update dependencies (#6267)

Signed-off-by: hongming <coder.scala@gmail.com>

vendor/k8s.io/gengo/v2/parser/parse.go

@@ -0,0 +1,821 @@
+/*
+Copyright 2015 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 parser
+
+import (
+	"errors"
+	"fmt"
+	"go/ast"
+	"go/constant"
+	"go/token"
+	gotypes "go/types"
+	"path/filepath"
+	"sort"
+	"strings"
+	"time"
+
+	"golang.org/x/tools/go/packages"
+	"k8s.io/gengo/v2/types"
+	"k8s.io/klog/v2"
+)
+
+// Parser lets you add all the go files in all the packages that you care
+// about, then constructs the type source data.
+type Parser struct {
+	// Map of package paths to definitions.  These keys should be canonical
+	// Go import paths (example.com/foo/bar) and not local paths (./foo/bar).
+	goPkgs map[string]*packages.Package
+
+	// Keep track of which packages were directly requested (as opposed to
+	// those which are transitively loaded).
+	userRequested map[string]bool
+
+	// Keep track of which packages have already been scanned for types.
+	fullyProcessed map[string]bool
+
+	// Build tags to set when loading packages.
+	buildTags []string
+
+	// Tracks accumulated parsed files, so we can do position lookups later.
+	fset *token.FileSet
+
+	// All comments from everywhere in every parsed file.  This map is keyed by
+	// the file-line on which the comment block ends, which makes it easy to
+	// look up comments which immediately precede a given obect (e.g. a type or
+	// function definition), which is what we almost always want.  We need this
+	// because Go's own ast package does a very poor job of handling comments.
+	endLineToCommentGroup map[fileLine]*ast.CommentGroup
+}
+
+// key type for finding comments.
+type fileLine struct {
+	file string
+	line int
+}
+
+// New constructs a new Parser.
+func New() *Parser {
+	return NewWithOptions(Options{})
+}
+
+func NewWithOptions(opts Options) *Parser {
+	return &Parser{
+		goPkgs:                map[string]*packages.Package{},
+		userRequested:         map[string]bool{},
+		fullyProcessed:        map[string]bool{},
+		fset:                  token.NewFileSet(),
+		endLineToCommentGroup: map[fileLine]*ast.CommentGroup{},
+		buildTags:             opts.BuildTags,
+	}
+}
+
+// Options holds optional settings for the Parser.
+type Options struct {
+	// BuildTags is a list of optional tags to be specified when loading
+	// packages.
+	BuildTags []string
+}
+
+// FindPackages expands the provided patterns into a list of Go import-paths,
+// much like `go list -find`.
+func (p *Parser) FindPackages(patterns ...string) ([]string, error) {
+	return p.findPackages(nil, patterns...)
+}
+
+// baseCfg is an optional (may be nil) config which might be injected by tests.
+func (p *Parser) findPackages(baseCfg *packages.Config, patterns ...string) ([]string, error) {
+	toFind := make([]string, 0, len(patterns))
+	results := make([]string, 0, len(patterns))
+	for _, pat := range patterns {
+		if pkg := p.goPkgs[pat]; pkg != nil {
+			results = append(results, pkg.PkgPath)
+		} else {
+			toFind = append(toFind, pat)
+		}
+	}
+	if len(toFind) == 0 {
+		return results, nil
+	}
+
+	cfg := packages.Config{
+		Mode:       packages.NeedName | packages.NeedFiles,
+		BuildFlags: []string{"-tags", strings.Join(p.buildTags, ",")},
+		Tests:      false,
+	}
+	if baseCfg != nil {
+		// This is to support tests, e.g. to inject a fake GOPATH or CWD.
+		cfg.Dir = baseCfg.Dir
+		cfg.Env = baseCfg.Env
+	}
+
+	pkgs, err := packages.Load(&cfg, toFind...)
+	if err != nil {
+		return nil, fmt.Errorf("error loading packages: %w", err)
+	}
+	var allErrs []error
+	for _, pkg := range pkgs {
+		results = append(results, pkg.PkgPath)
+
+		// pkg.Errors is not a slice of `error`, but concrete types.  We have
+		// to iteratively convert each one into `error`.
