Documentation ¶
Overview ¶
Package runtime includes helper functions for working with API objects that follow the kubernetes API object conventions, which are:
- Your API objects have a common metadata struct member, TypeMeta.
- Your code refers to an internal set of API objects.
- In a separate package, you have an external set of API objects.
- The external set is considered to be versioned, and no breaking changes are ever made to it (fields may be added but not changed or removed).
- As your api evolves, you'll make an additional versioned package with every major change.
- Versioned packages have conversion functions which convert to and from the internal version.
- You'll continue to support older versions according to your deprecation policy, and you can easily provide a program/library to update old versions into new versions because of 5.
- All of your serializations and deserializations are handled in a centralized place.
Package runtime provides a conversion helper to make 5 easy, and the Encode/Decode/DecodeInto trio to accomplish 7. You can also register additional "codecs" which use a version of your choice. It's recommended that you register your types with runtime in your package's init function.
As a bonus, a few common types useful from all api objects and versions are provided in types.go.
Index ¶
- func EncodeOrDie(codec Codec, obj Object) string
- func GetItemsPtr(list Object) (interface{}, error)
- func IsListType(obj Object) bool
- func IsNotRegisteredError(err error) bool
- func SetList(list Object, objects []Object) error
- type Codec
- type Decoder
- type EmbeddedObject
- type Encoder
- type Object
- type ObjectConvertor
- type ObjectTyper
- type PluginBase
- type RawExtension
- type ResourceVersioner
- type Scheme
- func (s *Scheme) AddConversionFuncs(conversionFuncs ...interface{}) error
- func (s *Scheme) AddKnownTypeWithName(version, kind string, obj Object)
- func (s *Scheme) AddKnownTypes(version string, types ...Object)
- func (s *Scheme) AddStructFieldConversion(srcFieldType interface{}, srcFieldName string, destFieldType interface{}, ...) error
- func (s *Scheme) Convert(in, out interface{}) error
- func (s *Scheme) ConvertToVersion(in Object, outVersion string) (Object, error)
- func (s *Scheme) Copy(obj Object) (Object, error)
- func (s *Scheme) CopyOrDie(obj Object) Object
- func (s *Scheme) DataVersionAndKind(data []byte) (version, kind string, err error)
- func (s *Scheme) Decode(data []byte) (Object, error)
- func (s *Scheme) DecodeInto(data []byte, obj Object) error
- func (s *Scheme) EncodeToVersion(obj Object, destVersion string) (data []byte, err error)
- func (s *Scheme) KnownTypes(version string) map[string]reflect.Type
- func (s *Scheme) Log(l conversion.DebugLogger)
- func (s *Scheme) New(versionName, typeName string) (Object, error)
- func (s *Scheme) ObjectVersionAndKind(obj Object) (version, kind string, err error)
- type SelfLinker
- type TypeMeta
- type Unknown
Constants ¶
This section is empty.
Variables ¶
This section is empty.
Functions ¶
func EncodeOrDie ¶
EncodeOrDie is a version of Encode which will panic instead of returning an error. For tests.
func GetItemsPtr ¶
GetItemsPtr returns a pointer to the list object's Items member. If 'list' doesn't have an Items member, it's not really a list type and an error will be returned. This function will either return a pointer to a slice, or an error, but not both.
func IsListType ¶ added in v0.5.2
func IsNotRegisteredError ¶ added in v0.7.0
IsNotRegisteredError returns true if the error indicates the provided object or input data is not registered.
Types ¶
type Decoder ¶
type Decoder interface { Decode(data []byte) (Object, error) DecodeInto(data []byte, obj Object) error }
Decoder defines methods for deserializing API objects into a given type
type EmbeddedObject ¶
type EmbeddedObject struct {
Object
}
EmbeddedObject has appropriate encoder and decoder functions, such that on the wire, it's stored as a []byte, but in memory, the contained object is accessible as an Object via the Get() function. Only valid API objects may be stored via EmbeddedObject. The purpose of this is to allow an API object of type known only at runtime to be embedded within other API objects.
Note that object assumes that you've registered all of your api types with the api package.
