.Handler(MyBarHandler, 200))
The Use
method can be used with groups to register middlewares after their creation.
grp.Use(middleware1, middleware2, ...)
The subpackage tonic handles path/query/header/body parameters binding in a single consolidated input object which allows you to remove all the boilerplate code that retrieves and tests the presence of various parameters. The OpenAPI generator make use of the input/output types informations of a tonic-wrapped handler reported by tonic to document the operation in the specification.
The handlers wrapped with tonic must follow the following signature.
func(*gin.Context, [input object ptr]) ([output object], error)
Input and output objects are both optional, as such, the minimal accepted signature is:
func(*gin.Context) error
To wrap a handler with tonic, use the tonic.Handler
method. It takes a function that follow the above signature and a default status code and return a gin.HandlerFunc
function that can be used when you register a route with Fizz of Gin.
Output objects can be of any type, and will be marshalled to the desired media type. Note that the input object MUST always be a pointer to a struct, or the tonic wrapping will panic at runtime.
If you use closures as handlers, please note that they will all have the same name, and the generator will return an error. To overcome this problem, you have to explicitely set the ID of an operation when you register the handler.
func MyHandler() gin.HandlerFunc {
return tonic.Handler(func(c *gin.Context) error {}, 200)
}
fizz.GET("/foo", []fizz.OperationOption{
fizz.ID("MyOperationID")
}, MyHandler())
tonic uses three struct tags to recognize the parameters it should bind to the input object of your tonic-wrapped handlers:
path
: bind from the request pathquery
: bind from the query stringheader
: bind from the request header
The fields that doesn't use one of these tags will be considered as part of the request body.
The value of each struct tag represents the name of the field in each location, with options.
type MyHandlerParams struct {
ID int64 `path:"id"`
Foo string `query:"foo"`
Bar time.Time `header:"x-foo-bar"`
}
tonic will automatically convert the value extracted from the location described by the tag to the appropriate type before binding.
NOTE: A path parameter is always required and will appear required in the spec regardless of the validate
tag content.
You can use additional tags. Some will be interpreted by tonic, others will be exclusively used to enrich the OpenAPI specification.
default
: tonic will bind this value if none was passed with the request. This should not be used if a field is also required. Read the documentation (section Common Mistakes) for more informations about this behaviour.description
: Add a description of the field in the spec.deprecated
: Indicates if the field is deprecated. Accepted values are 1, t, T, TRUE, true, True, 0, f, F, FALSE. Invalid value are considered to be false.enum
: A coma separated list of acceptable values for the parameter.format
: Override the format of the field in the specification. Read the documentation for more informations.validate
: Field validation rules. Read the documentation for more informations.explode
: Specifies whether arrays should generate separate parameters for each array item or object property (limited to query parameters with form style). Accepted values are 1, t, T, TRUE, true, True, 0, f, F, FALSE. Invalid value are considered to be false.
The JSON/XML encoders usually omit a field that has the tag "-"
. This behaviour is reproduced by the OpenAPI generator ; a field with this tag won't appear in the properties of the schema.
In the following example, the field Input
is used only for binding request body parameters and won't appear in the output encoding while Output
will be marshaled but will not be used for parameters binding.
type Model struct {
Input string `json:"-"`
Output string `json:"output" binding:"-"`
}
If you want to make a request body field mandatory, you can use the tag validate:"required"
. The validator used by tonic will ensure that the field is present.
To be able to make a difference between a missing value and the zero value of a type, use a pointer.
To explicitly ignore a parameter from the request body, use the tag binding:"-"
.
Note that the OpenAPI generator will ignore request body parameters for the routes with a method that is one of GET
, DELETE
or HEAD
.
GET, DELETE and HEAD are no longer allowed to have request body because it does not have defined semantics as per RFC 7231. source
The OpenAPI generator recognize some tags of the go-playground/validator.v8 package and translate those to the properties of the schema that are taken from the JSON Schema definition.
The supported tags are: len, max, min, eq, gt, gte, lt, lte.