+		var errs []error
+		for _, e := range pkg.Errors {
+			errs = append(errs, e)
+		}
+		if len(errs) > 0 {
+			allErrs = append(allErrs, fmt.Errorf("error(s) in %q:\n%w", pkg.PkgPath, errors.Join(errs...)))
+		}
+	}
+	if len(allErrs) != 0 {
+		return nil, errors.Join(allErrs...)
+	}
+	return results, nil
+}
+
+// LoadPackages loads and parses the specified Go packages.  Specifically
+// named packages (without a trailing "/...") which do not exist or have no Go
+// files are an error.
+func (p *Parser) LoadPackages(patterns ...string) error {
+	_, err := p.loadPackages(patterns...)
+	return err
+}
+
+// LoadPackagesWithConfigForTesting loads and parses the specified Go packages with the
+// specified packages.Config as a starting point.  This is for testing, and
+// only the .Dir and .Env fields of the Config will be considered.
+func (p *Parser) LoadPackagesWithConfigForTesting(cfg *packages.Config, patterns ...string) error {
+	_, err := p.loadPackagesWithConfig(cfg, patterns...)
+	return err
+}
+
+// LoadPackagesTo loads and parses the specified Go packages, and inserts them
+// into the specified Universe. It returns the packages which match the
+// patterns, but loads all packages and their imports, recursively, into the
+// universe.  See NewUniverse for more.
+func (p *Parser) LoadPackagesTo(u *types.Universe, patterns ...string) ([]*types.Package, error) {
+	// Load Packages.
+	pkgs, err := p.loadPackages(patterns...)
+	if err != nil {
+		return nil, err
+	}
+
+	// Load types in all packages (it will internally filter).
+	if err := p.addPkgsToUniverse(pkgs, u); err != nil {
+		return nil, err
+	}
+
+	// Return the results as gengo types.Packages.
+	ret := make([]*types.Package, 0, len(pkgs))
+	for _, pkg := range pkgs {
+		ret = append(ret, u.Package(pkg.PkgPath))
+	}
+
+	return ret, nil
+}
+
+func (p *Parser) loadPackages(patterns ...string) ([]*packages.Package, error) {
+	return p.loadPackagesWithConfig(nil, patterns...)
+}
+
+// baseCfg is an optional (may be nil) config which might be injected by tests.
+func (p *Parser) loadPackagesWithConfig(baseCfg *packages.Config, patterns ...string) ([]*packages.Package, error) {
+	klog.V(5).Infof("loadPackages %q", patterns)
+
+	// Loading packages is slow - only do ones we know we have not already done
+	// (e.g. if a tool calls LoadPackages itself).
+	existingPkgs, netNewPkgs, err := p.alreadyLoaded(baseCfg, patterns...)
+	if err != nil {
+		return nil, err
+	}
+	if vlog := klog.V(5); vlog.Enabled() {
+		if len(existingPkgs) > 0 {
+			keys := make([]string, 0, len(existingPkgs))
+			for _, p := range existingPkgs {
+				keys = append(keys, p.PkgPath)
+			}
+			vlog.Infof("  already have: %q", keys)
+		}
+		if len(netNewPkgs) > 0 {
+			vlog.Infof("  to be loaded: %q", netNewPkgs)
+		}
+	}
+
+	// If these were not user-requested before, they are now.
+	for _, pkg := range existingPkgs {
+		if !p.userRequested[pkg.PkgPath] {
+			p.userRequested[pkg.PkgPath] = true
+		}
+	}
+	for _, pkg := range netNewPkgs {
+		if !p.userRequested[pkg] {
+			p.userRequested[pkg] = true
+		}
+	}
+
+	if len(netNewPkgs) == 0 {
+		return existingPkgs, nil
+	}
+
+	cfg := packages.Config{
+		Mode: packages.NeedName |
+			packages.NeedFiles | packages.NeedImports | packages.NeedDeps |
+			packages.NeedModule | packages.NeedTypes | packages.NeedSyntax,
+		BuildFlags: []string{"-tags", strings.Join(p.buildTags, ",")},
+		Fset:       p.fset,
+		Tests:      false,
+	}
+	if baseCfg != nil {
+		// This is to support tests, e.g. to inject a fake GOPATH or CWD.
