feat: filepath filter for code search (#6143)

Added support for searching content in a specific directory or file.

Reviewed-on: https://codeberg.org/forgejo/forgejo/pulls/6143
Reviewed-by: Gusted <gusted@noreply.codeberg.org>
Reviewed-by: 0ko <0ko@noreply.codeberg.org>
Co-authored-by: Shiny Nematoda <snematoda.751k2@aleeas.com>
Co-committed-by: Shiny Nematoda <snematoda.751k2@aleeas.com>

modules/git/grep.go

@@ -36,13 +36,15 @@ const (
 	RegExpGrepMode
 )
 
+var GrepSearchOptions = [3]string{"exact", "union", "regexp"}
+
 type GrepOptions struct {
 	RefName           string
 	MaxResultLimit    int
 	MatchesPerFile    int // >= git 2.38
 	ContextLineNumber int
 	Mode              grepMode
-	PathSpec          []setting.Glob
+	Filename          string
 }
 
 func (opts *GrepOptions) ensureDefaults() {
@@ -112,12 +114,38 @@ func GrepSearch(ctx context.Context, repo *Repository, search string, opts GrepO
 	}
 
 	// pathspec
-	files := make([]string, 0,
+	includeLen := len(setting.Indexer.IncludePatterns)
-		len(setting.Indexer.IncludePatterns)+
+	if len(opts.Filename) > 0 {
-			len(setting.Indexer.ExcludePatterns)+
+		includeLen = 1
-			len(opts.PathSpec))
+	}
-	for _, expr := range append(setting.Indexer.IncludePatterns, opts.PathSpec...) {
+	files := make([]string, 0, len(setting.Indexer.ExcludePatterns)+includeLen)
-		files = append(files, ":"+expr.Pattern())
+	if len(opts.Filename) > 0 && len(setting.Indexer.IncludePatterns) > 0 {
+		// if the both a global include pattern and the per search path is defined
+		// we only include results where the path matches the globally set pattern
+		// (eg, global pattern = "src/**" and path = "node_modules/")
+
+		// FIXME: this is a bit too restrictive, and fails to consider cases where the
+		// gloabally set include pattern refers to a file than a directory
+		// (eg, global pattern = "**.go" and path = "modules/git")
+		exprMatched := false
+		for _, expr := range setting.Indexer.IncludePatterns {
+			if expr.Match(opts.Filename) {
+				files = append(files, ":(literal)"+opts.Filename)
+				exprMatched = true
+				break
+			}
+		}
+		if !exprMatched {
+			log.Warn("git-grep: filepath %s does not match any include pattern", opts.Filename)
+		}
+	} else if len(opts.Filename) > 0 {
+		// if the path is only set we just include results that matches it
+		files = append(files, ":(literal)"+opts.Filename)
+	} else {
+		// otherwise if global include patterns are set include results that strictly match them
+		for _, expr := range setting.Indexer.IncludePatterns {
+			files = append(files, ":"+expr.Pattern())
+		}
 	}
 	for _, expr := range setting.Indexer.ExcludePatterns {
 		files = append(files, ":^"+expr.Pattern())

modules/git/grep_test.go

@@ -89,6 +89,20 @@ func TestGrepSearch(t *testing.T) {
 		},
 	}, res)
 
+	res, err = GrepSearch(context.Background(), repo, "world", GrepOptions{
+		MatchesPerFile: 1,
+		Filename:       "java-hello/",
+	})
+	require.NoError(t, err)
+	assert.Equal(t, []*GrepResult{
+		{
+			Filename:          "java-hello/main.java",
+			LineNumbers:       []int{1},
+			LineCodes:         []string{"public class HelloWorld"},
+			HighlightedRanges: [][3]int{{0, 18, 23}},
+		},
+	}, res)
+
 	res, err = GrepSearch(context.Background(), repo, "no-such-content", GrepOptions{})
 	require.NoError(t, err)
 	assert.Empty(t, res)

