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 <[email protected]>

Author: wxiaoguang

docs/content/administration/repo-indexer.en-us.md

@@ -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):

docs/content/installation/comparison.en-us.md

@@ -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 |

modules/git/command.go

@@ -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

modules/git/git.go

@@ -341,7 +341,7 @@ func checkGitVersionCompatibility(gitVer *version.Version) 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)
 	}
 
@@ -364,7 +364,7 @@ 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 {
@@ -382,7 +382,7 @@ 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 {
@@ -403,7 +403,7 @@ func configUnsetAll(key, value string) error {
 		}
 		return nil
 	}
-	if err.IsExitCode(1) {
+	if IsErrorExitCode(err, 1) {
 		// not exist
 		return nil
 	}

modules/git/grep.go

@@ -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
+}

modules/git/grep_test.go

@@ -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)
+}

modules/indexer/code/search.go

@@ -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
@@ -91,37 +105,24 @@ 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,
 		CommitID:    result.CommitID,
 		UpdatedUnix: result.UpdatedUnix,
 		Language:    result.Language,
 		Color:       result.Color,
-		Lines:       lines,
+		Lines:       HighlightSearchResultCode(result.Filename, lineNums, formattedLinesBuffer.String()),
 	}, nil
 }
 

options/locale/locale_en-US.ini

@@ -172,6 +172,7 @@ 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...