+		cfg.Dir = baseCfg.Dir
+		cfg.Env = baseCfg.Env
+	}
+
+	tBefore := time.Now()
+	pkgs, err := packages.Load(&cfg, netNewPkgs...)
+	if err != nil {
+		return nil, fmt.Errorf("error loading packages: %w", err)
+	}
+	klog.V(5).Infof("  loaded %d pkg(s) in %v", len(pkgs), time.Since(tBefore))
+
+	// Handle any errors.
+	collectErrors := func(pkg *packages.Package) error {
+		var errs []error
+		for _, e := range pkg.Errors {
+			if e.Kind == packages.ListError || e.Kind == packages.ParseError {
+				errs = append(errs, e)
+			}
+		}
+		if len(errs) > 0 {
+			return fmt.Errorf("error(s) in %q:\n%w", pkg.PkgPath, errors.Join(errs...))
+		}
+		return nil
+	}
+	if err := forEachPackageRecursive(pkgs, collectErrors); err != nil {
+		return nil, err
+	}
+
+	// Finish integrating packages into our state.
+	absorbPkg := func(pkg *packages.Package) error {
+		p.goPkgs[pkg.PkgPath] = pkg
+
+		for _, f := range pkg.Syntax {
+			for _, c := range f.Comments {
+				// We need to do this on _every_ pkg, not just user-requested
+				// ones, because some generators look at tags in other
+				// packages.
+				//
+				// TODO: It would be nice if we only did this on user-requested
+				// packages.  The problem is that we don't always know which
+				// other packages will need this information, and even when we
+				// do we may have already loaded the package (as a transitive
+				// dep) and might have stored pointers into it.  Doing a
+				// thorough "reload" without invalidating all those pointers is
+				// a problem for another day.
+				position := p.fset.Position(c.End()) // Fset is synchronized
+				p.endLineToCommentGroup[fileLine{position.Filename, position.Line}] = c
+			}
+		}
+
+		return nil
+	}
+	if err := forEachPackageRecursive(pkgs, absorbPkg); err != nil {
+		return nil, err
+	}
+
+	return append(existingPkgs, pkgs...), nil
+}
+
+// alreadyLoaded figures out which of the specified patterns have already been loaded
+// and which have not, and returns those respectively.
+// baseCfg is an optional (may be nil) config which might be injected by tests.
+func (p *Parser) alreadyLoaded(baseCfg *packages.Config, patterns ...string) ([]*packages.Package, []string, error) {
+	existingPkgs := make([]*packages.Package, 0, len(patterns))
+	netNewPkgs := make([]string, 0, len(patterns))
+
+	// Expand and canonicalize the requested patterns.  This should be fast.
+	if pkgPaths, err := p.findPackages(baseCfg, patterns...); err != nil {
+		return nil, nil, err
+	} else {
+		for _, pkgPath := range pkgPaths {
+			if pkg := p.goPkgs[pkgPath]; pkg != nil {
+				existingPkgs = append(existingPkgs, pkg)
+			} else {
+				netNewPkgs = append(netNewPkgs, pkgPath)
+			}
+		}
+	}
+	return existingPkgs, netNewPkgs, nil
+}
+
+// forEachPackageRecursive will run the provided function on all of the specified
+// packages, and on their imports recursively.  Errors are accumulated and
+// returned as via errors.Join.
+func forEachPackageRecursive(pkgs []*packages.Package, fn func(pkg *packages.Package) error) error {
+	seen := map[string]bool{} // PkgPaths we have already visited
+	var errs []error
+	for _, pkg := range pkgs {
+		errs = append(errs, recursePackage(pkg, fn, seen)...)
+	}
+	if len(errs) > 0 {
+		return errors.Join(errs...)
+	}
+	return nil
+}
+
+func recursePackage(pkg *packages.Package, fn func(pkg *packages.Package) error, seen map[string]bool) []error {
+	if seen[pkg.PkgPath] {
+		return nil
+	}
+	var errs []error
+	seen[pkg.PkgPath] = true
+	if err := fn(pkg); err != nil {
+		errs = append(errs, err)
+	}
+	for _, imp := range pkg.Imports {
+		errs = append(errs, recursePackage(imp, fn, seen)...)