modules/indexer/code/bleve/bleve.go

@@ -17,6 +17,7 @@ import (
 	"code.gitea.io/gitea/modules/charset"
 	"code.gitea.io/gitea/modules/git"
 	"code.gitea.io/gitea/modules/gitrepo"
+	tokenizer_hierarchy "code.gitea.io/gitea/modules/indexer/code/bleve/tokenizer/hierarchy"
 	"code.gitea.io/gitea/modules/indexer/code/internal"
 	indexer_internal "code.gitea.io/gitea/modules/indexer/internal"
 	inner_bleve "code.gitea.io/gitea/modules/indexer/internal/bleve"
@@ -56,6 +57,7 @@ func addUnicodeNormalizeTokenFilter(m *mapping.IndexMappingImpl) error {
 type RepoIndexerData struct {
 	RepoID    int64
 	CommitID  string
+	Filename  string
 	Content   string
 	Language  string
 	UpdatedAt time.Time
@@ -69,7 +71,8 @@ func (d *RepoIndexerData) Type() string {
 const (
 	repoIndexerAnalyzer      = "repoIndexerAnalyzer"
 	repoIndexerDocType       = "repoIndexerDocType"
-	repoIndexerLatestVersion = 6
+	pathHierarchyAnalyzer    = "pathHierarchyAnalyzer"
+	repoIndexerLatestVersion = 7
 )
 
 // generateBleveIndexMapping generates a bleve index mapping for the repo indexer
@@ -89,6 +92,11 @@ func generateBleveIndexMapping() (mapping.IndexMapping, error) {
 	docMapping.AddFieldMappingsAt("Language", termFieldMapping)
 	docMapping.AddFieldMappingsAt("CommitID", termFieldMapping)
 
+	pathFieldMapping := bleve.NewTextFieldMapping()
+	pathFieldMapping.IncludeInAll = false
+	pathFieldMapping.Analyzer = pathHierarchyAnalyzer
+	docMapping.AddFieldMappingsAt("Filename", pathFieldMapping)
+
 	timeFieldMapping := bleve.NewDateTimeFieldMapping()
 	timeFieldMapping.IncludeInAll = false
 	docMapping.AddFieldMappingsAt("UpdatedAt", timeFieldMapping)
@@ -103,6 +111,13 @@ func generateBleveIndexMapping() (mapping.IndexMapping, error) {
 		"token_filters": []string{unicodeNormalizeName, camelcase.Name, lowercase.Name},
 	}); err != nil {
 		return nil, err
+	} else if err := mapping.AddCustomAnalyzer(pathHierarchyAnalyzer, map[string]any{
+		"type":          analyzer_custom.Name,
+		"char_filters":  []string{},
+		"tokenizer":     tokenizer_hierarchy.Name,
+		"token_filters": []string{unicodeNormalizeName},
+	}); err != nil {
+		return nil, err
 	}
 	mapping.DefaultAnalyzer = repoIndexerAnalyzer
 	mapping.AddDocumentMapping(repoIndexerDocType, docMapping)
@@ -178,6 +193,7 @@ func (b *Indexer) addUpdate(ctx context.Context, batchWriter git.WriteCloserErro
 	return batch.Index(id, &RepoIndexerData{
 		RepoID:    repo.ID,
 		CommitID:  commitSha,
+		Filename:  update.Filename,
 		Content:   string(charset.ToUTF8DropErrors(fileContents, charset.ConvertOpts{})),
 		Language:  analyze.GetCodeLanguage(update.Filename, fileContents),
 		UpdatedAt: time.Now().UTC(),
@@ -266,22 +282,30 @@ func (b *Indexer) Search(ctx context.Context, opts *internal.SearchOptions) (int
 		indexerQuery = keywordQuery
 	}
 
+	opts.Filename = strings.Trim(opts.Filename, "/")
+	if len(opts.Filename) > 0 {
+		// we use a keyword analyzer for the query than path hierarchy analyzer
+		// to match only the exact path
+		// eg, a query for modules/indexer/code
+		// should not provide results for modules/ nor modules/indexer
+		indexerQuery = bleve.NewConjunctionQuery(
+			indexerQuery,
+			inner_bleve.MatchQuery(opts.Filename, "Filename", analyzer_keyword.Name, 0),
+		)
+	}
+
 	// Save for reuse without language filter
 	facetQuery := indexerQuery
 	if len(opts.Language) > 0 {
-		languageQuery := bleve.NewMatchQuery(opts.Language)
-		languageQuery.FieldVal = "Language"
-		languageQuery.Analyzer = analyzer_keyword.Name
-
 		indexerQuery = bleve.NewConjunctionQuery(
 			indexerQuery,
-			languageQuery,
+			inner_bleve.MatchQuery(opts.Language, "Language", analyzer_keyword.Name, 0),
 		)
 	}
 
