OpenAPI 3.0 and 3.1 and 3.2 (and Swagger v2) implementation for Go (parsing, converting, validation, and more)
OpenAPI 3.0 and 3.1 and 3.2 (and Swagger v2) implementation for Go (parsing, converting, validation, and more)
A Go project for handling OpenAPI files. We target:
v2.0 (formerly known as Swagger)v3.0v3.1v3.2 Partially: Media Type Object itemSchema, Path Item Object query (the HTTP QUERY method) and additionalOperations (custom HTTP methods).Licensed under the MIT License.
The project has received pull requests from many people. Thanks to everyone!
Please, give back to this project by becoming a sponsor.
Here's some projects that depend on kin-openapi:
net/http"Be sure to check OpenAPI Initiative's great tooling list as well as OpenAPI.Tools.
*openapi3.Schema values for Go types.go run github.com/getkin/kin-openapi/cmd/validate@latest [--defaults] [--examples] [--ext] [--patterns] -- <local YAML or JSON file>
Use openapi3.Loader, which resolves all references:
loader := openapi3.NewLoader()
doc, err := loader.LoadFromFile("my-openapi-spec.json")
When IncludeOrigin is enabled, the loader records the file, line, and column of each element in the OpenAPI document. This is useful for tools that need to report errors or changes with precise source locations (e.g. linters, diff tools, editors).
loader := openapi3.NewLoader()
loader.IncludeOrigin = true
doc, err := loader.LoadFromFile("my-openapi-spec.json")
// Each element has an Origin field with source location info
fmt.Println(doc.Info.Origin.Key.File) // "my-openapi-spec.json"
fmt.Println(doc.Info.Origin.Key.Line) // 2
fmt.Println(doc.Info.Origin.Key.Column) // 1
The Origin struct contains three parts:
Key — the location of the object itself (file, line, column).Fields — locations of scalar fields within the object (e.g. origin.Fields["description"] gives the line of the description field).Sequences — locations of items in sequence-valued fields. For example, origin.Sequences["enum"] gives the location of each item in an enum array. This is used for fields like enum, required, and servers where the individual items are scalars and don't have their own Origin field.Origin data is populated by an internal post-processing step after YAML decoding — it is not part of the OpenAPI spec itself. For this reason, Origin fields are excluded from serialization. If you marshal a loaded document back to JSON/YAML, origin data will not appear in the output.
Each validation error carries a stable, kebab-case code (e.g. operation-responses-required), independent of the message text, so tools can suppress specific findings, assign per-rule severities, or emit machine-readable diagnostics. The full catalog is available from openapi3.ValidationErrorCodes().
err := doc.Validate(ctx, openapi3.EnableMultiError())
for _, e := range err.(openapi3.MultiError) {
var coded openapi3.CodedError
if errors.As(e, &coded) {
fmt.Println(coded.Code(), e) // e.g. "operation-responses-required value of responses must be an object"
}
}
loader := openapi3.NewLoader()
doc, _ := loader.LoadFromData([]byte(`...`))
_ = doc.Validate(loader.Context)
router, _ := gorillamux.NewRouter(doc)
route, pathParams, _ := router.FindRoute(httpRequest)
// Do something with route.Operation
…
By default, the library parses a body of the HTTP request and response of a few content types e.g. "text/plain" or "application/json".
To support other content types you must register decoders for them:
…
By default, the library checks unique items using the following predefined function:
func isSliceOfUniqueItems(xs []any) bool {
s := len(xs)
m := make(map[string]struct{}, s)
for _, x := range xs {
key, _ := json.Marshal(&x)
m[string(key)] = struct{}{}
}
return s == len(m)
}
In the predefined function json.Marshal is used to generate a string that can
be used as a map key which is to check the uniqueness of an array
when the array items are objects or arrays. You can register
you own function according to your input data to get better performance:
func main() {
// ...
// Register a customized function used to check uniqueness of array.
openapi3.RegisterArrayUniqueItemsChecker(arrayUniqueItemsChecker)
// ... other validate codes
}
func arrayUniqueItemsChecker(items []any) bool {
// Check the uniqueness of the input slice
}
By default, the error message returned when validating a value includes the error reason, the schema, and the input value.
For example, given the following schema:
{
"type": "string",
"allOf": [
{ "pattern": "[A-Z]" },
{ "pattern": "[a-z]" },
{ "pattern": "[0-9]" },
{ "pattern": "[!@#$%^&*()_+=-?~]" }
]
}
Passing the input value "secret" to this schema will produce the following error message:
string doesn't match the regular expression "[A-Z]"
Schema:
{
"pattern": "[A-Z]"
}
Value:
"secret"
Including the original value in the error message can be helpful for debugging, but it may not be appropriate for sensitive information such as secrets.
To disable the extra details in the schema error message, you can set the openapi3.SchemaErrorDetailsDisabled option to true:
func main() {
// ...
// Disable schema error detailed error messages
openapi3.SchemaErrorDetailsDisabled = true
// ... other validate codes
}
This will shorten the error message to present only the reason:
string doesn't match the regular expression "[A-Z]"
For more fine-grained control over the error message, you can pass a custom openapi3filter.Options object to openapi3filter.RequestValidationInput that includes a openapi3filter.CustomSchemaErrorFunc.
func validationOptions() *openapi3filter.Options {
options := &openapi3filter.Options{}
options.WithCustomSchemaErrorFunc(safeErrorMessage)
return options
}
func safeErrorMessage(err *openapi3.SchemaError) string {
return err.Reason
}
This will change the schema validation errors to return only the Reason field, which is guaranteed to not include the original value.
ReferencesComponentInRootDocument is a useful helper function to check if a component reference
coincides with a reference in the root document's component objects fixed fields.
This can be used to determine if two schema definitions are of the same structure, helpful for code generation tools when generating go type models.
doc, err = loader.LoadFromFile("openapi.yml")
for _, path := range doc.Paths.InMatchingOrder() {
pathItem := doc.Paths.Find(path)
if pathItem.Get == nil || pathItem.Get.Responses.Status(200) {
continue
}
for _, s := range pathItem.Get.Responses.Status(200).Value.Content {
name, match := ReferencesComponentInRootDocument(doc, s.Schema)
fmt.Println(path, match, name) // /record true #/components/schemas/BookRecord
}
}
(*openapi3.PathItem).SetOperation(string, *Operation) no longer panics on unhandled HTTP methods: these are now stored in the new openapi3.PathItem.AdditionalOperations field (passing a nil operation deletes the entry).(*openapi3.PathItem).GetOperation(string) and (*openapi3.PathItem).Operations() now also report the new openapi3.PathItem.Query field (the OpenAPI 3.2 HTTP QUERY method) and the `openapi3.PathItem.Ad$ref sibling fields cause validation to fail
unevaluatedProperties silently ignored: 2020-12 validator falls back to the built-in one whenever a schema contains an internal $ref
please user OrderedMap instead of the golang's map
Path parameter name/template correspondence not checked when counts match
ValidateParameter: injecting query default with q.Add duplicates values when param is present but empty/unparseable
openapi3: decoding allocates ~65x the document size, mostly re-parsing JSON it just generated