+	}
+	return errs
+}
+
+// UserRequestedPackages fetches a list of the user-imported packages.
+func (p *Parser) UserRequestedPackages() []string {
+	// Iterate packages in a predictable order.
+	pkgPaths := make([]string, 0, len(p.userRequested))
+	for k := range p.userRequested {
+		pkgPaths = append(pkgPaths, string(k))
+	}
+	sort.Strings(pkgPaths)
+	return pkgPaths
+}
+
+// NewUniverse finalizes the loaded packages, searches through them for types
+// and produces a new Universe. The returned Universe has one types.Package
+// entry for each Go package that has been loaded, including all of their
+// dependencies, recursively.  It also has one entry, whose key is "", which
+// represents "builtin" types.
+func (p *Parser) NewUniverse() (types.Universe, error) {
+	u := types.Universe{}
+
+	pkgs := []*packages.Package{}
+	for _, path := range p.UserRequestedPackages() {
+		pkgs = append(pkgs, p.goPkgs[path])
+	}
+	if err := p.addPkgsToUniverse(pkgs, &u); err != nil {
+		return nil, err
+	}
+
+	return u, nil
+}
+
+// addCommentsToType takes any accumulated comment lines prior to obj and
+// attaches them to the type t.
+func (p *Parser) addCommentsToType(obj gotypes.Object, t *types.Type) {
+	t.CommentLines = p.docComment(obj.Pos())
+	t.SecondClosestCommentLines = p.priorDetachedComment(obj.Pos())
+}
+
+// packageDir tries to figure out the directory of the specified package.
+func packageDir(pkg *packages.Package) (string, error) {
+	// Sometimes Module is present but has no Dir, e.g. when it is vendored.
+	if pkg.Module != nil && pkg.Module.Dir != "" {
+		// NOTE: this will not work if tests are loaded, because Go mutates the
+		// Package.PkgPath.
+		subdir := strings.TrimPrefix(pkg.PkgPath, pkg.Module.Path)
+		return filepath.Join(pkg.Module.Dir, subdir), nil
+	}
+	if len(pkg.GoFiles) > 0 {
+		return filepath.Dir(pkg.GoFiles[0]), nil
+	}
+	if len(pkg.IgnoredFiles) > 0 {
+		return filepath.Dir(pkg.IgnoredFiles[0]), nil
+	}
+	return "", fmt.Errorf("can't find package dir for %q - no module info and no Go files", pkg.PkgPath)
+}
+
+// addPkgsToUniverse adds the packages, and all of their deps, recursively, to
+// the universe and (if needed) searches through them for types.
+func (p *Parser) addPkgsToUniverse(pkgs []*packages.Package, u *types.Universe) error {
+	addOne := func(pkg *packages.Package) error {
+		if err := p.addPkgToUniverse(pkg, u); err != nil {
+			return err
+		}
+		return nil
+	}
+	if err := forEachPackageRecursive(pkgs, addOne); err != nil {
+		return err
+	}
+	return nil
+}
+
+// addPkgToUniverse adds one package to the universe and (if needed) searches
+// through it for types.
+func (p *Parser) addPkgToUniverse(pkg *packages.Package, u *types.Universe) error {
+	pkgPath := pkg.PkgPath
+	if p.fullyProcessed[pkgPath] {
+		return nil
+	}
+
+	// This will get-or-create the Package.
+	gengoPkg := u.Package(pkgPath)
+
+	if gengoPkg.Dir == "" {
+		// We're keeping this package, though we might not fully process it.
+		if vlog := klog.V(5); vlog.Enabled() {
+			why := "user-requested"
+			if !p.userRequested[pkgPath] {
+				why = "dependency"
+			}
+			vlog.Infof("addPkgToUniverse %q (%s)", pkgPath, why)
+		}
+
+		absPath := ""
+		if dir, err := packageDir(pkg); err != nil {
+			return err
+		} else {
+			absPath = dir
+		}
+
+		gengoPkg.Path = pkg.PkgPath
+		gengoPkg.Dir = absPath
+	}
+
+	// If the package was not user-requested, we can stop here.