 	from, pageSize := opts.GetSkipTake()
 	searchRequest := bleve.NewSearchRequestOptions(indexerQuery, pageSize, from, false)
-	searchRequest.Fields = []string{"Content", "RepoID", "Language", "CommitID", "UpdatedAt"}
+	searchRequest.Fields = []string{"Content", "RepoID", "Filename", "Language", "CommitID", "UpdatedAt"}
 	searchRequest.IncludeLocations = true
 
 	if len(opts.Language) == 0 {
@@ -320,7 +344,7 @@ func (b *Indexer) Search(ctx context.Context, opts *internal.SearchOptions) (int
 			RepoID:      int64(hit.Fields["RepoID"].(float64)),
 			StartIndex:  startIndex,
 			EndIndex:    endIndex,
-			Filename:    internal.FilenameOfIndexerID(hit.ID),
+			Filename:    hit.Fields["Filename"].(string),
 			Content:     hit.Fields["Content"].(string),
 			CommitID:    hit.Fields["CommitID"].(string),
 			UpdatedUnix: updatedUnix,
@@ -333,7 +357,7 @@ func (b *Indexer) Search(ctx context.Context, opts *internal.SearchOptions) (int
 	if len(opts.Language) > 0 {
 		// Use separate query to go get all language counts
 		facetRequest := bleve.NewSearchRequestOptions(facetQuery, 1, 0, false)
-		facetRequest.Fields = []string{"Content", "RepoID", "Language", "CommitID", "UpdatedAt"}
+		facetRequest.Fields = []string{"Content", "RepoID", "Filename", "Language", "CommitID", "UpdatedAt"}
 		facetRequest.IncludeLocations = true
 		facetRequest.AddFacet("languages", bleve.NewFacetRequest("Language", 10))
 

modules/indexer/code/bleve/tokenizer/hierarchy/hierarchy.go

@@ -0,0 +1,69 @@
+// Copyright 2024 The Forgejo Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package hierarchy
+
+import (
+	"bytes"
+
+	"github.com/blevesearch/bleve/v2/analysis"
+	"github.com/blevesearch/bleve/v2/registry"
+)
+
+const Name = "path_hierarchy"
+
+type PathHierarchyTokenizer struct{}
+
+// Similar to elastic's path_hierarchy tokenizer
+// This tokenizes a given path into all the possible hierarchies
+// For example,
+// modules/indexer/code/search.go =>
+//
+//	modules/
+//	modules/indexer
+//	modules/indexer/code
+//	modules/indexer/code/search.go
+func (t *PathHierarchyTokenizer) Tokenize(input []byte) analysis.TokenStream {
+	// trim any extra slashes
+	input = bytes.Trim(input, "/")
+
+	// zero allocations until the nested directories exceed a depth of 8 (which is unlikely)
+	rv := make(analysis.TokenStream, 0, 8)
+	count, off := 1, 0
+
+	// iterate till all directory seperators
+	for i := bytes.IndexRune(input[off:], '/'); i != -1; i = bytes.IndexRune(input[off:], '/') {
+		// the index is relative to input[offest...]
+		// add this index to the accumlated offset to get the index of the current seperator in input[0...]
+		off += i
+		rv = append(rv, &analysis.Token{
+			Term:     input[:off], // take the slice, input[0...index of seperator]
+			Start:    0,
+			End:      off,
+			Position: count,
+			Type:     analysis.AlphaNumeric,
+		})
+		// increment the offset after considering the seperator
+		off++
+		count++
+	}
+
+	// the entire file path should always be the last token
+	rv = append(rv, &analysis.Token{
+		Term:     input,
+		Start:    0,
+		End:      len(input),
+		Position: count,
+		Type:     analysis.AlphaNumeric,
+	})
+
+	return rv
+}
+
+func TokenizerConstructor(config map[string]any, cache *registry.Cache) (analysis.Tokenizer, error) {
+	return &PathHierarchyTokenizer{}, nil
+}
+
+func init() {
+	registry.RegisterTokenizer(Name, TokenizerConstructor)
+}

