mirror of
https://github.com/ceph/ceph-csi.git
synced 2024-11-18 12:20:24 +00:00
e3bf375035
update kubernetes to latest v1.25.0 release. Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
472 lines
12 KiB
Go
472 lines
12 KiB
Go
/*
|
|
Copyright 2017 The Kubernetes Authors.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package schemaconv
|
|
|
|
import (
|
|
"errors"
|
|
"fmt"
|
|
"path"
|
|
"sort"
|
|
"strings"
|
|
|
|
"k8s.io/kube-openapi/pkg/util/proto"
|
|
"sigs.k8s.io/structured-merge-diff/v4/schema"
|
|
)
|
|
|
|
const (
|
|
quantityResource = "io.k8s.apimachinery.pkg.api.resource.Quantity"
|
|
)
|
|
|
|
// ToSchema converts openapi definitions into a schema suitable for structured
|
|
// merge (i.e. kubectl apply v2).
|
|
func ToSchema(models proto.Models) (*schema.Schema, error) {
|
|
return ToSchemaWithPreserveUnknownFields(models, false)
|
|
}
|
|
|
|
// ToSchemaWithPreserveUnknownFields converts openapi definitions into a schema suitable for structured
|
|
// merge (i.e. kubectl apply v2), it will preserve unknown fields if specified.
|
|
func ToSchemaWithPreserveUnknownFields(models proto.Models, preserveUnknownFields bool) (*schema.Schema, error) {
|
|
c := convert{
|
|
input: models,
|
|
preserveUnknownFields: preserveUnknownFields,
|
|
output: &schema.Schema{},
|
|
}
|
|
if err := c.convertAll(); err != nil {
|
|
return nil, err
|
|
}
|
|
c.addCommonTypes()
|
|
return c.output, nil
|
|
}
|
|
|
|
type convert struct {
|
|
input proto.Models
|
|
preserveUnknownFields bool
|
|
output *schema.Schema
|
|
|
|
currentName string
|
|
current *schema.Atom
|
|
errorMessages []string
|
|
}
|
|
|
|
func (c *convert) push(name string, a *schema.Atom) *convert {
|
|
return &convert{
|
|
input: c.input,
|
|
preserveUnknownFields: c.preserveUnknownFields,
|
|
output: c.output,
|
|
currentName: name,
|
|
current: a,
|
|
}
|
|
}
|
|
|
|
func (c *convert) top() *schema.Atom { return c.current }
|
|
|
|
func (c *convert) pop(c2 *convert) {
|
|
c.errorMessages = append(c.errorMessages, c2.errorMessages...)
|
|
}
|
|
|
|
func (c *convert) convertAll() error {
|
|
for _, name := range c.input.ListModels() {
|
|
model := c.input.LookupModel(name)
|
|
c.insertTypeDef(name, model)
|
|
}
|
|
if len(c.errorMessages) > 0 {
|
|
return errors.New(strings.Join(c.errorMessages, "\n"))
|
|
}
|
|
return nil
|
|
}
|
|
|
|
func (c *convert) reportError(format string, args ...interface{}) {
|
|
c.errorMessages = append(c.errorMessages,
|
|
c.currentName+": "+fmt.Sprintf(format, args...),
|
|
)
|
|
}
|
|
|
|
func (c *convert) insertTypeDef(name string, model proto.Schema) {
|
|
def := schema.TypeDef{
|
|
Name: name,
|
|
}
|
|
c2 := c.push(name, &def.Atom)
|
|
model.Accept(c2)
|
|
c.pop(c2)
|
|
if def.Atom == (schema.Atom{}) {
|
|
// This could happen if there were a top-level reference.
