Support repo code search without setting up an indexer (#29998)

By using git's ability, end users (especially small instance users) do
not need to enable the indexer, they could also benefit from the code
searching feature.

Fix #29996

![image](https://github.com/go-gitea/gitea/assets/2114189/11b7e458-88a4-480d-b4d7-72ee59406dd1)

![image](https://github.com/go-gitea/gitea/assets/2114189/0fe777d5-c95c-4288-a818-0427680805b6)

---------

Co-authored-by: silverwind <me@silverwind.io>
This commit is contained in:
wxiaoguang 2024-03-25 00:05:00 +08:00 committed by Shiny Nematoda
parent 488a99fb56
commit 1e7a6483b8
12 changed files with 254 additions and 60 deletions

View file

@ -17,6 +17,12 @@ menu:
# Repository indexer
## Builtin repository code search without indexer
Users could do repository-level code search without setting up a repository indexer.
The builtin code search is based on the `git grep` command, which is fast and efficient for small repositories.
Better code search support could be achieved by setting up the repository indexer.
## Setting up the repository indexer
Gitea can search through the files of the repositories by enabling this function in your [`app.ini`](administration/config-cheat-sheet.md):

View file

@ -87,6 +87,9 @@ _Symbols used in table:_
| Git Blame | ✓ | ✘ | ✓ | ✓ | ✓ | ✓ | ✓ | ✓ |
| Visual comparison of image changes | ✓ | ✘ | ✓ | ? | ? | ? | ✘ | ✘ |
- Gitea has builtin repository-level code search
- Better code search support could be achieved by [using a repository indexer](administration/repo-indexer.md)
## Issue Tracker
| Feature | Gitea | Gogs | GitHub EE | GitLab CE | GitLab EE | BitBucket | RhodeCode CE | RhodeCode EE |

View file

@ -367,7 +367,6 @@ type RunStdError interface {
error
Unwrap() error
Stderr() string
IsExitCode(code int) bool
}
type runStdError struct {
@ -392,9 +391,9 @@ func (r *runStdError) Stderr() string {
return r.stderr
}
func (r *runStdError) IsExitCode(code int) bool {
func IsErrorExitCode(err error, code int) bool {
var exitError *exec.ExitError
if errors.As(r.err, &exitError) {
if errors.As(err, &exitError) {
return exitError.ExitCode() == code
}
return false

View file

@ -340,7 +340,7 @@ func CheckGitVersionEqual(equal string) error {
func configSet(key, value string) error {
stdout, _, err := NewCommand(DefaultContext, "config", "--global", "--get").AddDynamicArguments(key).RunStdString(nil)
if err != nil && !err.IsExitCode(1) {
if err != nil && !IsErrorExitCode(err, 1) {
return fmt.Errorf("failed to get git config %s, err: %w", key, err)
}
@ -363,7 +363,7 @@ func configSetNonExist(key, value string) error {
// already exist
return nil
}
if err.IsExitCode(1) {
if IsErrorExitCode(err, 1) {
// not exist, set new config
_, _, err = NewCommand(DefaultContext, "config", "--global").AddDynamicArguments(key, value).RunStdString(nil)
if err != nil {
@ -381,7 +381,7 @@ func configAddNonExist(key, value string) error {
// already exist
return nil
}
if err.IsExitCode(1) {
if IsErrorExitCode(err, 1) {
// not exist, add new config
_, _, err = NewCommand(DefaultContext, "config", "--global", "--add").AddDynamicArguments(key, value).RunStdString(nil)
if err != nil {
@ -402,7 +402,7 @@ func configUnsetAll(key, value string) error {
}
return nil
}
if err.IsExitCode(1) {
if IsErrorExitCode(err, 1) {
// not exist
return nil
}