modules/indexer/code/bleve/tokenizer/hierarchy/hierarchy_test.go

@@ -0,0 +1,59 @@
+// Copyright 2024 The Forgejo Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package hierarchy
+
+import (
+	"testing"
+
+	"github.com/stretchr/testify/assert"
+)
+
+func TestIndexerBleveHierarchyTokenizer(t *testing.T) {
+	tokenizer := &PathHierarchyTokenizer{}
+	keywords := []struct {
+		Term    string
+		Results []string
+	}{
+		{
+			Term: "modules/indexer/code/search.go",
+			Results: []string{
+				"modules",
+				"modules/indexer",
+				"modules/indexer/code",
+				"modules/indexer/code/search.go",
+			},
+		},
+		{
+			Term: "/tmp/forgejo/",
+			Results: []string{
+				"tmp",
+				"tmp/forgejo",
+			},
+		},
+		{
+			Term: "a/b/c/d/e/f/g/h/i/j",
+			Results: []string{
+				"a",
+				"a/b",
+				"a/b/c",
+				"a/b/c/d",
+				"a/b/c/d/e",
+				"a/b/c/d/e/f",
+				"a/b/c/d/e/f/g",
+				"a/b/c/d/e/f/g/h",
+				"a/b/c/d/e/f/g/h/i",
+				"a/b/c/d/e/f/g/h/i/j",
+			},
+		},
+	}
+
+	for _, kw := range keywords {
+		tokens := tokenizer.Tokenize([]byte(kw.Term))
+		assert.Len(t, tokens, len(kw.Results))
+		for i, token := range tokens {
+			assert.Equal(t, i+1, token.Position)
+			assert.Equal(t, kw.Results[i], string(token.Term))
+		}
+	}
+}

modules/indexer/code/elasticsearch/elasticsearch.go

@@ -30,7 +30,7 @@ import (
 )
 
 const (
-	esRepoIndexerLatestVersion = 1
+	esRepoIndexerLatestVersion = 2
 	// multi-match-types, currently only 2 types are used
 	// Reference: https://www.elastic.co/guide/en/elasticsearch/reference/7.0/query-dsl-multi-match-query.html#multi-match-types
 	esMultiMatchTypeBestFields   = "best_fields"
@@ -57,6 +57,21 @@ func NewIndexer(url, indexerName string) *Indexer {
 
 const (
 	defaultMapping = `{
+		"settings": {
+			"analysis": {
+				"analyzer": {
+					"custom_path_tree": {
+						"tokenizer": "custom_hierarchy"
+					}
+				},
+				"tokenizer": {
+					"custom_hierarchy": {
+						"type": "path_hierarchy",
+						"delimiter": "/"
+					}
+				}
+			}
+		},
 		"mappings": {
 			"properties": {
 				"repo_id": {
@@ -72,6 +87,15 @@ const (
 					"type": "keyword",
 					"index": true
 				},
+				"filename": {
+					"type": "text",
+					"fields": {
+						"tree": {
+							"type": "text",
+							"analyzer": "custom_path_tree"
+						}
+					}
+				},
 				"language": {
 					"type": "keyword",
 					"index": true
@@ -138,6 +162,7 @@ func (b *Indexer) addUpdate(ctx context.Context, batchWriter git.WriteCloserErro
 				"repo_id":    repo.ID,
 				"content":    string(charset.ToUTF8DropErrors(fileContents, charset.ConvertOpts{})),
 				"commit_id":  sha,
+				"filename":   update.Filename,
 				"language":   analyze.GetCodeLanguage(update.Filename, fileContents),
 				"updated_at": timeutil.TimeStampNow(),
 			}),
@@ -267,7 +292,6 @@ func convertResult(searchResult *elastic.SearchResult, kw string, pageSize int)
 			panic(fmt.Sprintf("2===%#v", hit.Highlight))
 		}
 
