1
0
mirror of https://github.com/golang/go synced 2024-11-18 12:34:42 -07:00
go/internal/imports/mod.go

660 lines
18 KiB
Go
Raw Normal View History

package imports
import (
"bytes"
"context"
"encoding/json"
"fmt"
"io/ioutil"
"os"
"path"
"path/filepath"
"regexp"
"sort"
"strconv"
"strings"
"golang.org/x/mod/module"
"golang.org/x/tools/internal/gocommand"
"golang.org/x/tools/internal/gopathwalk"
)
// ModuleResolver implements resolver for modules using the go command as little
// as feasible.
type ModuleResolver struct {
env *ProcessEnv
moduleCacheDir string
dummyVendorMod *gocommand.ModuleJSON // If vendoring is enabled, the pseudo-module that represents the /vendor directory.
roots []gopathwalk.Root
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
scanSema chan struct{} // scanSema prevents concurrent scans and guards scannedRoots.
scannedRoots map[gopathwalk.Root]bool
initialized bool
main *gocommand.ModuleJSON
modsByModPath []*gocommand.ModuleJSON // All modules, ordered by # of path components in module Path...
modsByDir []*gocommand.ModuleJSON // ...or Dir.
// moduleCacheCache stores information about the module cache.
moduleCacheCache *dirInfoCache
otherCache *dirInfoCache
}
func newModuleResolver(e *ProcessEnv) *ModuleResolver {
r := &ModuleResolver{
env: e,
scanSema: make(chan struct{}, 1),
}
r.scanSema <- struct{}{}
return r
}
func (r *ModuleResolver) init() error {
if r.initialized {
return nil
}
inv := gocommand.Invocation{
BuildFlags: r.env.BuildFlags,
Env: r.env.env(),
Logf: r.env.Logf,
WorkingDir: r.env.WorkingDir,
}
mainMod, vendorEnabled, err := gocommand.VendorEnabled(context.TODO(), inv, r.env.GocmdRunner)
if err != nil {
return err
}
if mainMod != nil && vendorEnabled {
// Vendor mode is on, so all the non-Main modules are irrelevant,
// and we need to search /vendor for everything.
r.main = mainMod
r.dummyVendorMod = &gocommand.ModuleJSON{
Path: "",
Dir: filepath.Join(mainMod.Dir, "vendor"),
}
r.modsByModPath = []*gocommand.ModuleJSON{mainMod, r.dummyVendorMod}
r.modsByDir = []*gocommand.ModuleJSON{mainMod, r.dummyVendorMod}
} else {
// Vendor mode is off, so run go list -m ... to find everything.
r.initAllMods()
}
if gmc := r.env.Env["GOMODCACHE"]; gmc != "" {
r.moduleCacheDir = gmc
} else {
r.moduleCacheDir = filepath.Join(filepath.SplitList(r.env.gopath())[0], "/pkg/mod")
}
sort.Slice(r.modsByModPath, func(i, j int) bool {
count := func(x int) int {
return strings.Count(r.modsByModPath[x].Path, "/")
}
return count(j) < count(i) // descending order
})
sort.Slice(r.modsByDir, func(i, j int) bool {
count := func(x int) int {
return strings.Count(r.modsByDir[x].Dir, "/")
}
return count(j) < count(i) // descending order
})
r.roots = []gopathwalk.Root{
{filepath.Join(r.env.goroot(), "/src"), gopathwalk.RootGOROOT},
}
if r.main != nil {
r.roots = append(r.roots, gopathwalk.Root{r.main.Dir, gopathwalk.RootCurrentModule})
}
if vendorEnabled {
r.roots = append(r.roots, gopathwalk.Root{r.dummyVendorMod.Dir, gopathwalk.RootOther})
} else {
addDep := func(mod *gocommand.ModuleJSON) {
if mod.Replace == nil {
// This is redundant with the cache, but we'll skip it cheaply enough.
r.roots = append(r.roots, gopathwalk.Root{mod.Dir, gopathwalk.RootModuleCache})
} else {
r.roots = append(r.roots, gopathwalk.Root{mod.Dir, gopathwalk.RootOther})
}
}
// Walk dependent modules before scanning the full mod cache, direct deps first.