112
modules/git/grep.go Normal file
View file

@ -0,0 +1,112 @@
// Copyright 2024 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package git
import (
"bufio"
"bytes"
"context"
"errors"
"fmt"
"os"
"strconv"
"strings"
"code.gitea.io/gitea/modules/util"
)
type GrepResult struct {
Filename string
LineNumbers []int
LineCodes []string
}
type GrepOptions struct {
RefName string
ContextLineNumber int
IsFuzzy bool
}
func GrepSearch(ctx context.Context, repo *Repository, search string, opts GrepOptions) ([]*GrepResult, error) {
stdoutReader, stdoutWriter, err := os.Pipe()
if err != nil {
return nil, fmt.Errorf("unable to create os pipe to grep: %w", err)
}
defer func() {
_ = stdoutReader.Close()
_ = stdoutWriter.Close()
}()
/*
The output is like this ( "^@" means \x00):
HEAD:.air.toml
6^@bin = "gitea"
HEAD:.changelog.yml
2^@repo: go-gitea/gitea
*/
var results []*GrepResult
cmd := NewCommand(ctx, "grep", "--null", "--break", "--heading", "--fixed-strings", "--line-number", "--ignore-case", "--full-name")
cmd.AddOptionValues("--context", fmt.Sprint(opts.ContextLineNumber))
if opts.IsFuzzy {
words := strings.Fields(search)
for _, word := range words {
cmd.AddOptionValues("-e", strings.TrimLeft(word, "-"))
}
} else {
cmd.AddOptionValues("-e", strings.TrimLeft(search, "-"))
}
cmd.AddDynamicArguments(util.IfZero(opts.RefName, "HEAD"))
stderr := bytes.Buffer{}
err = cmd.Run(&RunOpts{
Dir: repo.Path,
Stdout: stdoutWriter,
Stderr: &stderr,
PipelineFunc: func(ctx context.Context, cancel context.CancelFunc) error {
_ = stdoutWriter.Close()
defer stdoutReader.Close()
isInBlock := false
scanner := bufio.NewScanner(stdoutReader)
var res *GrepResult
for scanner.Scan() {
line := scanner.Text()
if !isInBlock {
if _ /* ref */, filename, ok := strings.Cut(line, ":"); ok {
isInBlock = true
res = &GrepResult{Filename: filename}
results = append(results, res)
}
continue
}
if line == "" {
if len(results) >= 50 {
cancel()
break
}
isInBlock = false
continue
}
if line == "--" {
continue
}
if lineNum, lineCode, ok := strings.Cut(line, "\x00"); ok {
lineNumInt, _ := strconv.Atoi(lineNum)
res.LineNumbers = append(res.LineNumbers, lineNumInt)
res.LineCodes = append(res.LineCodes, lineCode)
}
}
return scanner.Err()
},
})
// git grep exits with 1 if no results are found
if IsErrorExitCode(err, 1) && stderr.Len() == 0 {
return nil, nil
}
if err != nil && !errors.Is(err, context.Canceled) {
return nil, fmt.Errorf("unable to run git grep: %w, stderr: %s", err, stderr.String())
}
return results, nil
}

41
modules/git/grep_test.go Normal file
View file

@ -0,0 +1,41 @@
// Copyright 2024 The Gitea Authors. All rights reserved.
// SPDX-License-Identifier: MIT
package git
import (
"context"
"path/filepath"
"testing"
"github.com/stretchr/testify/assert"
)
func TestGrepSearch(t *testing.T) {
repo, err := openRepositoryWithDefaultContext(filepath.Join(testReposDir, "language_stats_repo"))
assert.NoError(t, err)
defer repo.Close()
res, err := GrepSearch(context.Background(), repo, "void", GrepOptions{})
assert.NoError(t, err)
assert.Equal(t, []*GrepResult{
{
Filename: "java-hello/main.java",
LineNumbers: []int{3},
LineCodes: []string{" public static void main(String[] args)"},
},
{
Filename: "main.vendor.java",
LineNumbers: []int{3},
LineCodes: []string{" public static void main(String[] args)"},
},
}, res)
res, err = GrepSearch(context.Background(), repo, "no-such-content", GrepOptions{})
assert.NoError(t, err)
assert.Len(t, res, 0)
res, err = GrepSearch(context.Background(), &Repository{Path: "no-such-git-repo"}, "no-such-content", GrepOptions{})
assert.Error(t, err)
assert.Len(t, res, 0)
}

