初始提交: Gitea 项目代码

modules/indexer/code/bleve/token/path/path.go

@@ -0,0 +1,105 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package path
+
+import (
+	"slices"
+	"strings"
+
+	"github.com/blevesearch/bleve/v2/analysis"
+	"github.com/blevesearch/bleve/v2/registry"
+)
+
+const (
+	Name = "gitea/path"
+)
+
+type TokenFilter struct{}
+
+func NewTokenFilter() *TokenFilter {
+	return &TokenFilter{}
+}
+
+func TokenFilterConstructor(config map[string]any, cache *registry.Cache) (analysis.TokenFilter, error) {
+	return NewTokenFilter(), nil
+}
+
+func (s *TokenFilter) Filter(input analysis.TokenStream) analysis.TokenStream {
+	if len(input) == 1 {
+		// if there is only one token, we don't need to generate the reversed chain
+		return generatePathTokens(input, false)
+	}
+
+	normal := generatePathTokens(input, false)
+	reversed := generatePathTokens(input, true)
+
+	return append(normal, reversed...)
+}
+
+// Generates path tokens from the input tokens.
+// This mimics the behavior of the path hierarchy tokenizer in ES. It takes the input tokens and combine them, generating a term for each component
+// in tree (e.g., foo/bar/baz.md will generate foo, foo/bar, and foo/bar/baz.md).
+//
+// If the reverse flag is set, the order of the tokens is reversed (the same input will generate baz.md, baz.md/bar, baz.md/bar/foo). This is useful
+// to efficiently search for filenames without supplying the fullpath.
+func generatePathTokens(input analysis.TokenStream, reversed bool) analysis.TokenStream {
+	terms := make([]string, 0, len(input))
+	longestTerm := 0
+
+	if reversed {
+		slices.Reverse(input)
+	}
+
+	for i := range input {
+		var sb strings.Builder
+		sb.Write(input[0].Term)
+
+		for j := 1; j < i; j++ {
+			sb.WriteString("/")
+			sb.Write(input[j].Term)
+		}
+
+		term := sb.String()
+
+		if longestTerm < len(term) {
+			longestTerm = len(term)
+		}
+
+		terms = append(terms, term)
+	}
+
+	output := make(analysis.TokenStream, 0, len(terms))
+
+	for _, term := range terms {
+		var start, end int
+
+		if reversed {
+			start = 0
+			end = len(term)
+		} else {
+			start = longestTerm - len(term)
+			end = longestTerm
+		}
+
+		token := analysis.Token{
+			Position: 1,
+			Start:    start,
+			End:      end,
+			Type:     analysis.AlphaNumeric,
+			Term:     []byte(term),
+		}
+
+		output = append(output, &token)
+	}
+
+	return output
+}
+
+func init() {
+	// FIXME: move it to the bleve's init function, but do not call it in global init
+	err := registry.RegisterTokenFilter(Name, TokenFilterConstructor)
+	if err != nil {
+		panic(err)
+	}
+}

modules/indexer/code/bleve/token/path/path_test.go

@@ -0,0 +1,76 @@
+// Copyright 2024 The Gitea Authors. All rights reserved.
+// SPDX-License-Identifier: MIT
+
+package path
+
+import (
+	"fmt"
+	"testing"
+
+	"github.com/blevesearch/bleve/v2/analysis"
+	"github.com/blevesearch/bleve/v2/analysis/tokenizer/unicode"
+	"github.com/stretchr/testify/assert"
+)
+
+type Scenario struct {
+	Input  string
+	Tokens []string
+}
+
+func TestTokenFilter(t *testing.T) {
+	scenarios := []struct {
+		Input string
+		Terms []string
+	}{
+		{
+			Input: "Dockerfile",
+			Terms: []string{"Dockerfile"},
+		},
+		{
+			Input: "Dockerfile.rootless",
+			Terms: []string{"Dockerfile.rootless"},
+		},
+		{
+			Input: "a/b/c/Dockerfile.rootless",
+			Terms: []string{"a", "a/b", "a/b/c", "a/b/c/Dockerfile.rootless", "Dockerfile.rootless", "Dockerfile.rootless/c", "Dockerfile.rootless/c/b", "Dockerfile.rootless/c/b/a"},
+		},
+		{
+			Input: "",
+			Terms: []string{},
+		},
+	}
+
+	for _, scenario := range scenarios {
+		t.Run(fmt.Sprintf("ensure terms of '%s'", scenario.Input), func(t *testing.T) {
+			terms := extractTerms(scenario.Input)
+
+			assert.Len(t, terms, len(scenario.Terms))
+
+			for _, term := range terms {
+				assert.Contains(t, scenario.Terms, term)
+			}
+		})
+	}
+}
+
+func extractTerms(input string) []string {
+	tokens := tokenize(input)
+	filteredTokens := filter(tokens)
+	terms := make([]string, 0, len(filteredTokens))
+
+	for _, token := range filteredTokens {
+		terms = append(terms, string(token.Term))
+	}
+
+	return terms
+}
+
+func filter(input analysis.TokenStream) analysis.TokenStream {
+	filter := NewTokenFilter()
+	return filter.Filter(input)
+}
+
+func tokenize(input string) analysis.TokenStream {
+	tokenizer := unicode.NewUnicodeTokenizer()
+	return tokenizer.Tokenize([]byte(input))
+}