Akron | cc9a8a6 | 2025-06-25 11:56:28 +0200 | [diff] [blame^] | 1 | package matcher |
| 2 | |
| 3 | import ( |
| 4 | "fmt" |
| 5 | "sort" |
| 6 | "strings" |
| 7 | |
| 8 | "github.com/KorAP/KoralPipe-TermMapper/ast" |
| 9 | "github.com/KorAP/KoralPipe-TermMapper/parser" |
| 10 | "github.com/orisano/gosax" |
| 11 | ) |
| 12 | |
| 13 | // TokenSpan represents a token and its position in the snippet |
| 14 | type TokenSpan struct { |
| 15 | Text string // The actual token text |
| 16 | StartPos int // Character position where the token starts |
| 17 | EndPos int // Character position where the token ends |
| 18 | Annotations []string // All title attributes that annotate this token |
| 19 | } |
| 20 | |
| 21 | // SnippetMatcher extends the basic matcher to work with HTML/XML snippets |
| 22 | type SnippetMatcher struct { |
| 23 | matcher *Matcher |
| 24 | titleParser *parser.TitleAttributeParser |
| 25 | } |
| 26 | |
| 27 | // NewSnippetMatcher creates a new snippet matcher |
| 28 | func NewSnippetMatcher(pattern ast.Pattern, replacement ast.Replacement) (*SnippetMatcher, error) { |
| 29 | matcher, err := NewMatcher(pattern, replacement) |
| 30 | if err != nil { |
| 31 | return nil, fmt.Errorf("failed to create base matcher: %w", err) |
| 32 | } |
| 33 | |
| 34 | return &SnippetMatcher{ |
| 35 | matcher: matcher, |
| 36 | titleParser: parser.NewTitleAttributeParser(), |
| 37 | }, nil |
| 38 | } |
| 39 | |
| 40 | // ParseSnippet parses an HTML/XML snippet and extracts tokens with their annotations |
| 41 | func (sm *SnippetMatcher) ParseSnippet(snippet string) ([]TokenSpan, error) { |
| 42 | tokens := make([]TokenSpan, 0) |
| 43 | |
| 44 | // Stack to track nested spans and their annotations |
| 45 | type spanInfo struct { |
| 46 | title string |
| 47 | level int |
| 48 | } |
| 49 | spanStack := make([]spanInfo, 0) |
| 50 | |
| 51 | // Current position tracking |
| 52 | var currentPos int |
| 53 | |
| 54 | reader := strings.NewReader(snippet) |
| 55 | r := gosax.NewReader(reader) |
| 56 | |
| 57 | for { |
| 58 | e, err := r.Event() |
| 59 | if err != nil { |
| 60 | return nil, fmt.Errorf("failed to parse snippet: %w", err) |
| 61 | } |
| 62 | |
| 63 | if e.Type() == 8 { // gosax.EventEOF |
| 64 | break |
| 65 | } |
| 66 | |
| 67 | switch e.Type() { |
| 68 | case 1: // gosax.EventStart |
| 69 | // Parse start element |
| 70 | startElem, err := gosax.StartElement(e.Bytes) |
| 71 | if err != nil { |
| 72 | continue // Skip invalid elements |
| 73 | } |
| 74 | |
| 75 | if startElem.Name.Local == "span" { |
| 76 | // Look for title attribute |
| 77 | var title string |
| 78 | for _, attr := range startElem.Attr { |
| 79 | if attr.Name.Local == "title" { |
| 80 | title = attr.Value |
| 81 | break |
| 82 | } |
| 83 | } |
| 84 | spanStack = append(spanStack, spanInfo{title: title, level: len(spanStack)}) |
| 85 | } |
| 86 | |
| 87 | case 2: // gosax.EventEnd |
| 88 | // Parse end element |
| 89 | endElem := gosax.EndElement(e.Bytes) |
| 90 | if endElem.Name.Local == "span" && len(spanStack) > 0 { |
| 91 | spanStack = spanStack[:len(spanStack)-1] |
| 92 | } |
| 93 | |
| 94 | case 3: // gosax.EventText |