View file

@ -70,13 +70,27 @@ func writeStrings(buf *bytes.Buffer, strs ...string) error {
return nil
}
func HighlightSearchResultCode(filename string, lineNums []int, code string) []ResultLine {
// we should highlight the whole code block first, otherwise it doesn't work well with multiple line highlighting
hl, _ := highlight.Code(filename, "", code)
highlightedLines := strings.Split(string(hl), "\n")
// The lineNums outputted by highlight.Code might not match the original lineNums, because "highlight" removes the last `\n`
lines := make([]ResultLine, min(len(highlightedLines), len(lineNums)))
for i := 0; i < len(lines); i++ {
lines[i].Num = lineNums[i]
lines[i].FormattedContent = template.HTML(highlightedLines[i])
}
return lines
}
func searchResult(result *internal.SearchResult, startIndex, endIndex int) (*Result, error) {
startLineNum := 1 + strings.Count(result.Content[:startIndex], "\n")
var formattedLinesBuffer bytes.Buffer
contentLines := strings.SplitAfter(result.Content[startIndex:endIndex], "\n")
lines := make([]ResultLine, 0, len(contentLines))
lineNums := make([]int, 0, len(contentLines))
index := startIndex
for i, line := range contentLines {
var err error
@ -91,29 +105,16 @@ func searchResult(result *internal.SearchResult, startIndex, endIndex int) (*Res
line[closeActiveIndex:],
)
} else {
err = writeStrings(&formattedLinesBuffer,
line,
)
err = writeStrings(&formattedLinesBuffer, line)
}
if err != nil {
return nil, err
}
lines = append(lines, ResultLine{Num: startLineNum + i})
lineNums = append(lineNums, startLineNum+i)
index += len(line)
}
// we should highlight the whole code block first, otherwise it doesn't work well with multiple line highlighting
hl, _ := highlight.Code(result.Filename, "", formattedLinesBuffer.String())
highlightedLines := strings.Split(string(hl), "\n")
// The lines outputted by highlight.Code might not match the original lines, because "highlight" removes the last `\n`
lines = lines[:min(len(highlightedLines), len(lines))]
highlightedLines = highlightedLines[:len(lines)]
for i := 0; i < len(lines); i++ {
lines[i].FormattedContent = template.HTML(highlightedLines[i])
}
return &Result{
RepoID: result.RepoID,
Filename: result.Filename,
@ -121,7 +122,7 @@ func searchResult(result *internal.SearchResult, startIndex, endIndex int) (*Res
UpdatedUnix: result.UpdatedUnix,
Language: result.Language,
Color: result.Color,
Lines: lines,
Lines: HighlightSearchResultCode(result.Filename, lineNums, formattedLinesBuffer.String()),
}, nil
}

View file

@ -171,6 +171,7 @@ org_kind = Search orgs...
team_kind = Search teams...
code_kind = Search code...
code_search_unavailable = Code search is currently not available. Please contact the site administrator.
code_search_by_git_grep = Current code search results are provided by "git grep". There might be better results if site administrator enables Repository Indexer.
package_kind = Search packages...
project_kind = Search projects...
branch_kind = Search branches...

View file

@ -5,9 +5,11 @@ package repo
import (
"net/http"
"strings"
"code.gitea.io/gitea/models/db"
"code.gitea.io/gitea/modules/base"
"code.gitea.io/gitea/modules/git"
code_indexer "code.gitea.io/gitea/modules/indexer/code"
"code.gitea.io/gitea/modules/setting"
"code.gitea.io/gitea/services/context"
@ -17,11 +19,6 @@ const tplSearch base.TplName = "repo/search"
// Search render repository search page
func Search(ctx *context.Context) {
if !setting.Indexer.RepoIndexerEnabled {
ctx.Redirect(ctx.Repo.RepoLink)
return
}
language := ctx.FormTrim("l")
keyword := ctx.FormTrim("q")
@ -41,19 +38,55 @@ func Search(ctx *context.Context) {
if page <= 0 {
page = 1
}
total, searchResults, searchResultLanguages, err := code_indexer.PerformSearch(ctx, []int64{ctx.Repo.Repository.ID},
language, keyword, page, setting.UI.RepoSearchPagingNum, isMatch)
if err != nil {
if code_indexer.IsAvailable(ctx) {
ctx.ServerError("SearchResults", err)
var total int
var searchResults []*code_indexer.Result
var searchResultLanguages []*code_indexer.SearchResultLanguages
if setting.Indexer.RepoIndexerEnabled {
var err error
total, searchResults, searchResultLanguages, err = code_indexer.PerformSearch(ctx, &code_indexer.SearchOptions{
RepoIDs: []int64{ctx.Repo.Repository.ID},
Keyword: keyword,
IsKeywordFuzzy: isFuzzy,
Language: language,
Paginator: &db.ListOptions{
Page: page,
PageSize: setting.UI.RepoSearchPagingNum,
},
})
if err != nil {
if code_indexer.IsAvailable(ctx) {
ctx.ServerError("SearchResults", err)
return
}
ctx.Data["CodeIndexerUnavailable"] = true
} else {
ctx.Data["CodeIndexerUnavailable"] = !code_indexer.IsAvailable(ctx)
}
} else {
res, err := git.GrepSearch(ctx, ctx.Repo.GitRepo, keyword, git.GrepOptions{ContextLineNumber: 3, IsFuzzy: isFuzzy})
if err != nil {
ctx.ServerError("GrepSearch", err)
return
}
ctx.Data["CodeIndexerUnavailable"] = true
} else {
ctx.Data["CodeIndexerUnavailable"] = !code_indexer.IsAvailable(ctx)
total = len(res)
pageStart := min((page-1)*setting.UI.RepoSearchPagingNum, len(res))
pageEnd := min(page*setting.UI.RepoSearchPagingNum, len(res))
res = res[pageStart:pageEnd]
for _, r := range res {
searchResults = append(searchResults, &code_indexer.Result{
RepoID: ctx.Repo.Repository.ID,
Filename: r.Filename,
CommitID: ctx.Repo.CommitID,
// UpdatedUnix: not supported yet
// Language: not supported yet
// Color: not supported yet
Lines: code_indexer.HighlightSearchResultCode(r.Filename, r.LineNumbers, strings.Join(r.LineCodes, "\n")),
})
}
}
ctx.Data["CodeIndexerEnabled"] = setting.Indexer.RepoIndexerEnabled
ctx.Data["Repo"] = ctx.Repo.Repository
ctx.Data["SourcePath"] = ctx.Repo.Repository.Link()
ctx.Data["SearchResults"] = searchResults