+	if !p.userRequested[pkgPath] {
+		return nil
+	}
+
+	// Mark it as done, so we don't ever re-process it.
+	p.fullyProcessed[pkgPath] = true
+	gengoPkg.Name = pkg.Name
+
+	// For historical reasons we treat files named "doc.go" specially.
+	// TODO: It would be nice to not do this and instead treat package
+	// doc-comments as the "global" config place.  This would require changing
+	// most generators and input files.
+	for _, f := range pkg.Syntax {
+		// This gets the filename for the ast.File.  Iterating pkg.GoFiles is
+		// documented as unreliable.
+		pos := p.fset.Position(f.FileStart)
+		if filepath.Base(pos.Filename) == "doc.go" {
+			gengoPkg.Comments = []string{}
+			for i := range f.Comments {
+				gengoPkg.Comments = append(gengoPkg.Comments, splitLines(f.Comments[i].Text())...)
+			}
+			if f.Doc != nil {
+				gengoPkg.DocComments = splitLines(f.Doc.Text())
+			}
+		}
+	}
+
+	// Walk all the types, recursively and save them for later access.
+	s := pkg.Types.Scope()
+	for _, n := range s.Names() {
+		switch obj := s.Lookup(n).(type) {
+		case *gotypes.TypeName:
+			t := p.walkType(*u, nil, obj.Type())
+			p.addCommentsToType(obj, t)
+		case *gotypes.Func:
+			// We only care about functions, not concrete/abstract methods.
+			if obj.Type() != nil && obj.Type().(*gotypes.Signature).Recv() == nil {
+				t := p.addFunction(*u, nil, obj)
+				p.addCommentsToType(obj, t)
+			}
+		case *gotypes.Var:
+			if !obj.IsField() {
+				t := p.addVariable(*u, nil, obj)
+				p.addCommentsToType(obj, t)
+			}
+		case *gotypes.Const:
+			t := p.addConstant(*u, nil, obj)
+			p.addCommentsToType(obj, t)
+		default:
+			klog.Infof("addPkgToUniverse %q: unhandled object of type %T: %v", pkgPath, obj, obj)
+		}
+	}
+
+	// Add all of this package's imports.
+	importedPkgs := []string{}
+	for _, imp := range pkg.Imports {
+		if err := p.addPkgToUniverse(imp, u); err != nil {
+			return err
+		}
+		importedPkgs = append(importedPkgs, imp.PkgPath)
+	}
+	sort.Strings(importedPkgs)
+	u.AddImports(pkg.PkgPath, importedPkgs...)
+
+	return nil
+}
+
+// If the specified position has a "doc comment", return that.
+func (p *Parser) docComment(pos token.Pos) []string {
+	// An object's doc comment always ends on the line before the object's own
+	// declaration.
+	c1 := p.priorCommentLines(pos, 1)
+	return splitLines(c1.Text()) // safe even if c1 is nil
+}
+
+// If there is a detached (not immediately before a declaration) comment,
+// return that.
+func (p *Parser) priorDetachedComment(pos token.Pos) []string {
+	// An object's doc comment always ends on the line before the object's own
+	// declaration.
+	c1 := p.priorCommentLines(pos, 1)
+
+	// Using a literal "2" here is brittle in theory (it means literally 2
+	// lines), but in practice Go code is gofmt'ed (which elides repeated blank
+	// lines), so it works.
+	var c2 *ast.CommentGroup
+	if c1 == nil {
+		c2 = p.priorCommentLines(pos, 2)
+	} else {
+		c2 = p.priorCommentLines(c1.List[0].Slash, 2)
+	}
+	return splitLines(c2.Text()) // safe even if c1 is nil
+}
+
+// If there's a comment block which ends nlines before pos, return it.