-		repoID, fileName := internal.ParseIndexerID(hit.Id)
 		res := make(map[string]any)
 		if err := json.Unmarshal(hit.Source, &res); err != nil {
 			return 0, nil, nil, err
@@ -276,8 +300,8 @@ func convertResult(searchResult *elastic.SearchResult, kw string, pageSize int)
 		language := res["language"].(string)
 
 		hits = append(hits, &internal.SearchResult{
-			RepoID:      repoID,
+			RepoID:      int64(res["repo_id"].(float64)),
-			Filename:    fileName,
+			Filename:    res["filename"].(string),
 			CommitID:    res["commit_id"].(string),
 			Content:     res["content"].(string),
 			UpdatedUnix: timeutil.TimeStamp(res["updated_at"].(float64)),
@@ -326,6 +350,9 @@ func (b *Indexer) Search(ctx context.Context, opts *internal.SearchOptions) (int
 		repoQuery := elastic.NewTermsQuery("repo_id", repoStrs...)
 		query = query.Must(repoQuery)
 	}
+	if len(opts.Filename) > 0 {
+		query = query.Filter(elastic.NewTermsQuery("filename.tree", opts.Filename))
+	}
 
 	var (
 		start, pageSize = opts.GetSkipTake()

modules/indexer/code/indexer_test.go

@@ -34,10 +34,11 @@ func testIndexer(name string, t *testing.T, indexer internal.Indexer) {
 		err := index(git.DefaultContext, indexer, repoID)
 		require.NoError(t, err)
 		keywords := []struct {
-			RepoIDs []int64
+			RepoIDs  []int64
-			Keyword string
+			Keyword  string
-			IDs     []int64
+			IDs      []int64
-			Langs   int
+			Langs    int
+			Filename string
 		}{
 			{
 				RepoIDs: nil,
@@ -51,6 +52,20 @@ func testIndexer(name string, t *testing.T, indexer internal.Indexer) {
 				IDs:     []int64{},
 				Langs:   0,
 			},
+			{
+				RepoIDs:  nil,
+				Keyword:  "Description",
+				IDs:      []int64{},
+				Langs:    0,
+				Filename: "NOT-README.md",
+			},
+			{
+				RepoIDs:  nil,
+				Keyword:  "Description",
+				IDs:      []int64{repoID},
+				Langs:    1,
+				Filename: "README.md",
+			},
 			{
 				RepoIDs: nil,
 				Keyword: "Description for",
@@ -86,6 +101,7 @@ func testIndexer(name string, t *testing.T, indexer internal.Indexer) {
 						Page:     1,
 						PageSize: 10,
 					},
+					Filename:       kw.Filename,
 					IsKeywordFuzzy: true,
 				})
 				require.NoError(t, err)

modules/indexer/code/internal/indexer.go

@@ -24,6 +24,7 @@ type SearchOptions struct {
 	RepoIDs  []int64
 	Keyword  string
 	Language string
+	Filename string
 
 	IsKeywordFuzzy bool
 

modules/indexer/code/internal/util.go

@@ -3,30 +3,8 @@
 
 package internal
 
-import (
+import "code.gitea.io/gitea/modules/indexer/internal"
-	"strings"
-
-	"code.gitea.io/gitea/modules/indexer/internal"
-	"code.gitea.io/gitea/modules/log"
-)
 
 func FilenameIndexerID(repoID int64, filename string) string {
 	return internal.Base36(repoID) + "_" + filename
 }
-
-func ParseIndexerID(indexerID string) (int64, string) {
-	index := strings.IndexByte(indexerID, '_')
-	if index == -1 {
-		log.Error("Unexpected ID in repo indexer: %s", indexerID)
-	}
-	repoID, _ := internal.ParseBase36(indexerID[:index])
-	return repoID, indexerID[index+1:]
-}
-
-func FilenameOfIndexerID(indexerID string) string {
-	index := strings.IndexByte(indexerID, '_')
-	if index == -1 {
-		log.Error("Unexpected ID in repo indexer: %s", indexerID)
-	}
-	return indexerID[index+1:]
-}

modules/indexer/code/search.go

@@ -35,6 +35,8 @@ type SearchResultLanguages = internal.SearchResultLanguages
 
 type SearchOptions = internal.SearchOptions
 
+var CodeSearchOptions = [2]string{"exact", "fuzzy"}
+
 func indices(content string, selectionStartIndex, selectionEndIndex int) (int, int) {
 	startIndex := selectionStartIndex
 	numLinesBefore := 0