for _, mod := range r.modsByModPath {
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
if !mod.Indirect && !mod.Main {
addDep(mod)
}
}
for _, mod := range r.modsByModPath {
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
if mod.Indirect && !mod.Main {
addDep(mod)
}
}
r.roots = append(r.roots, gopathwalk.Root{r.moduleCacheDir, gopathwalk.RootModuleCache})
}
r.scannedRoots = map[gopathwalk.Root]bool{}
if r.moduleCacheCache == nil {
r.moduleCacheCache = &dirInfoCache{
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
dirs: map[string]*directoryPackageInfo{},
listeners: map[*int]cacheListener{},
}
}
if r.otherCache == nil {
r.otherCache = &dirInfoCache{
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
dirs: map[string]*directoryPackageInfo{},
listeners: map[*int]cacheListener{},
}
}
r.initialized = true
return nil
}
func (r *ModuleResolver) initAllMods() error {
stdout, err := r.env.invokeGo(context.TODO(), "list", "-m", "-json", "...")
if err != nil {
return err
}
for dec := json.NewDecoder(stdout); dec.More(); {
mod := &gocommand.ModuleJSON{}
if err := dec.Decode(mod); err != nil {
return err
}
if mod.Dir == "" {
if r.env.Logf != nil {
r.env.Logf("module %v has not been downloaded and will be ignored", mod.Path)
}
// Can't do anything with a module that's not downloaded.
continue
}
// golang/go#36193: the go command doesn't always clean paths.
mod.Dir = filepath.Clean(mod.Dir)
r.modsByModPath = append(r.modsByModPath, mod)
r.modsByDir = append(r.modsByDir, mod)
if mod.Main {
r.main = mod
}
}
return nil
}
func (r *ModuleResolver) ClearForNewScan() {
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
<-r.scanSema
r.scannedRoots = map[gopathwalk.Root]bool{}
r.otherCache = &dirInfoCache{
dirs: map[string]*directoryPackageInfo{},
listeners: map[*int]cacheListener{},
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
r.scanSema <- struct{}{}
}
func (r *ModuleResolver) ClearForNewMod() {
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
<-r.scanSema
*r = ModuleResolver{
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
env: r.env,
moduleCacheCache: r.moduleCacheCache,
otherCache: r.otherCache,
scanSema: r.scanSema,
}
r.init()
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
r.scanSema <- struct{}{}
}
// findPackage returns the module and directory that contains the package at
// the given import path, or returns nil, "" if no module is in scope.
func (r *ModuleResolver) findPackage(importPath string) (*gocommand.ModuleJSON, string) {
// This can't find packages in the stdlib, but that's harmless for all
// the existing code paths.
for _, m := range r.modsByModPath {
if !strings.HasPrefix(importPath, m.Path) {
continue
}
pathInModule := importPath[len(m.Path):]
pkgDir := filepath.Join(m.Dir, pathInModule)
if r.dirIsNestedModule(pkgDir, m) {
continue
}
if info, ok := r.cacheLoad(pkgDir); ok {
if loaded, err := info.reachedStatus(nameLoaded); loaded {
if err != nil {
continue // No package in this dir.
}
return m, pkgDir
}
if scanned, err := info.reachedStatus(directoryScanned); scanned && err != nil {
continue // Dir is unreadable, etc.
}
// This is slightly wrong: a directory doesn't have to have an
// importable package to count as a package for package-to-module
// resolution. package main or _test files should count but
// don't.
// TODO(heschi): fix this.
if _, err := r.cachePackageName(info); err == nil {
return m, pkgDir
}
}
// Not cached. Read the filesystem.
pkgFiles, err := ioutil.ReadDir(pkgDir)
if err != nil {
continue
}
// A module only contains a package if it has buildable go
// files in that directory. If not, it could be provided by an
// outer module. See #29736.
for _, fi := range pkgFiles {
if ok, _ := r.env.buildContext().MatchFile(pkgDir, fi.Name()); ok {
return m, pkgDir
}
}
}
return nil, ""
}
func (r *ModuleResolver) cacheLoad(dir string) (directoryPackageInfo, bool) {
if info, ok := r.moduleCacheCache.Load(dir); ok {
return info, ok
}
return r.otherCache.Load(dir)
}
func (r *ModuleResolver) cacheStore(info directoryPackageInfo) {
if info.rootType == gopathwalk.RootModuleCache {
r.moduleCacheCache.Store(info.dir, info)
} else {
r.otherCache.Store(info.dir, info)
}
}
func (r *ModuleResolver) cacheKeys() []string {
return append(r.moduleCacheCache.Keys(), r.otherCache.Keys()...)