+func (p *Parser) priorCommentLines(pos token.Pos, lines int) *ast.CommentGroup {
+	position := p.fset.Position(pos)
+	key := fileLine{position.Filename, position.Line - lines}
+	return p.endLineToCommentGroup[key]
+}
+
+func splitLines(str string) []string {
+	return strings.Split(strings.TrimRight(str, "\n"), "\n")
+}
+
+func goFuncNameToName(in string) types.Name {
+	name := strings.TrimPrefix(in, "func ")
+	nameParts := strings.Split(name, "(")
+	return goNameToName(nameParts[0])
+}
+
+func goVarNameToName(in string) types.Name {
+	nameParts := strings.Split(in, " ")
+	// nameParts[0] is "var".
+	// nameParts[2:] is the type of the variable, we ignore it for now.
+	return goNameToName(nameParts[1])
+}
+
+func goNameToName(in string) types.Name {
+	// Detect anonymous type names. (These may have '.' characters because
+	// embedded types may have packages, so we detect them specially.)
+	if strings.HasPrefix(in, "struct{") ||
+		strings.HasPrefix(in, "<-chan") ||
+		strings.HasPrefix(in, "chan<-") ||
+		strings.HasPrefix(in, "chan ") ||
+		strings.HasPrefix(in, "func(") ||
+		strings.HasPrefix(in, "func (") ||
+		strings.HasPrefix(in, "*") ||
+		strings.HasPrefix(in, "map[") ||
+		strings.HasPrefix(in, "[") {
+		return types.Name{Name: in}
+	}
+
+	// Otherwise, if there are '.' characters present, the name has a
+	// package path in front.
+	nameParts := strings.Split(in, ".")
+	name := types.Name{Name: in}
+	if n := len(nameParts); n >= 2 {
+		// The final "." is the name of the type--previous ones must
+		// have been in the package path.
+		name.Package, name.Name = strings.Join(nameParts[:n-1], "."), nameParts[n-1]
+	}
+	return name
+}
+
+func (p *Parser) convertSignature(u types.Universe, t *gotypes.Signature) *types.Signature {
+	signature := &types.Signature{}
+	for i := 0; i < t.Params().Len(); i++ {
+		signature.Parameters = append(signature.Parameters, p.walkType(u, nil, t.Params().At(i).Type()))
+		signature.ParameterNames = append(signature.ParameterNames, t.Params().At(i).Name())
+	}
+	for i := 0; i < t.Results().Len(); i++ {
+		signature.Results = append(signature.Results, p.walkType(u, nil, t.Results().At(i).Type()))
+		signature.ResultNames = append(signature.ResultNames, t.Results().At(i).Name())
+	}
+	if r := t.Recv(); r != nil {
+		signature.Receiver = p.walkType(u, nil, r.Type())
+	}
+	signature.Variadic = t.Variadic()
+	return signature
+}
+
+// walkType adds the type, and any necessary child types.
+func (p *Parser) walkType(u types.Universe, useName *types.Name, in gotypes.Type) *types.Type {
+	// Most of the cases are underlying types of the named type.
+	name := goNameToName(in.String())
+	if useName != nil {
+		name = *useName
+	}
+
+	switch t := in.(type) {
+	case *gotypes.Struct:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Struct
+		for i := 0; i < t.NumFields(); i++ {
+			f := t.Field(i)
+			m := types.Member{
+				Name:         f.Name(),
+				Embedded:     f.Anonymous(),
+				Tags:         t.Tag(i),
+				Type:         p.walkType(u, nil, f.Type()),
+				CommentLines: p.docComment(f.Pos()),
+			}
+			out.Members = append(out.Members, m)
+		}
+		return out
+	case *gotypes.Map:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Map
+		out.Elem = p.walkType(u, nil, t.Elem())
+		out.Key = p.walkType(u, nil, t.Key())
+		return out
+	case *gotypes.Pointer:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Pointer
+		out.Elem = p.walkType(u, nil, t.Elem())
+		return out
+	case *gotypes.Slice:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Slice
+		out.Elem = p.walkType(u, nil, t.Elem())
+		return out
+	case *gotypes.Array:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Array
+		out.Elem = p.walkType(u, nil, t.Elem())
+		out.Len = in.(*gotypes.Array).Len()
+		return out
+	case *gotypes.Chan:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Chan
+		out.Elem = p.walkType(u, nil, t.Elem())
+		// TODO: need to store direction, otherwise raw type name
+		// cannot be properly written.