|
|
return
|
|
}
|
|
c.output.Types = append(c.output.Types, def)
|
|
}
|
|
|
|
func (c *convert) addCommonTypes() {
|
|
c.output.Types = append(c.output.Types, untypedDef)
|
|
c.output.Types = append(c.output.Types, deducedDef)
|
|
}
|
|
|
|
var untypedName string = "__untyped_atomic_"
|
|
|
|
var untypedDef schema.TypeDef = schema.TypeDef{
|
|
Name: untypedName,
|
|
Atom: schema.Atom{
|
|
Scalar: ptr(schema.Scalar("untyped")),
|
|
List: &schema.List{
|
|
ElementType: schema.TypeRef{
|
|
NamedType: &untypedName,
|
|
},
|
|
ElementRelationship: schema.Atomic,
|
|
},
|
|
Map: &schema.Map{
|
|
ElementType: schema.TypeRef{
|
|
NamedType: &untypedName,
|
|
},
|
|
ElementRelationship: schema.Atomic,
|
|
},
|
|
},
|
|
}
|
|
|
|
var deducedName string = "__untyped_deduced_"
|
|
|
|
var deducedDef schema.TypeDef = schema.TypeDef{
|
|
Name: deducedName,
|
|
Atom: schema.Atom{
|
|
Scalar: ptr(schema.Scalar("untyped")),
|
|
List: &schema.List{
|
|
ElementType: schema.TypeRef{
|
|
NamedType: &untypedName,
|
|
},
|
|
ElementRelationship: schema.Atomic,
|
|
},
|
|
Map: &schema.Map{
|
|
ElementType: schema.TypeRef{
|
|
NamedType: &deducedName,
|
|
},
|
|
ElementRelationship: schema.Separable,
|
|
},
|
|
},
|
|
}
|
|
|
|
func (c *convert) makeRef(model proto.Schema, preserveUnknownFields bool) schema.TypeRef {
|
|
var tr schema.TypeRef
|
|
if r, ok := model.(*proto.Ref); ok {
|
|
if r.Reference() == "io.k8s.apimachinery.pkg.runtime.RawExtension" {
|
|
return schema.TypeRef{
|
|
NamedType: &untypedName,
|
|
}
|
|
}
|
|
// reference a named type
|
|
_, n := path.Split(r.Reference())
|
|
tr.NamedType = &n
|
|
|
|
ext := model.GetExtensions()
|
|
if val, ok := ext["x-kubernetes-map-type"]; ok {
|
|
switch val {
|
|
case "atomic":
|
|
relationship := schema.Atomic
|
|
tr.ElementRelationship = &relationship
|
|
case "granular":
|
|
relationship := schema.Separable
|
|
tr.ElementRelationship = &relationship
|
|
default:
|
|
c.reportError("unknown map type %v", val)
|
|
}
|
|
}
|
|
} else {
|
|
// compute the type inline
|
|
c2 := c.push("inlined in "+c.currentName, &tr.Inlined)
|
|
c2.preserveUnknownFields = preserveUnknownFields
|
|
model.Accept(c2)
|
|
c.pop(c2)
|
|
|
|
if tr == (schema.TypeRef{}) {
|
|
// emit warning?
|
|
tr.NamedType = &untypedName
|
|
}
|
|
}
|
|
return tr
|
|
}
|
|
|
|
func makeUnions(extensions map[string]interface{}) ([]schema.Union, error) {
|
|
schemaUnions := []schema.Union{}
|
|
if iunions, ok := extensions["x-kubernetes-unions"]; ok {
|
|
unions, ok := iunions.([]interface{})
|
|
if !ok {
|
|
return nil, fmt.Errorf(`"x-kubernetes-unions" should be a list, got %#v`, unions)
|
|
}
|
|
for _, iunion := range unions {
|
|
union, ok := iunion.(map[interface{}]interface{})
|
|
if !ok {
|
|
return nil, fmt.Errorf(`"x-kubernetes-unions" items should be a map of string to unions, got %#v`, iunion)
|
|
}
|
|
unionMap := map[string]interface{}{}
|
|
for k, v := range union {
|
|
key, ok := k.(string)
|
|
if !ok {
|
|
return nil, fmt.Errorf(`"x-kubernetes-unions" has non-string key: %#v`, k)
|
|
}
|
|
unionMap[key] = v
|
|
}
|
|
schemaUnion, err := makeUnion(unionMap)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
schemaUnions = append(schemaUnions, schemaUnion)
|
|
}
|
|
}
|
|
|
|
// Make sure we have no overlap between unions
|
|
fs := map[string]struct{}{}
|
|
for _, u := range schemaUnions {
|
|
if u.Discriminator != nil {
|
|
if _, ok := fs[*u.Discriminator]; ok {
|
|
return nil, fmt.Errorf("%v field appears multiple times in unions", *u.Discriminator)
|
|
}
|
|
fs[*u.Discriminator] = struct{}{}
|
|
}
|
|
for _, f := range u.Fields {
|
|
if _, ok := fs[f.FieldName]; ok {
|
|
return nil, fmt.Errorf("%v field appears multiple times in unions", f.FieldName)
|
|
}
|
|
fs[f.FieldName] = struct{}{}
|
|
}
|
|
}
|
|
|
|
return schemaUnions, nil
|
|
}
|
|
|
|
func makeUnion(extensions map[string]interface{}) (schema.Union, error) {
|
|
union := schema.Union{
|
|
Fields: []schema.UnionField{},
|
|
}
|
|
|
|
if idiscriminator, ok := extensions["discriminator"]; ok {
|
|
discriminator, ok := idiscriminator.(string)
|
|
if !ok {
|
|
return schema.Union{}, fmt.Errorf(`"discriminator" must be a string, got: %#v`, idiscriminator)
|
|
}
|
|
union.Discriminator = &discriminator
|
|
}
|
|
|
|
if ifields, ok := extensions["fields-to-discriminateBy"]; ok {
|
|
fields, ok := ifields.(map[interface{}]interface{})
|
|
if !ok {
|
|
return schema.Union{}, fmt.Errorf(`"fields-to-discriminateBy" must be a map[string]string, got: %#v`, ifields)
|
|
}
|
|
// Needs sorted keys by field.