}
// cachePackageName caches the package name for a dir already in the cache.
func (r *ModuleResolver) cachePackageName(info directoryPackageInfo) (string, error) {
if info.rootType == gopathwalk.RootModuleCache {
return r.moduleCacheCache.CachePackageName(info)
}
return r.otherCache.CachePackageName(info)
}
func (r *ModuleResolver) cacheExports(ctx context.Context, env *ProcessEnv, info directoryPackageInfo) (string, []string, error) {
if info.rootType == gopathwalk.RootModuleCache {
return r.moduleCacheCache.CacheExports(ctx, env, info)
}
return r.otherCache.CacheExports(ctx, env, info)
}
// findModuleByDir returns the module that contains dir, or nil if no such
// module is in scope.
func (r *ModuleResolver) findModuleByDir(dir string) *gocommand.ModuleJSON {
// This is quite tricky and may not be correct. dir could be:
// - a package in the main module.
// - a replace target underneath the main module's directory.
// - a nested module in the above.
// - a replace target somewhere totally random.
// - a nested module in the above.
// - in the mod cache.
// - in /vendor/ in -mod=vendor mode.
// - nested module? Dunno.
// Rumor has it that replace targets cannot contain other replace targets.
for _, m := range r.modsByDir {
if !strings.HasPrefix(dir, m.Dir) {
continue
}
if r.dirIsNestedModule(dir, m) {
continue
}
return m
}
return nil
}
// dirIsNestedModule reports if dir is contained in a nested module underneath
// mod, not actually in mod.
func (r *ModuleResolver) dirIsNestedModule(dir string, mod *gocommand.ModuleJSON) bool {
if !strings.HasPrefix(dir, mod.Dir) {
return false
}
if r.dirInModuleCache(dir) {
// Nested modules in the module cache are pruned,
// so it cannot be a nested module.
return false
}
if mod != nil && mod == r.dummyVendorMod {
// The /vendor pseudomodule is flattened and doesn't actually count.
return false
}
modDir, _ := r.modInfo(dir)
if modDir == "" {
return false
}
return modDir != mod.Dir
}
func (r *ModuleResolver) modInfo(dir string) (modDir string, modName string) {
readModName := func(modFile string) string {
modBytes, err := ioutil.ReadFile(modFile)
if err != nil {
return ""
}
return modulePath(modBytes)
}
if r.dirInModuleCache(dir) {
matches := modCacheRegexp.FindStringSubmatch(dir)
index := strings.Index(dir, matches[1]+"@"+matches[2])
modDir := filepath.Join(dir[:index], matches[1]+"@"+matches[2])
return modDir, readModName(filepath.Join(modDir, "go.mod"))
}
for {
if info, ok := r.cacheLoad(dir); ok {
return info.moduleDir, info.moduleName
}
f := filepath.Join(dir, "go.mod")
info, err := os.Stat(f)
if err == nil && !info.IsDir() {
return dir, readModName(f)
}
d := filepath.Dir(dir)
if len(d) >= len(dir) {
return "", "" // reached top of file system, no go.mod
}
dir = d
}
}
func (r *ModuleResolver) dirInModuleCache(dir string) bool {
if r.moduleCacheDir == "" {
return false
}
return strings.HasPrefix(dir, r.moduleCacheDir)
}
func (r *ModuleResolver) loadPackageNames(importPaths []string, srcDir string) (map[string]string, error) {
if err := r.init(); err != nil {
return nil, err
}
names := map[string]string{}
for _, path := range importPaths {
_, packageDir := r.findPackage(path)
if packageDir == "" {
continue
}
name, err := packageDirToName(packageDir)
if err != nil {
continue
}
names[path] = name
}
return names, nil
}
func (r *ModuleResolver) scan(ctx context.Context, callback *scanCallback) error {
if err := r.init(); err != nil {
return err
}
processDir := func(info directoryPackageInfo) {
// Skip this directory if we were not able to get the package information successfully.
if scanned, err := info.reachedStatus(directoryScanned); !scanned || err != nil {
return
}
pkg, err := r.canonicalize(info)
if err != nil {
return
}
if !callback.dirFound(pkg) {
return
}
pkg.packageName, err = r.cachePackageName(info)
if err != nil {
return
}
if !callback.packageNameLoaded(pkg) {
return
}
_, exports, err := r.loadExports(ctx, pkg, false)
if err != nil {
return
}
callback.exportsLoaded(pkg, exports)
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
// Start processing everything in the cache, and listen for the new stuff
// we discover in the walk below.
stop1 := r.moduleCacheCache.ScanAndListen(ctx, processDir)
defer stop1()
stop2 := r.otherCache.ScanAndListen(ctx, processDir)
defer stop2()
// We assume cached directories are fully cached, including all their
// children, and have not changed. We can skip them.
skip := func(root gopathwalk.Root, dir string) bool {
info, ok := r.cacheLoad(dir)
if !ok {
return false
}
// This directory can be skipped as long as we have already scanned it.