routers/web/explore/code.go

@@ -35,6 +35,7 @@ func Code(ctx *context.Context) {
 
 	language := ctx.FormTrim("l")
 	keyword := ctx.FormTrim("q")
+	path := ctx.FormTrim("path")
 
 	isFuzzy := ctx.FormOptionalBool("fuzzy").ValueOrDefault(true)
 	if mode := ctx.FormTrim("mode"); len(mode) > 0 {
@@ -91,6 +92,7 @@ func Code(ctx *context.Context) {
 			Keyword:        keyword,
 			IsKeywordFuzzy: isFuzzy,
 			Language:       language,
+			Filename:       path,
 			Paginator: &db.ListOptions{
 				Page:     page,
 				PageSize: setting.UI.RepoSearchPagingNum,

routers/web/repo/search.go

@@ -54,6 +54,7 @@ func Search(ctx *context.Context) {
 	language := ctx.FormTrim("l")
 	keyword := ctx.FormTrim("q")
 
+	path := ctx.FormTrim("path")
 	mode := ExactSearchMode
 	if modeStr := ctx.FormString("mode"); len(modeStr) > 0 {
 		mode = searchModeFromString(modeStr)
@@ -63,6 +64,7 @@ func Search(ctx *context.Context) {
 
 	ctx.Data["Keyword"] = keyword
 	ctx.Data["Language"] = language
+	ctx.Data["CodeSearchPath"] = path
 	ctx.Data["CodeSearchMode"] = mode.String()
 	ctx.Data["PageIsViewCode"] = true
 
@@ -86,6 +88,7 @@ func Search(ctx *context.Context) {
 			Keyword:        keyword,
 			IsKeywordFuzzy: mode == FuzzySearchMode,
 			Language:       language,
+			Filename:       path,
 			Paginator: &db.ListOptions{
 				Page:     page,
 				PageSize: setting.UI.RepoSearchPagingNum,
@@ -100,11 +103,12 @@ func Search(ctx *context.Context) {
 		} else {
 			ctx.Data["CodeIndexerUnavailable"] = !code_indexer.IsAvailable(ctx)
 		}
-		ctx.Data["CodeSearchOptions"] = []string{"exact", "fuzzy"}
+		ctx.Data["CodeSearchOptions"] = code_indexer.CodeSearchOptions
 	} else {
 		grepOpt := git.GrepOptions{
 			ContextLineNumber: 1,
 			RefName:           ctx.Repo.RefName,
+			Filename:          path,
 		}
 		switch mode {
 		case FuzzySearchMode:
@@ -130,10 +134,12 @@ func Search(ctx *context.Context) {
 				// UpdatedUnix: not supported yet
 				// Language:    not supported yet
 				// Color:       not supported yet
-				Lines: code_indexer.HighlightSearchResultCode(r.Filename, r.LineNumbers, r.HighlightedRanges, strings.Join(r.LineCodes, "\n")),
+				Lines: code_indexer.HighlightSearchResultCode(
+					r.Filename, r.LineNumbers, r.HighlightedRanges,
+					strings.Join(r.LineCodes, "\n")),
 			})
 		}
-		ctx.Data["CodeSearchOptions"] = []string{"exact", "union", "regexp"}
+		ctx.Data["CodeSearchOptions"] = git.GrepSearchOptions
 	}
 
 	ctx.Data["CodeIndexerDisabled"] = !setting.Indexer.RepoIndexerEnabled

routers/web/repo/view.go

@@ -39,6 +39,7 @@ import (
 	"code.gitea.io/gitea/modules/git"
 	"code.gitea.io/gitea/modules/gitrepo"
 	"code.gitea.io/gitea/modules/highlight"
+	code_indexer "code.gitea.io/gitea/modules/indexer/code"
 	"code.gitea.io/gitea/modules/lfs"
 	"code.gitea.io/gitea/modules/log"
 	"code.gitea.io/gitea/modules/markup"
@@ -1152,6 +1153,12 @@ PostRecentBranchCheck:
 	ctx.Data["TreeNames"] = treeNames
 	ctx.Data["BranchLink"] = branchLink
 	ctx.Data["CodeIndexerDisabled"] = !setting.Indexer.RepoIndexerEnabled
+	if setting.Indexer.RepoIndexerEnabled {
+		ctx.Data["CodeIndexerUnavailable"] = !code_indexer.IsAvailable(ctx)
+		ctx.Data["CodeSearchOptions"] = code_indexer.CodeSearchOptions
+	} else {
+		ctx.Data["CodeSearchOptions"] = git.GrepSearchOptions
+	}
 	ctx.HTML(http.StatusOK, tplRepoHome)
 }
 