View file

@ -5,29 +5,18 @@
{{template "base/alert" .}}
{{template "repo/code/recently_pushed_new_branches" .}}
{{if and (not .HideRepoInfo) (not .IsBlame)}}
<div class="ui repo-description gt-word-break">
<div id="repo-desc" class="tw-text-16">
<div class="repo-description">
<div id="repo-desc" class="gt-word-break tw-text-16">
{{$description := .Repository.DescriptionHTML $.Context}}
{{if $description}}<span class="description">{{$description | RenderCodeBlock}}</span>{{else if .IsRepositoryAdmin}}<span class="no-description text-italic">{{ctx.Locale.Tr "repo.no_desc"}}</span>{{end}}
<a class="link" href="{{.Repository.Website}}">{{.Repository.Website}}</a>
</div>
{{if .RepoSearchEnabled}}
<div class="ui repo-search">
<form class="ui form ignore-dirty" action="{{.RepoLink}}/search" method="get">
<div class="field">
<div class="ui small action input{{if .CodeIndexerUnavailable}} disabled left icon{{end}}"{{if .CodeIndexerUnavailable}} data-tooltip-content="{{ctx.Locale.Tr "search.code_search_unavailable"}}"{{end}}>
<input name="q" value="{{.Keyword}}"{{if .CodeIndexerUnavailable}} disabled{{end}} placeholder="{{ctx.Locale.Tr "search.code_kind"}}">
{{if .CodeIndexerUnavailable}}
<i class="icon">{{svg "octicon-alert"}}</i>
{{end}}
<button class="ui small icon button"{{if .CodeIndexerUnavailable}} disabled{{end}} type="submit">
{{svg "octicon-search"}}
</button>
</div>
</div>
</form>
<form class="ignore-dirty" action="{{.RepoLink}}/search" method="get">
<div class="ui small action input">
<input name="q" value="{{.Keyword}}" placeholder="{{ctx.Locale.Tr "search.code_kind"}}">
{{template "shared/search/button"}}
</div>
{{end}}
</form>
</div>
<div class="tw-flex tw-items-center tw-flex-wrap tw-gap-1" id="repo-topics">
{{range .Topics}}<a class="ui repo-topic large label topic gt-m-0" href="{{AppSubUrl}}/explore/repos?q={{.Name}}&topic=1">{{.Name}}</a>{{end}}

View file

@ -11,9 +11,16 @@
<div class="ui error message">
<p>{{ctx.Locale.Tr "search.code_search_unavailable"}}</p>
</div>
{{else if .SearchResults}}
{{template "shared/search/code/results" .}}
{{else if .Keyword}}
<div>{{ctx.Locale.Tr "search.no_results"}}</div>
{{else}}
{{if not .CodeIndexerEnabled}}
<div class="ui message">
<p>{{ctx.Locale.Tr "search.code_search_by_git_grep"}}</p>
</div>
{{end}}
{{if .SearchResults}}
{{template "shared/search/code/results" .}}
{{else if .Keyword}}
<div>{{ctx.Locale.Tr "search.no_results"}}</div>
{{end}}
{{end}}
</div>

View file

@ -1,3 +1,4 @@
{{if or .result.Language (not .result.UpdatedUnix.IsZero)}}
<div class="ui bottom attached table segment tw-flex tw-items-center tw-justify-between">
<div class="tw-flex tw-items-center gt-ml-4">
{{if .result.Language}}
@ -10,3 +11,4 @@
{{end}}
</div>
</div>
{{end}}