fork of https://github.com/sourcegraph/zoekt
gitindex: diff config updates for existing clones (#1019)
This supersedes #635 by porting its selective config-sync idea onto
current main with a smaller, easier-to-read shape. Clone orchestration
stays in gitindex/clone.go, while config argument generation and
existing-clone sync logic now live in gitindex/clone_config.go.
For existing clones, we now diff each zoekt.* setting before writing.
Unchanged values are skipped, changed values are updated, and settings
that disappear are removed. CloneRepo returns the repo destination only
when a setting change actually happened so the caller can trigger
reindexing only when needed.
cmd/zoekt-mirror-github was also cleaned up so optional integer metadata
keys are only added to the config map when present, which avoids pushing
empty config values downstream.
Note: On #635 review it mentioned using go-git. This commit initially
explored that but it ended up being a _lot_ more code due to missing
utilities around easily setting values based on a git config string.
author
Keegan Carruthers-Smith
committer
GitHub
date
(Mar 23, 2026, 2:11 PM +0200)
commit
1e121443
1e121443f9f27758c3cb4f887bd4d028d66b0994
parent
ed735fa5
ed735fa5e9b8cc822e026f996551b8dfb20dfbad
+108
-50
+8
-9
cmd/zoekt-mirror-github/main.go
+8
-9
cmd/zoekt-mirror-github/main.go
···
292
292
return allRepos, nil
293
293
}
294
294
295
-
func itoa(p *int) string {
296
-
if p != nil {
297
-
return strconv.Itoa(*p)
295
+
func setOptionalIntConfig(config map[string]string, key string, value *int) {
296
+
if value != nil {
297
+
config[key] = strconv.Itoa(*value)
298
298
}
299
-
return ""
300
299
}
301
300
302
301
func cloneRepos(destDir string, repos []*github.Repository) error {
···
311
310
"zoekt.web-url": *r.HTMLURL,
312
311
"zoekt.name": filepath.Join(host.Hostname(), *r.FullName),
313
312
314
-
"zoekt.github-stars": itoa(r.StargazersCount),
315
-
"zoekt.github-watchers": itoa(r.WatchersCount),
316
-
"zoekt.github-subscribers": itoa(r.SubscribersCount),
317
-
"zoekt.github-forks": itoa(r.ForksCount),
318
-
319
313
"zoekt.archived": marshalBool(r.Archived != nil && *r.Archived),
320
314
"zoekt.fork": marshalBool(r.Fork != nil && *r.Fork),
321
315
"zoekt.public": marshalBool(r.Private == nil || !*r.Private),
322
316
}
317
+
setOptionalIntConfig(config, "zoekt.github-stars", r.StargazersCount)
318
+
setOptionalIntConfig(config, "zoekt.github-watchers", r.WatchersCount)
319
+
setOptionalIntConfig(config, "zoekt.github-subscribers", r.SubscribersCount)
320
+
setOptionalIntConfig(config, "zoekt.github-forks", r.ForksCount)
321
+
323
322
dest, err := gitindex.CloneRepo(destDir, *r.FullName, *r.CloneURL, config)
324
323
if err != nil {
325
324
return err
+7
-41
gitindex/clone.go
+7
-41
gitindex/clone.go
···
18
18
"bytes"
19
19
"fmt"
20
20
"log"
21
-
"maps"
22
21
"os"
23
22
"os/exec"
24
23
"path/filepath"
25
-
"sort"
26
24
27
25
git "github.com/go-git/go-git/v5"
28
26
"github.com/go-git/go-git/v5/config"
29
27
)
30
28
31
-
// Updates the zoekt.* git config options after a repo is cloned.
32
-
// Once a repo is cloned, we can no longer use the --config flag to update all
33
-
// of it's zoekt.* settings at once. `git config` is limited to one option at once.
34
-
func updateZoektGitConfig(repoDest string, settings map[string]string) error {
35
-
var keys []string
36
-
for k := range settings {
37
-
keys = append(keys, k)
38
-
}
39
-
sort.Strings(keys)
40
-
41
-
for _, k := range keys {
42
-
if settings[k] != "" {
43
-
if err := exec.Command("git", "-C", repoDest, "config", k, settings[k]).Run(); err != nil {
44
-
return err
45
-
}
46
-
}
47
-
}
48
-
49
-
return nil
50
-
}
51
-
52
29
// CloneRepo clones one repository, adding the given config
53
30
// settings. It returns the bare repo directory. The `name` argument
54
31
// determines where the repo is stored relative to `destDir`. Returns
···
61
38
62
39
repoDest := filepath.Join(parent, filepath.Base(name)+".git")
63
40
if _, err := os.Lstat(repoDest); err == nil {
64
-
// Repository exists, ensure settings are in sync including the clone URL
65
-
settings := maps.Clone(settings)
66
-
settings["remote.origin.url"] = cloneURL
67
-
if err := updateZoektGitConfig(repoDest, settings); err != nil {
41
+
// Repository exists, ensure zoekt settings are in sync.