routers/web/user/code.go

@@ -39,6 +39,7 @@ func CodeSearch(ctx *context.Context) {
 
 	language := ctx.FormTrim("l")
 	keyword := ctx.FormTrim("q")
+	path := ctx.FormTrim("path")
 
 	isFuzzy := ctx.FormOptionalBool("fuzzy").ValueOrDefault(true)
 	if mode := ctx.FormTrim("mode"); len(mode) > 0 {
@@ -88,6 +89,7 @@ func CodeSearch(ctx *context.Context) {
 			Keyword:        keyword,
 			IsKeywordFuzzy: isFuzzy,
 			Language:       language,
+			Filename:       path,
 			Paginator: &db.ListOptions{
 				Page:     page,
 				PageSize: setting.UI.RepoSearchPagingNum,

templates/repo/home.tmpl

@@ -11,12 +11,6 @@
 				{{if $description}}<span class="description">{{$description | RenderCodeBlock}}</span>{{else}}<span class="no-description text-italic">{{ctx.Locale.Tr "repo.no_desc"}}</span>{{end}}
 				{{if .Repository.Website}}<a class="link" href="{{.Repository.Website}}">{{.Repository.Website}}</a>{{end}}
 			</div>
-			<form class="ignore-dirty" action="{{.RepoLink}}/search/{{if .CodeIndexerDisabled}}{{.BranchNameSubURL}}{{end}}" method="get" data-test-tag="codesearch">
-				<div class="ui small action input">
-					<input name="q" value="{{.Keyword}}" placeholder="{{ctx.Locale.Tr "search.code_kind"}}">
-					{{template "shared/search/button"}}
-				</div>
-			</form>
 		</div>
 		<div class="tw-flex tw-items-center tw-flex-wrap tw-gap-2 tw-my-2" id="repo-topics">
 			{{/* it should match the code in issue-home.js */}}
@@ -158,6 +152,22 @@
 		{{else if .IsBlame}}
 			{{template "repo/blame" .}}
 		{{else}}{{/* IsViewDirectory */}}
+			{{/* display the search bar only if */}}
+			{{$isCommit := StringUtils.HasPrefix .BranchNameSubURL "commit"}}
+			{{if and (not $isCommit) (or .CodeIndexerDisabled (and (not .TagName) (eq .Repository.DefaultBranch .BranchName)))}}
+				<div class="code-search tw-w-full tw-py-2 tw-px-2 tw-bg-box-header tw-rounded-t tw-border tw-border-secondary tw-border-b-0">
+					<form class="ui form ignore-dirty" action="{{.RepoLink}}/search/{{if .CodeIndexerDisabled}}{{.BranchNameSubURL}}{{end}}" method="get" data-test-tag="codesearch">
+						<input type="hidden" name="path" value="{{.TreePath | PathEscapeSegments}}">
+						{{template "shared/search/combo_multi"
+								dict
+									"Value"       .Keyword
+									"Disabled"    .CodeIndexerUnavailable
+									"Placeholder" (ctx.Locale.Tr "search.code_kind")
+									"Selected"    (index .CodeSearchOptions 0)
+									"Options"     .CodeSearchOptions}}
+					</form>
+				</div>
+			{{end}}
 			{{template "repo/view_list" .}}
 		{{end}}
 	</div>