|
|
keys := []string{}
|
|
for ifield := range fields {
|
|
field, ok := ifield.(string)
|
|
if !ok {
|
|
return schema.Union{}, fmt.Errorf(`"fields-to-discriminateBy": field must be a string, got: %#v`, ifield)
|
|
}
|
|
keys = append(keys, field)
|
|
|
|
}
|
|
sort.Strings(keys)
|
|
reverseMap := map[string]struct{}{}
|
|
for _, field := range keys {
|
|
value := fields[field]
|
|
discriminated, ok := value.(string)
|
|
if !ok {
|
|
return schema.Union{}, fmt.Errorf(`"fields-to-discriminateBy"/%v: value must be a string, got: %#v`, field, value)
|
|
}
|
|
union.Fields = append(union.Fields, schema.UnionField{
|
|
FieldName: field,
|
|
DiscriminatorValue: discriminated,
|
|
})
|
|
|
|
// Check that we don't have the same discriminateBy multiple times.
|
|
if _, ok := reverseMap[discriminated]; ok {
|
|
return schema.Union{}, fmt.Errorf("Multiple fields have the same discriminated name: %v", discriminated)
|
|
}
|
|
reverseMap[discriminated] = struct{}{}
|
|
}
|
|
}
|
|
|
|
if union.Discriminator != nil && len(union.Fields) == 0 {
|
|
return schema.Union{}, fmt.Errorf("discriminator set to %v, but no fields in union", *union.Discriminator)
|
|
}
|
|
return union, nil
|
|
}
|
|
|
|
func (c *convert) VisitKind(k *proto.Kind) {
|
|
preserveUnknownFields := c.preserveUnknownFields
|
|
if p, ok := k.GetExtensions()["x-kubernetes-preserve-unknown-fields"]; ok && p == true {
|
|
preserveUnknownFields = true
|
|
}
|
|
|
|
a := c.top()
|
|
a.Map = &schema.Map{}
|
|
for _, name := range k.FieldOrder {
|
|
member := k.Fields[name]
|
|
tr := c.makeRef(member, preserveUnknownFields)
|
|
a.Map.Fields = append(a.Map.Fields, schema.StructField{
|
|
Name: name,
|
|
Type: tr,
|
|
Default: member.GetDefault(),
|
|
})
|
|
}
|
|
|
|
unions, err := makeUnions(k.GetExtensions())
|
|
if err != nil {
|
|
c.reportError(err.Error())
|
|
return
|
|
}
|
|
// TODO: We should check that the fields and discriminator
|
|
// specified in the union are actual fields in the struct.