// Packages with errors will continue to have errors, so there is no need
// to rescan them.
packageScanned, _ := info.reachedStatus(directoryScanned)
return packageScanned
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
// Add anything new to the cache, and process it if we're still listening.
add := func(root gopathwalk.Root, dir string) {
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
r.cacheStore(r.scanDirForPackage(root, dir))
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
// r.roots and the callback are not necessarily safe to use in the
// goroutine below. Process them eagerly.
roots := filterRoots(r.roots, callback.rootFound)
// We can't cancel walks, because we need them to finish to have a usable
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
// cache. Instead, run them in a separate goroutine and detach.
scanDone := make(chan struct{})
go func() {
select {
case <-ctx.Done():
return
case <-r.scanSema:
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
defer func() { r.scanSema <- struct{}{} }()
// We have the lock on r.scannedRoots, and no other scans can run.
for _, root := range roots {
if ctx.Err() != nil {
return
}
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
if r.scannedRoots[root] {
continue
}
gopathwalk.WalkSkip([]gopathwalk.Root{root}, add, skip, gopathwalk.Options{Logf: r.env.Logf, ModulesEnabled: true})
internal/imports: don't block completions on walks Filesystem walks of large GOPATHs/module caches can take seconds, especially on systems with slow filesystems like MacOS and WSL. We don't want to block completion requests on walks finishing. At the same time, cancelling a walk midway through results in an unusable cache, where we don't know which parts have been scanned so far. The best option is to run the walks in a separate goroutine. Then we can detach and let them finish. On the other side, we need to be able to reattach for the next completion request. Introduce a new method on caches, ScanAndListen, which first processes all the items in the cache, then notifies of any new items. This allows us to reattach to an existing scan without missing anything. The background scan introduces concurrency to the resolvers where there wasn't any before. We can't use mutexes, because there's no way to stop Lock() when a context expires. Use a 1-element semaphore channel to accomplish the same effect. Along the way: Only rescan GOPATH if the resolver has been cleared. None of this makes sense for GOPATH without that. Fix a bug where we were scanning the main module twice in module mode. Stop loading exports in module tests, it slows them down a ton. Change-Id: I978efae733ccba0c0cdc8e8fe6892bf5f15feac8 Reviewed-on: https://go-review.googlesource.com/c/tools/+/213217 Run-TryBot: Heschi Kreinick <heschi@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Rebecca Stambler <rstambler@golang.org>
2020-01-02 14:10:45 -07:00
r.scannedRoots[root] = true
}
close(scanDone)
}()
select {
case <-ctx.Done():
case <-scanDone:
}
return nil
}
func (r *ModuleResolver) scoreImportPath(ctx context.Context, path string) int {
if _, ok := stdlib[path]; ok {
return MaxRelevance
}
mod, _ := r.findPackage(path)
return modRelevance(mod)
}
func modRelevance(mod *gocommand.ModuleJSON) int {
switch {
case mod == nil: // out of scope
return MaxRelevance - 4
case mod.Indirect:
return MaxRelevance - 3
case !mod.Main:
return MaxRelevance - 2
default:
return MaxRelevance - 1 // main module ties with stdlib
}
}
// canonicalize gets the result of canonicalizing the packages using the results
// of initializing the resolver from 'go list -m'.
func (r *ModuleResolver) canonicalize(info directoryPackageInfo) (*pkg, error) {
// Packages in GOROOT are already canonical, regardless of the std/cmd modules.
if info.rootType == gopathwalk.RootGOROOT {
return &pkg{
importPathShort: info.nonCanonicalImportPath,
dir: info.dir,
packageName: path.Base(info.nonCanonicalImportPath),
relevance: MaxRelevance,
}, nil
}
importPath := info.nonCanonicalImportPath
mod := r.findModuleByDir(info.dir)
// Check if the directory is underneath a module that's in scope.