templates/shared/search/code/results.tmpl

@@ -1,7 +1,7 @@
 <div class="flex-text-block tw-flex-wrap">
 	{{range $term := .SearchResultLanguages}}
 	<a class="ui {{if eq $.Language $term.Language}}primary{{end}} basic label tw-m-0"
-		href="?q={{$.Keyword}}{{if ne $.Language $term.Language}}&l={{$term.Language}}{{end}}&mode={{$.CodeSearchMode}}">
+		href="?q={{$.Keyword}}{{if ne $.Language $term.Language}}&l={{$term.Language}}{{end}}&mode={{$.CodeSearchMode}}&path={{$.CodeSearchPath}}">
 		<i class="color-icon tw-mr-2" style="background-color: {{$term.Color}}"></i>
 		{{$term.Language}}
 		<div class="detail">{{$term.Count}}</div>
@@ -13,7 +13,7 @@
 		{{$repo := or $.Repo (index $.RepoMaps .RepoID)}}
 		<details class="tw-group diff-file-box diff-box file-content non-diff-file-content repo-search-result" open>
 			<summary class="tw-list-none">
-				<h4 class="ui top attached header tw-font-normal tw-flex tw-flex-wrap tw-transform-reset">
+				<h4 class="ui top attached header tw-font-normal tw-flex tw-items-center tw-flex-wrap tw-transform-reset">
 					<span class="tw-h-4 tw-transition -tw-rotate-90 group-open:tw-rotate-0">
 						{{svg "octicon-chevron-down"}}
 					</span>

templates/shared/search/code/search.tmpl

@@ -1,4 +1,5 @@
 <form class="ui form ignore-dirty">
+	<input type="hidden" name="path" value="{{.CodeSearchPath}}">
 	{{template "shared/search/combo_multi"
 		dict
 			"Value" .Keyword
@@ -14,6 +15,23 @@
 			<p>{{ctx.Locale.Tr "search.code_search_unavailable"}}</p>
 		</div>
 	{{else}}
+		{{if .CodeSearchPath}}
+			<div class="tw-mb-4">
+				<span class="breadcrumb">
+					<a class="section" href="?q={{.Keyword}}&mode={{.CodeSearchMode}}">@</a>
+					{{$href := ""}}
+					{{- range $i, $path := StringUtils.Split .CodeSearchPath "/" -}}
+						{{if eq $i 0}}
+							{{$href = $path}}
+						{{else}}
+							{{$href = StringUtils.Join (StringUtils.Make $href $path) "/"}}
+						{{end}}
+						<span class="breadcrumb-divider">/</span>
+						<span class="section"><a href="?q={{$.Keyword}}&mode={{$.CodeSearchMode}}&path={{$href}}">{{$path}}</a></span>
+					{{- end -}}
+				</span>
+			</div>
+		{{end}}
 		{{if .CodeIndexerDisabled}}
 			<div class="ui message" data-test-tag="grep">
 				<p>{{ctx.Locale.Tr "search.code_search_by_git_grep"}}</p>

tests/integration/repo_test.go

@@ -1009,16 +1009,29 @@ func TestRepoCodeSearchForm(t *testing.T) {
 		resp := MakeRequest(t, req, http.StatusOK)
 
 		htmlDoc := NewHTMLParser(t, resp.Body)
-		action, exists := htmlDoc.doc.Find("form[data-test-tag=codesearch]").Attr("action")
+		formEl := htmlDoc.doc.Find("form[data-test-tag=codesearch]")
+
+		action, exists := formEl.Attr("action")
 		assert.True(t, exists)
-
 		branchSubURL := "/branch/master"
-
 		if indexer {
 			assert.NotContains(t, action, branchSubURL)
 		} else {
 			assert.Contains(t, action, branchSubURL)
 		}
+
+		filepath, exists := formEl.Find("input[name=path]").Attr("value")
+		assert.True(t, exists)
+		assert.Empty(t, filepath)
+
+		req = NewRequest(t, "GET", "/user2/glob/src/branch/master/x/y")
+		resp = MakeRequest(t, req, http.StatusOK)
+
+		filepath, exists = NewHTMLParser(t, resp.Body).doc.
+			Find("form[data-test-tag=codesearch] input[name=path]").
+			Attr("value")
+		assert.True(t, exists)
+		assert.Equal(t, "x/y", filepath)
 	}
 
 	t.Run("indexer disabled", func(t *testing.T) {

web_src/css/repo.css

@@ -389,6 +389,11 @@ td .commit-summary {
   cursor: default;
 }
 
+.code-search + #repo-files-table {
+  border-top-left-radius: 0;
+  border-top-right-radius: 0;
+}
+
 .view-raw {
   display: flex;
   justify-content: center;