Based on the type of the field that carry the tag, the fields maximum
, minimum
, minLength
, maxLength
, minIntems
, maxItems
, minProperties
and maxProperties
of its JSON Schema will be filled accordingly.
To serve the generated OpenAPI specification in either JSON
or YAML
format, use the handler returned by the fizz.OpenAPI
method.
To enrich the specification, you can provide additional informations. Head to the OpenAPI 3 spec for more informations about the API informations that you can specify, or take a look at the type openapi.Info
in the file openapi/spec.go.
infos := &openapi.Info{
Title: "Fruits Market",
Description: `This is a sample Fruits market server.`,
Version: "1.0.0",
}
f.GET("/openapi.json", nil, fizz.OpenAPI(infos, "json"))
NOTE: The generator will never panic. However, it is strongly recommended to call fizz.Errors
to retrieve and handle the errors that may have occured during the generation of the specification before starting your API.
The output types of your handlers are registered as components within the generated specification. By default, the name used for each component is composed of the package and type name concatenated using CamelCase style, and does not contain the full import path. As such, please ensure that you don't use the same type name in two eponym package in your application.
The names of the components can be customized in two different ways.
Override the name of a type globally before registering your handlers. This has the highest precedence.
f := fizz.New()
f.Generator().OverrideTypeName(reflect.TypeOf(T{}), "OverridedName")
Implements the openapi.Typer
interface on your types.
func (*T) TypeName() string { return "OverridedName" }
WARNING: You MUST not rely on the method receiver to return the name, because the method will be called on a new instance created by the generator with the reflect
package.
The spec generator creates OpenAPI schemas for your types based on their reflection kind.
If you want to control the output schema of a type manually, you can implement the DataType
interface for this type.
For example, given a UUID version 4 type, declared as a struct, that should appear as a string with a custom format.
type UUIDv4 struct { ... }
func (*UUIDv4) Format() string { return "uuid" }
func (*UUIDv4) Type() string { return "string" }
The schema of the type will look like the following instead of describing all the fields of the struct.
{
"type": "string",
"format": "uuid"
}
WARNING: You MUST not rely on the method receivers to return the type and format, because these methods will be called on a new instance created by the generator with the reflect
package.
You can also override manually the type and format using OverrideDataType()
. This has the highest precedence.
fizz.Generator().OverrideDataType(reflect.TypeOf(&UUIDv4{}), "string", "uuid")
Fizz supports some native and imported types. A schema with a proper type and format will be generated automatically, removing the need for creating your own custom schema.
- time.Time
- time.Duration
- net.URL
- net.IP
Note that, according to the doc, the inherent version of the address is a semantic property, and thus cannot be determined by Fizz. Therefore, the format returned is simplyip
. If you want to specify the version, you can use the tagsformat:"ipv4"
orformat:"ipv6"
. - uuid.UUID
Throughout the specification description fields are noted as supporting CommonMark markdown formatting. Where OpenAPI tooling renders rich text it MUST support, at a minimum, markdown syntax as described by CommonMark 0.27. Tooling MAY choose to ignore some CommonMark features to address security concerns. source
To help you write markdown descriptions in Go, a simple builder is available in the sub-package markdown
. This is quite handy to avoid conflicts with backticks that are both used in Go for litteral multi-lines strings and code blocks in markdown.
- Since OpenAPI is based on the JSON Schema specification itself, objects (Go maps) with keys that are not of type
string
are not supported and will be ignored during the generation of the specification. - Recursive embedding of the same type is not supported, at any level of recursion. The generator will warn and skip the offending fields.
type A struct { Foo int *A // ko, embedded and same type as parent A *A // ok, not embedded *B // ok, different type } type B struct { Bar string *A // ko, type B is embedded in type A *C // ok, type C does not contains an embedded field of type A } type C struct { Baz bool }
A simple runnable API is available in examples/market
.
go build
./market
# Retrieve the specification marshaled in JSON.
curl -i http://localhost:4242/openapi.json
Fizz is based on gin-gonic/gin and use gadgeto/tonic. ❤️