|
|
a.Map.Unions = unions
|
|
|
|
if preserveUnknownFields {
|
|
a.Map.ElementType = schema.TypeRef{
|
|
NamedType: &deducedName,
|
|
}
|
|
}
|
|
|
|
ext := k.GetExtensions()
|
|
if val, ok := ext["x-kubernetes-map-type"]; ok {
|
|
switch val {
|
|
case "atomic":
|
|
a.Map.ElementRelationship = schema.Atomic
|
|
case "granular":
|
|
a.Map.ElementRelationship = schema.Separable
|
|
default:
|
|
c.reportError("unknown map type %v", val)
|
|
}
|
|
}
|
|
}
|
|
|
|
func toStringSlice(o interface{}) (out []string, ok bool) {
|
|
switch t := o.(type) {
|
|
case []interface{}:
|
|
for _, v := range t {
|
|
switch vt := v.(type) {
|
|
case string:
|
|
out = append(out, vt)
|
|
}
|
|
}
|
|
return out, true
|
|
}
|
|
return nil, false
|
|
}
|
|
|
|
func (c *convert) VisitArray(a *proto.Array) {
|
|
atom := c.top()
|
|
atom.List = &schema.List{
|
|
ElementRelationship: schema.Atomic,
|
|
}
|
|
l := atom.List
|
|
l.ElementType = c.makeRef(a.SubType, c.preserveUnknownFields)
|
|
|
|
ext := a.GetExtensions()
|
|
|
|
if val, ok := ext["x-kubernetes-list-type"]; ok {
|
|
if val == "atomic" {
|
|
l.ElementRelationship = schema.Atomic
|
|
} else if val == "set" {
|
|
l.ElementRelationship = schema.Associative
|
|
} else if val == "map" {
|
|
l.ElementRelationship = schema.Associative
|
|
if keys, ok := ext["x-kubernetes-list-map-keys"]; ok {
|
|
if keyNames, ok := toStringSlice(keys); ok {
|
|
l.Keys = keyNames
|
|
} else {
|
|
c.reportError("uninterpreted map keys: %#v", keys)
|
|
}
|
|
} else {
|
|
c.reportError("missing map keys")
|
|
}
|
|
} else {
|
|
c.reportError("unknown list type %v", val)
|
|
l.ElementRelationship = schema.Atomic
|
|
}
|
|
} else if val, ok := ext["x-kubernetes-patch-strategy"]; ok {
|
|
if val == "merge" || val == "merge,retainKeys" {
|
|
l.ElementRelationship = schema.Associative
|
|
if key, ok := ext["x-kubernetes-patch-merge-key"]; ok {
|
|
if keyName, ok := key.(string); ok {
|
|
l.Keys = []string{keyName}
|
|
} else {
|
|
c.reportError("uninterpreted merge key: %#v", key)
|
|
}
|
|
} else {
|
|
// It's not an error for this to be absent, it
|
|
// means it's a set.
|
|
}
|
|
} else if val == "retainKeys" {
|
|
} else {
|
|
c.reportError("unknown patch strategy %v", val)
|
|
l.ElementRelationship = schema.Atomic
|
|
}
|
|
}
|
|
}
|
|
|
|
func (c *convert) VisitMap(m *proto.Map) {
|
|
a := c.top()
|
|
a.Map = &schema.Map{}
|
|
a.Map.ElementType = c.makeRef(m.SubType, c.preserveUnknownFields)
|
|
|
|
ext := m.GetExtensions()
|
|
if val, ok := ext["x-kubernetes-map-type"]; ok {
|
|
switch val {
|
|
case "atomic":
|
|
a.Map.ElementRelationship = schema.Atomic
|
|
case "granular":
|
|
a.Map.ElementRelationship = schema.Separable
|
|
default:
|
|
c.reportError("unknown map type %v", val)
|
|
}
|
|
}
|
|
}
|
|
|
|
func ptr(s schema.Scalar) *schema.Scalar { return &s }
|
|
|
|
func (c *convert) VisitPrimitive(p *proto.Primitive) {
|
|
a := c.top()
|
|
if c.currentName == quantityResource {
|
|
a.Scalar = ptr(schema.Scalar("untyped"))
|
|
} else {
|
|
switch p.Type {
|
|
case proto.Integer:
|
|
a.Scalar = ptr(schema.Numeric)
|
|
case proto.Number:
|
|
a.Scalar = ptr(schema.Numeric)
|
|
case proto.String:
|
|
switch p.Format {
|
|
case "":
|
|
a.Scalar = ptr(schema.String)
|
|
case "byte":
|
|
// byte really means []byte and is encoded as a string.
|
|
a.Scalar = ptr(schema.String)
|
|
case "int-or-string":
|
|
a.Scalar = ptr(schema.Scalar("untyped"))
|
|
case "date-time":
|
|
a.Scalar = ptr(schema.Scalar("untyped"))
|
|
default:
|
|
a.Scalar = ptr(schema.Scalar("untyped"))
|
|
}
|
|
case proto.Boolean:
|
|
a.Scalar = ptr(schema.Boolean)
|
|
default:
|
|
a.Scalar = ptr(schema.Scalar("untyped"))
|
|
}
|
|
}
|
|
}
|
|
|
|
func (c *convert) VisitArbitrary(a *proto.Arbitrary) {
|
|
*c.top() = deducedDef.Atom
|
|
}
|
|
|
|
func (c *convert) VisitReference(proto.Reference) {
|
|
// Do nothing, we handle references specially
|
|
}
|