EmbeddedObject and RawExtension can be used together to allow for API object extensions: see the comment for RawExtension.
type Object ¶
type Object interface { // This function is used only to enforce membership. It's never called. // TODO: Consider mass rename in the future to make it do something useful. IsAnAPIObject() }
All api types must support the Object interface. It's deliberately tiny so that this is not an onerous burden. Implement it with a pointer reciever; this will allow us to use the go compiler to check the one thing about our objects that it's capable of checking for us.
func ExtractList ¶
ExtractList returns obj's Items element as an array of runtime.Objects. Returns an error if obj is not a List type (does not have an Items member).
type ObjectConvertor ¶ added in v0.5.2
type ObjectConvertor interface {
ConvertToVersion(in Object, outVersion string) (out Object, err error)
}
ObjectConvertor converts an object to a different version.
type ObjectTyper ¶ added in v0.5.2
type ObjectTyper interface { DataVersionAndKind([]byte) (version, kind string, err error) ObjectVersionAndKind(Object) (version, kind string, err error) }
ObjectTyper contains methods for extracting the APIVersion and Kind of objects.
type PluginBase ¶
type PluginBase struct {
Kind string `json:"kind,omitempty"`
}
PluginBase is like TypeMeta, but it's intended for plugin objects that won't ever be encoded except while embedded in other objects.
type RawExtension ¶
type RawExtension struct {
RawJSON []byte
}
RawExtension is used with EmbeddedObject to do a two-phase encoding of extension objects.
To use this, make a field which has RawExtension as its type in your external, versioned struct, and EmbeddedObject in your internal struct. You also need to register your various plugin types.
// Internal package:
type MyAPIObject struct { runtime.TypeMeta `json:",inline"` MyPlugin runtime.EmbeddedObject `json:"myPlugin"` }
type PluginA struct { runtime.PluginBase `json:",inline"` AOption string `json:"aOption"` }
// External package:
type MyAPIObject struct { runtime.TypeMeta `json:",inline"` MyPlugin runtime.RawExtension `json:"myPlugin"` }
type PluginA struct { runtime.PluginBase `json:",inline"` AOption string `json:"aOption"` }
// On the wire, the JSON will look something like this:
{ "kind":"MyAPIObject", "apiVersion":"v1beta1", "myPlugin": { "kind":"PluginA", "aOption":"foo", }, }
So what happens? Decode first uses json or yaml to unmarshal the serialized data into your external MyAPIObject. That causes the raw JSON to be stored, but not unpacked. The next step is to copy (using pkg/conversion) into the internal struct. The runtime package's DefaultScheme has conversion functions installed which will unpack the JSON stored in RawExtension, turning it into the correct object type, and storing it in the EmbeddedObject. (TODO: In the case where the object is of an unknown type, a runtime.Unknown object will be created and stored.)
func (RawExtension) MarshalJSON ¶
func (re RawExtension) MarshalJSON() ([]byte, error)
Marshal may get called on pointers or values, so implement MarshalJSON on value. http://stackoverflow.com/questions/21390979/custom-marshaljson-never-gets-called-in-go
func (*RawExtension) UnmarshalJSON ¶
func (re *RawExtension) UnmarshalJSON(in []byte) error
type ResourceVersioner ¶
type ResourceVersioner interface { SetResourceVersion(obj Object, version string) error ResourceVersion(obj Object) (string, error) }
ResourceVersioner provides methods for setting and retrieving the resource version from an API object.
type Scheme ¶
type Scheme struct {
// contains filtered or unexported fields
}
Scheme defines methods for serializing and deserializing API objects. It is an adaptation of conversion's Scheme for our API objects.
func NewScheme ¶
func NewScheme() *Scheme
NewScheme creates a new Scheme. This scheme is pluggable by default.
func (*Scheme) AddConversionFuncs ¶
AddConversionFuncs adds a function to the list of conversion functions. The given function should know how to convert between two API objects. We deduce how to call it from the types of its two parameters; see the comment for Converter.Register.
Note that, if you need to copy sub-objects that didn't change, it's safe to call Convert() inside your conversionFuncs, as long as you don't start a conversion chain that's infinitely recursive.
Also note that the default behavior, if you don't add a conversion function, is to sanely copy fields that have the same names. It's OK if the destination type has extra fields, but it must not remove any. So you only need to add a conversion function for things with changed/removed fields.
func (*Scheme) AddKnownTypeWithName ¶
AddKnownTypeWithName is like AddKnownTypes, but it lets you specify what this type should be encoded as. Useful for testing when you don't want to make multiple packages to define your structs.
func (*Scheme) AddKnownTypes ¶
AddKnownTypes registers the types of the arguments to the marshaller of the package api. Encode() refuses the object unless its type is registered with AddKnownTypes.
func (*Scheme) AddStructFieldConversion ¶ added in v0.6.0
func (s *Scheme) AddStructFieldConversion(srcFieldType interface{}, srcFieldName string, destFieldType interface{}, destFieldName string) error
AddStructFieldConversion allows you to specify a mechanical copy for a moved or renamed struct field without writing an entire conversion function. See the comment in conversion.Converter.SetStructFieldCopy for parameter details. Call as many times as needed, even on the same fields.