+		return out
+	case *gotypes.Basic:
+		out := u.Type(types.Name{
+			Package: "", // This is a magic package name in the Universe.
+			Name:    t.Name(),
+		})
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Unsupported
+		return out
+	case *gotypes.Signature:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Func
+		out.Signature = p.convertSignature(u, t)
+		return out
+	case *gotypes.Interface:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Interface
+		t.Complete()
+		for i := 0; i < t.NumMethods(); i++ {
+			if out.Methods == nil {
+				out.Methods = map[string]*types.Type{}
+			}
+			method := t.Method(i)
+			name := goNameToName(method.String())
+			mt := p.walkType(u, &name, method.Type())
+			mt.CommentLines = p.docComment(method.Pos())
+			out.Methods[method.Name()] = mt
+		}
+		return out
+	case *gotypes.Named:
+		var out *types.Type
+		switch t.Underlying().(type) {
+		case *gotypes.Named, *gotypes.Basic, *gotypes.Map, *gotypes.Slice:
+			name := goNameToName(t.String())
+			out = u.Type(name)
+			if out.Kind != types.Unknown {
+				return out
+			}
+			out.Kind = types.Alias
+			out.Underlying = p.walkType(u, nil, t.Underlying())
+		default:
+			// gotypes package makes everything "named" with an
+			// underlying anonymous type--we remove that annoying
+			// "feature" for users. This flattens those types
+			// together.
+			name := goNameToName(t.String())
+			if out := u.Type(name); out.Kind != types.Unknown {
+				return out // short circuit if we've already made this.
+			}
+			out = p.walkType(u, &name, t.Underlying())
+		}
+		// If the underlying type didn't already add methods, add them.
+		// (Interface types will have already added methods.)
+		if len(out.Methods) == 0 {
+			for i := 0; i < t.NumMethods(); i++ {
+				if out.Methods == nil {
+					out.Methods = map[string]*types.Type{}
+				}
+				method := t.Method(i)
+				name := goNameToName(method.String())
+				mt := p.walkType(u, &name, method.Type())
+				mt.CommentLines = p.docComment(method.Pos())
+				out.Methods[method.Name()] = mt
+			}
+		}
+		return out
+	default:
+		out := u.Type(name)
+		if out.Kind != types.Unknown {
+			return out
+		}
+		out.Kind = types.Unsupported
+		klog.Warningf("Making unsupported type entry %q for: %#v\n", out, t)
+		return out
+	}
+}
+
+func (p *Parser) addFunction(u types.Universe, useName *types.Name, in *gotypes.Func) *types.Type {
+	name := goFuncNameToName(in.String())
+	if useName != nil {
+		name = *useName
+	}
+	out := u.Function(name)
+	out.Kind = types.DeclarationOf
+	out.Underlying = p.walkType(u, nil, in.Type())
+	return out
+}
+
+func (p *Parser) addVariable(u types.Universe, useName *types.Name, in *gotypes.Var) *types.Type {
+	name := goVarNameToName(in.String())
+	if useName != nil {
+		name = *useName
+	}
+	out := u.Variable(name)
+	out.Kind = types.DeclarationOf
+	out.Underlying = p.walkType(u, nil, in.Type())
+	return out
+}
+
+func (p *Parser) addConstant(u types.Universe, useName *types.Name, in *gotypes.Const) *types.Type {
+	name := goVarNameToName(in.String())
+	if useName != nil {
+		name = *useName
+	}
+	out := u.Constant(name)
+	out.Kind = types.DeclarationOf
+	out.Underlying = p.walkType(u, nil, in.Type())
+
+	var constval string
+
+	// For strings, we use `StringVal()` to get the un-truncated,
+	// un-quoted string. For other values, `.String()` is preferable to
+	// get something relatively human readable (especially since for
+	// floating point types, `ExactString()` will generate numeric
+	// expressions using `big.(*Float).Text()`.
+	switch in.Val().Kind() {
+	case constant.String:
+		constval = constant.StringVal(in.Val())
+	default:
+		constval = in.Val().String()
+	}
+
+	out.ConstValue = &constval
+	return out
+}