42
+
hadUpdate, err := updateZoektGitConfig(repoDest, settings)
43
+
if err != nil {
68
44
return "", fmt.Errorf("failed to update repository settings: %w", err)
69
45
}
70
-
return "", nil
71
-
}
72
-
73
-
var keys []string
74
-
for k := range settings {
75
-
keys = append(keys, k)
76
-
}
77
-
sort.Strings(keys)
78
-
79
-
var config []string
80
-
for _, k := range keys {
81
-
if settings[k] != "" {
82
-
config = append(config, "--config", k+"="+settings[k])
46
+
if hadUpdate {
47
+
return repoDest, nil
83
48
}
49
+
return "", nil
84
50
}
85
51
86
52
cmd := exec.Command(
87
53
"git", "clone", "--bare", "--verbose", "--progress",
88
54
)
89
-
cmd.Args = append(cmd.Args, config...)
55
+
cmd.Args = append(cmd.Args, cloneConfigArgs(settings)...)
90
56
cmd.Args = append(cmd.Args, cloneURL, repoDest)
91
57
92
58
// Prevent prompting
+93
gitindex/clone_config.go
+93
gitindex/clone_config.go
···
1
+
package gitindex
2
+
3
+
import (
4
+
"bytes"
5
+
"errors"
6
+
"fmt"
7
+
"maps"
8
+
"os/exec"
9
+
"slices"
10
+
"strings"
11
+
)
12
+
13
+
func sortedKeys(settings map[string]string) []string {
14
+
return slices.Sorted(maps.Keys(settings))
15
+
}
16
+
17
+
func cloneConfigArgs(settings map[string]string) []string {
18
+
args := make([]string, 0, len(settings)*2)
19
+
for _, key := range sortedKeys(settings) {
20
+
if value := settings[key]; value != "" {
21
+
args = append(args, "--config", key+"="+value)
22
+
}
23
+
}
24
+
return args
25
+
}
26
+
27
+
// updateZoektGitConfig applies zoekt.* settings to an existing clone.
28
+
// It returns whether the repository config changed.
29
+
func updateZoektGitConfig(repoDest string, settings map[string]string) (bool, error) {
30
+
changed := false
31
+
for _, key := range sortedKeys(settings) {
32
+
updated, err := syncGitConfigOption(repoDest, key, settings[key])
33
+
if err != nil {
34
+
return false, err
35
+
}
36
+
changed = changed || updated
37
+
}
38
+
return changed, nil
39
+
}
40
+
41
+
func syncGitConfigOption(repoDest, key, value string) (bool, error) {
42
+
current, ok, err := repoConfigValue(repoDest, key)
43
+
if err != nil {
44
+
return false, err
45
+
}
46
+
47
+
if value == "" {
48
+
if !ok {
49
+
return false, nil
50
+
}
51
+
if err := unsetRepoConfigValue(repoDest, key); err != nil {
52
+
return false, err
53
+
}
54
+
return true, nil
55
+
}
56
+
57
+
if ok && current == value {
58
+
return false, nil
59
+
}
60
+
if err := setRepoConfigValue(repoDest, key, value); err != nil {
61
+
return false, err
62
+
}
63
+
return true, nil
64
+
}
65
+
66
+
func repoConfigValue(repoDest, key string) (string, bool, error) {
67
+
cmd := exec.Command("git", "-C", repoDest, "config", "--get", key)
68
+
var out bytes.Buffer
69
+
cmd.Stdout = &out
70
+
if err := cmd.Run(); err == nil {
71
+
return strings.TrimSuffix(out.String(), "\n"), true, nil
72
+
} else {
73
+
var exitErr *exec.ExitError
74
+
if errors.As(err, &exitErr) && exitErr.ExitCode() == 1 {
75
+
return "", false, nil
76
+
}
77
+
return "", false, fmt.Errorf("git config --get %q: %w", key, err)
78
+
}
79
+
}
80
+
81
+
func setRepoConfigValue(repoDest, key, value string) error {
82
+
if err := exec.Command("git", "-C", repoDest, "config", "--replace-all", key, value).Run(); err != nil {
83
+
return fmt.Errorf("git config --replace-all %q: %w", key, err)
84
+
}
85
+
return nil
86
+
}
87
+
88
+
func unsetRepoConfigValue(repoDest, key string) error {
89
+
if err := exec.Command("git", "-C", repoDest, "config", "--unset-all", key).Run(); err != nil {
90
+
return fmt.Errorf("git config --unset-all %q: %w", key, err)
91
+
}
92
+
return nil
93
+
}