func (*Scheme) Convert ¶
Convert will attempt to convert in into out. Both must be pointers. For easy testing of conversion functions. Returns an error if the conversion isn't possible.
func (*Scheme) ConvertToVersion ¶ added in v0.5.2
ConvertToVersion attempts to convert an input object to its matching Kind in another version within this scheme. Will return an error if the provided version does not contain the inKind (or a mapping by name defined with AddKnownTypeWithName). Will also return an error if the conversion does not result in a valid Object being returned.
func (*Scheme) Copy ¶
Copy does a deep copy of an API object. Useful mostly for tests. TODO(dbsmith): implement directly instead of via Encode/Decode TODO(claytonc): Copy cannot be used for objects which do not encode type information, such as lists of runtime.Objects
func (*Scheme) DataVersionAndKind ¶ added in v0.5.2
DataVersionAndKind will return the APIVersion and Kind of the given wire-format encoding of an API Object, or an error.
func (*Scheme) Decode ¶
Decode converts a YAML or JSON string back into a pointer to an api object. Deduces the type based upon the APIVersion and Kind fields, which are set by Encode. Only versioned objects (APIVersion != "") are accepted. The object will be converted into the in-memory unversioned type before being returned.
func (*Scheme) DecodeInto ¶
DecodeInto parses a YAML or JSON string and stores it in obj. Returns an error if data.Kind is set and doesn't match the type of obj. Obj should be a pointer to an api type. If obj's APIVersion doesn't match that in data, an attempt will be made to convert data into obj's version.
func (*Scheme) EncodeToVersion ¶
EncodeToVersion turns the given api object into an appropriate JSON string. Will return an error if the object doesn't have an embedded TypeMeta. Obj may be a pointer to a struct, or a struct. If a struct, a copy must be made. If a pointer, the object may be modified before encoding, but will be put back into its original state before returning.
Memory/wire format differences:
- Having to keep track of the Kind and APIVersion fields makes tests very annoying, so the rule is that they are set only in wire format (json), not when in native (memory) format. This is possible because both pieces of information are implicit in the go typed object.
- An exception: note that, if there are embedded API objects of known type, for example, PodList{... Items []Pod ...}, these embedded objects must be of the same version of the object they are embedded within, and their APIVersion and Kind must both be empty.
- Note that the exception does not apply to the APIObject type, which recursively does Encode()/Decode(), and is capable of expressing any API object.
- Only versioned objects should be encoded. This means that, if you pass a native object, Encode will convert it to a versioned object. For example, an api.Pod will get converted to a v1beta1.Pod. However, if you pass in an object that's already versioned (v1beta1.Pod), Encode will not modify it.
The purpose of the above complex conversion behavior is to allow us to change the memory format yet not break compatibility with any stored objects, whether they be in our storage layer (e.g., etcd), or in user's config files.
func (*Scheme) KnownTypes ¶
KnownTypes returns the types known for the given version. Return value must be treated as read-only.
func (*Scheme) Log ¶
func (s *Scheme) Log(l conversion.DebugLogger)
Log sets a logger on the scheme. For test purposes only
type SelfLinker ¶
type SelfLinker interface { SetSelfLink(obj Object, selfLink string) error SelfLink(obj Object) (string, error) // Knowing Name is sometimes necessary to use a SelfLinker. Name(obj Object) (string, error) // Knowing Namespace is sometimes necessary to use a SelfLinker Namespace(obj Object) (string, error) }
SelfLinker provides methods for setting and retrieving the SelfLink field of an API object.
type TypeMeta ¶
type TypeMeta struct { APIVersion string `json:"apiVersion,omitempty" yaml:"apiVersion,omitempty"` Kind string `json:"kind,omitempty" yaml:"kind,omitempty"` }
TypeMeta is shared by all top level objects. The proper way to use it is to inline it in your type, like this:
type MyAwesomeAPIObject struct { runtime.TypeMeta `json:",inline"` ... // other fields }
func (*MyAwesomeAPIObject) IsAnAPIObject() {}
TypeMeta is provided here for convenience. You may use it directly from this package or define your own with the same fields.
type Unknown ¶
type Unknown struct { TypeMeta `json:",inline"` // RawJSON will hold the complete JSON of the object which couldn't be matched // with a registered type. Most likely, nothing should be done with this // except for passing it through the system. RawJSON []byte }
Unknown allows api objects with unknown types to be passed-through. This can be used to deal with the API objects from a plug-in. Unknown objects still have functioning TypeMeta features-- kind, version, resourceVersion, etc. TODO: Not implemented yet!
func (*Unknown) IsAnAPIObject ¶
func (*Unknown) IsAnAPIObject()