if mod != nil {
// It is. If dir is the target of a replace directive,
// our guessed import path is wrong. Use the real one.
if mod.Dir == info.dir {
importPath = mod.Path
} else {
dirInMod := info.dir[len(mod.Dir)+len("/"):]
importPath = path.Join(mod.Path, filepath.ToSlash(dirInMod))
}
} else if !strings.HasPrefix(importPath, info.moduleName) {
// The module's name doesn't match the package's import path. It
// probably needs a replace directive we don't have.
return nil, fmt.Errorf("package in %q is not valid without a replace statement", info.dir)
}
res := &pkg{
importPathShort: importPath,
dir: info.dir,
relevance: modRelevance(mod),
}
// We may have discovered a package that has a different version
// in scope already. Canonicalize to that one if possible.
if _, canonicalDir := r.findPackage(importPath); canonicalDir != "" {
res.dir = canonicalDir
}
return res, nil
}
func (r *ModuleResolver) loadExports(ctx context.Context, pkg *pkg, includeTest bool) (string, []string, error) {
if err := r.init(); err != nil {
return "", nil, err
}
if info, ok := r.cacheLoad(pkg.dir); ok && !includeTest {
return r.cacheExports(ctx, r.env, info)
}
return loadExportsFromFiles(ctx, r.env, pkg.dir, includeTest)
}
func (r *ModuleResolver) scanDirForPackage(root gopathwalk.Root, dir string) directoryPackageInfo {
subdir := ""
if dir != root.Path {
subdir = dir[len(root.Path)+len("/"):]
}
importPath := filepath.ToSlash(subdir)
if strings.HasPrefix(importPath, "vendor/") {
// Only enter vendor directories if they're explicitly requested as a root.
return directoryPackageInfo{
status: directoryScanned,
err: fmt.Errorf("unwanted vendor directory"),
}
}
switch root.Type {
case gopathwalk.RootCurrentModule:
importPath = path.Join(r.main.Path, filepath.ToSlash(subdir))
case gopathwalk.RootModuleCache:
matches := modCacheRegexp.FindStringSubmatch(subdir)
if len(matches) == 0 {
return directoryPackageInfo{
status: directoryScanned,
err: fmt.Errorf("invalid module cache path: %v", subdir),
}
}
modPath, err := module.UnescapePath(filepath.ToSlash(matches[1]))
if err != nil {
if r.env.Logf != nil {
r.env.Logf("decoding module cache path %q: %v", subdir, err)
}
return directoryPackageInfo{
status: directoryScanned,
err: fmt.Errorf("decoding module cache path %q: %v", subdir, err),
}
}
importPath = path.Join(modPath, filepath.ToSlash(matches[3]))
}
modDir, modName := r.modInfo(dir)
result := directoryPackageInfo{
status: directoryScanned,
dir: dir,
rootType: root.Type,
nonCanonicalImportPath: importPath,
moduleDir: modDir,
moduleName: modName,
}
if root.Type == gopathwalk.RootGOROOT {
// stdlib packages are always in scope, despite the confusing go.mod
return result
}
return result
}
// modCacheRegexp splits a path in a module cache into module, module version, and package.
var modCacheRegexp = regexp.MustCompile(`(.*)@([^/\\]*)(.*)`)
var (
slashSlash = []byte("//")
moduleStr = []byte("module")
)
// modulePath returns the module path from the gomod file text.
// If it cannot find a module path, it returns an empty string.
// It is tolerant of unrelated problems in the go.mod file.
//
// Copied from cmd/go/internal/modfile.
func modulePath(mod []byte) string {
for len(mod) > 0 {
line := mod
mod = nil
if i := bytes.IndexByte(line, '\n'); i >= 0 {
line, mod = line[:i], line[i+1:]
}
if i := bytes.Index(line, slashSlash); i >= 0 {
line = line[:i]
}
line = bytes.TrimSpace(line)
if !bytes.HasPrefix(line, moduleStr) {
continue
}
line = line[len(moduleStr):]
n := len(line)
line = bytes.TrimSpace(line)
if len(line) == n || len(line) == 0 {
continue
}
if line[0] == '"' || line[0] == '`' {
p, err := strconv.Unquote(string(line))
if err != nil {
return "" // malformed quoted string or multiline module path
}
return p
}
return string(line)
}
return "" // missing module path
}