distribution/registry/storage/catalog.go
Stephen J Day bba5a0d05c
registry/storage: more efficient path compare in catalog
Previous component-wise path comparison is recursive and generates a
large amount of garbage. This more efficient version simply replaces the
path comparison with the zero-value to sort before everything. We do
this by replacing the byte-wise comparison that swaps a single character
inline for the separator comparison, such that separators sort first.

The resulting implementation provides component-wise path comparison
with no cost incurred for allocation or stack frame.

Direction of the comparison is also reversed to match Go style.

Signed-off-by: Stephen J Day <stephen.day@docker.com>
2016-08-08 15:55:26 -07:00

144 lines
3 KiB
Go

package storage
import (
"errors"
"io"
"path"
"strings"
"github.com/docker/distribution/context"
"github.com/docker/distribution/registry/storage/driver"
)
// ErrFinishedWalk is used when the called walk function no longer wants
// to accept any more values. This is used for pagination when the
// required number of repos have been found.
var ErrFinishedWalk = errors.New("finished walk")
// Returns a list, or partial list, of repositories in the registry.
// Because it's a quite expensive operation, it should only be used when building up
// an initial set of repositories.
func (reg *registry) Repositories(ctx context.Context, repos []string, last string) (n int, err error) {
var foundRepos []string
if len(repos) == 0 {
return 0, errors.New("no space in slice")
}
root, err := pathFor(repositoriesRootPathSpec{})
if err != nil {
return 0, err
}
err = Walk(ctx, reg.blobStore.driver, root, func(fileInfo driver.FileInfo) error {
filePath := fileInfo.Path()
// lop the base path off
repoPath := filePath[len(root)+1:]
_, file := path.Split(repoPath)
if file == "_layers" {
repoPath = strings.TrimSuffix(repoPath, "/_layers")
if lessPath(last, repoPath) {
foundRepos = append(foundRepos, repoPath)
}
return ErrSkipDir
} else if strings.HasPrefix(file, "_") {
return ErrSkipDir
}
// if we've filled our array, no need to walk any further
if len(foundRepos) == len(repos) {
return ErrFinishedWalk
}
return nil
})
n = copy(repos, foundRepos)
// Signal that we have no more entries by setting EOF
if len(foundRepos) <= len(repos) && (err == nil || err == ErrSkipDir) {
err = io.EOF
}
return n, err
}
// Enumerate applies ingester to each repository
func (reg *registry) Enumerate(ctx context.Context, ingester func(string) error) error {
repoNameBuffer := make([]string, 100)
var last string
for {
n, err := reg.Repositories(ctx, repoNameBuffer, last)
if err != nil && err != io.EOF {
return err
}
if n == 0 {
break
}
last = repoNameBuffer[n-1]
for i := 0; i < n; i++ {
repoName := repoNameBuffer[i]
err = ingester(repoName)
if err != nil {
return err
}
}
if err == io.EOF {
break
}
}
return nil
}
// lessPath returns true if one path a is less than path b.
//
// A component-wise comparison is done, rather than the lexical comparison of
// strings.
func lessPath(a, b string) bool {
// we provide this behavior by making separator always sort first.
return compareReplaceInline(a, b, '/', '\x00') < 0
}
// compareReplaceInline modifies runtime.cmpstring to replace old with new
// during a byte-wise comparison.
func compareReplaceInline(s1, s2 string, old, new byte) int {
l := len(s1)
if len(s2) < l {
l = len(s2)
}
for i := 0; i < l; i++ {
c1, c2 := s1[i], s2[i]
if c1 == old {
c1 = new
}
if c2 == old {
c2 = new
}
if c1 < c2 {
return -1
}
if c1 > c2 {
return +1
}
}
if len(s1) < len(s2) {
return -1
}
if len(s1) > len(s2) {
return +1
}
return 0
}