Skip to content
Branch: master
Find file Copy path
Find file Copy path
Fetching contributors…
Cannot retrieve contributors at this time
243 lines (206 sloc) 7.19 KB
Copyright 2019 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
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
See the License for the specific language governing permissions and
limitations under the License.
package crd
import (
apiext ""
// TypeIdent represents some type in a Package.
type TypeIdent struct {
Package *loader.Package
Name string
func (t TypeIdent) String() string {
return fmt.Sprintf("%q.%s", t.Package.ID, t.Name)
// PackageOverride overrides the loading of some package
// (potentially setting custom schemata, etc). It must
// call AddPackage if it wants to continue with the default
// loading behavior.
type PackageOverride func(p *Parser, pkg *loader.Package)
// Parser knows how to parse out CRD information and generate
// OpenAPI schemata from some collection of types and markers.
// Most methods on Parser cache their results automatically,
// and thus may be called any number of times.
type Parser struct {
Collector *markers.Collector
// Types contains the known TypeInfo for this parser.
Types map[TypeIdent]*markers.TypeInfo
// Schemata contains the known OpenAPI JSONSchemata for this parser.
Schemata map[TypeIdent]apiext.JSONSchemaProps
// GroupVersions contains the known group-versions of each package in this parser.
GroupVersions map[*loader.Package]schema.GroupVersion
// CustomResourceDefinitions contains the known CustomResourceDefinitions for types in this parser.
CustomResourceDefinitions map[schema.GroupKind]apiext.CustomResourceDefinition
// FlattenedSchemata contains fully flattened schemata for use in building
// CustomResourceDefinition validation. Each schema has been flattened by the flattener,
// and then embedded fields have been flattened with FlattenEmbedded.
FlattenedSchemata map[TypeIdent]apiext.JSONSchemaProps
// PackageOverrides indicates that the loading of any package with
// the given path should be handled by the given overrider.
PackageOverrides map[string]PackageOverride
// checker stores persistent partial type-checking/reference-traversal information.
Checker *loader.TypeChecker
// packages marks packages as loaded, to avoid re-loading them.
packages map[*loader.Package]struct{}
flattener *Flattener
func (p *Parser) init() {
if p.packages == nil {
p.packages = make(map[*loader.Package]struct{})
if p.flattener == nil {
p.flattener = &Flattener{
Parser: p,
if p.Schemata == nil {
p.Schemata = make(map[TypeIdent]apiext.JSONSchemaProps)
if p.Types == nil {
p.Types = make(map[TypeIdent]*markers.TypeInfo)
if p.PackageOverrides == nil {
p.PackageOverrides = make(map[string]PackageOverride)
if p.GroupVersions == nil {
p.GroupVersions = make(map[*loader.Package]schema.GroupVersion)
if p.CustomResourceDefinitions == nil {
p.CustomResourceDefinitions = make(map[schema.GroupKind]apiext.CustomResourceDefinition)
if p.FlattenedSchemata == nil {
p.FlattenedSchemata = make(map[TypeIdent]apiext.JSONSchemaProps)
// indexTypes loads all types in the package into Types.
func (p *Parser) indexTypes(pkg *loader.Package) {
// autodetect
pkgMarkers, err := markers.PackageMarkers(p.Collector, pkg)
if err != nil {
} else {
if skipPkg := pkgMarkers.Get("kubebuilder:skip"); skipPkg != nil {
if nameVal := pkgMarkers.Get("groupName"); nameVal != nil {
versionVal := pkg.Name // a reasonable guess
if versionMarker := pkgMarkers.Get("versionName"); versionMarker != nil {
versionVal = versionMarker.(string)
p.GroupVersions[pkg] = schema.GroupVersion{
Version: versionVal,
Group: nameVal.(string),
if err := markers.EachType(p.Collector, pkg, func(info *markers.TypeInfo) {
ident := TypeIdent{
Package: pkg,
Name: info.Name,
p.Types[ident] = info
}); err != nil {
// LookupType fetches type info from Types.
func (p *Parser) LookupType(pkg *loader.Package, name string) *markers.TypeInfo {
return p.Types[TypeIdent{Package: pkg, Name: name}]
// NeedSchemaFor indicates that a schema should be generated for the given type.
func (p *Parser) NeedSchemaFor(typ TypeIdent) {
if _, knownSchema := p.Schemata[typ]; knownSchema {
info, knownInfo := p.Types[typ]
if !knownInfo {
typ.Package.AddError(fmt.Errorf("unknown type %s", typ))
// avoid tripping recursive schemata, like ManagedFields, by adding an empty WIP schema
p.Schemata[typ] = apiext.JSONSchemaProps{}
schemaCtx := newSchemaContext(typ.Package, p)
ctxForInfo := schemaCtx.ForInfo(info)
pkgMarkers, err := markers.PackageMarkers(p.Collector, typ.Package)
if err != nil {
ctxForInfo.PackageMarkers = pkgMarkers
schema := infoToSchema(ctxForInfo)
p.Schemata[typ] = *schema
func (p *Parser) NeedFlattenedSchemaFor(typ TypeIdent) {
if _, knownSchema := p.FlattenedSchemata[typ]; knownSchema {
partialFlattened := p.flattener.FlattenType(typ)
fullyFlattened := FlattenEmbedded(partialFlattened, typ.Package)
p.FlattenedSchemata[typ] = *fullyFlattened
// NeedCRDFor lives off in spec.go
// AddPackage indicates that types and type-checking information is needed
// for the the given package, *ignoring* overrides.
// Generally, consumers should call NeedPackage, while PackageOverrides should
// call AddPackage to continue with the normal loading procedure.
func (p *Parser) AddPackage(pkg *loader.Package) {
if _, checked := p.packages[pkg]; checked {
p.Checker.Check(pkg, filterTypesForCRDs)
p.packages[pkg] = struct{}{}
// NeedPackage indicates that types and type-checking information
// is needed for the given package.
func (p *Parser) NeedPackage(pkg *loader.Package) {
if _, checked := p.packages[pkg]; checked {
// overrides are going to be written without vendor. This is why we index by the actual
// object when we can.
if override, overridden := p.PackageOverrides[loader.NonVendorPath(pkg.PkgPath)]; overridden {
override(p, pkg)
p.packages[pkg] = struct{}{}
// filterTypesForCRDs filters out all nodes that aren't used in CRD generation,
// like interfaces and struct fields without JSON tag.
func filterTypesForCRDs(node ast.Node) bool {
switch node := node.(type) {
case *ast.InterfaceType:
// skip interfaces, we never care about references in them
return false
case *ast.StructType:
return true
case *ast.Field:
_, hasTag := loader.ParseAstTag(node.Tag).Lookup("json")
// fields without JSON tags mean we have custom serialization,
// so only visit fields with tags.
return hasTag
return true
You can’t perform that action at this time.