| 95 | // Process character data |
| 96 | charData, err := gosax.CharData(e.Bytes) |
| 97 | if err != nil { |
| 98 | continue |
| 99 | } |
| 100 | |
| 101 | text := string(charData) |
| 102 | trimmed := strings.TrimSpace(text) |
| 103 | if trimmed != "" && len(spanStack) > 0 { |
| 104 | // Only create tokens if we're inside at least one span |
| 105 | // Collect all annotations from the current span stack |
| 106 | annotations := make([]string, 0) |
| 107 | for _, span := range spanStack { |
| 108 | if span.title != "" { |
| 109 | annotations = append(annotations, span.title) |
| 110 | } |
| 111 | } |
| 112 | |
| 113 | // Create token span |
| 114 | token := TokenSpan{ |
| 115 | Text: trimmed, |
| 116 | StartPos: currentPos, |
| 117 | EndPos: currentPos + len(trimmed), |
| 118 | Annotations: annotations, |
| 119 | } |
| 120 | tokens = append(tokens, token) |
| 121 | } |
| 122 | currentPos += len(text) |
| 123 | } |
| 124 | } |
| 125 | |
| 126 | // Sort tokens by start position to ensure proper order |
| 127 | sort.Slice(tokens, func(i, j int) bool { |
| 128 | return tokens[i].StartPos < tokens[j].StartPos |
| 129 | }) |
| 130 | |
| 131 | return tokens, nil |
| 132 | } |
| 133 | |
| 134 | // CheckToken checks if a token's annotations match the pattern |
| 135 | func (sm *SnippetMatcher) CheckToken(token TokenSpan) (bool, error) { |
| 136 | if len(token.Annotations) == 0 { |
| 137 | return false, nil |
| 138 | } |
| 139 | |
| 140 | // Parse all annotations into AST terms |
| 141 | terms, err := sm.titleParser.ParseTitleAttributesToTerms(token.Annotations) |
| 142 | if err != nil { |
| 143 | return false, fmt.Errorf("failed to parse token annotations: %w", err) |
| 144 | } |
| 145 | |
| 146 | if len(terms) == 0 { |
| 147 | return false, nil |
| 148 | } |
| 149 | |
| 150 | // Create a TermGroup with AND relation for all annotations |
| 151 | var nodeToMatch ast.Node |
| 152 | if len(terms) == 1 { |
| 153 | nodeToMatch = terms[0] |
| 154 | } else { |
| 155 | nodeToMatch = &ast.TermGroup{ |
| 156 | Operands: terms, |
| 157 | Relation: ast.AndRelation, |
| 158 | } |
| 159 | } |
| 160 | |
| 161 | // Check if the constructed node matches our pattern |
| 162 | return sm.matcher.Match(nodeToMatch), nil |
| 163 | } |
| 164 | |
| 165 | // CheckTokenSequence checks if a sequence of tokens matches the pattern |
| 166 | func (sm *SnippetMatcher) CheckTokenSequence(tokens []TokenSpan) (bool, error) { |
| 167 | if len(tokens) == 0 { |
| 168 | return false, nil |
| 169 | } |
| 170 | |
| 171 | // For token sequences, we need to check different strategies: |
| 172 | // 1. Check if any individual token matches |
| 173 | // 2. Check if the combined annotations of all tokens match |
| 174 | |
| 175 | // Strategy 1: Check individual tokens |
| 176 | for _, token := range tokens { |
| 177 | matches, err := sm.CheckToken(token) |
| 178 | if err != nil { |
| 179 | return false, err |
| 180 | } |
| 181 | if matches { |
| 182 | return true, nil |
| 183 | } |
| 184 | } |
| 185 | |
| 186 | // Strategy 2: Check combined annotations |
| 187 | allAnnotations := make([]string, 0) |
| 188 | for _, token := range tokens { |
| 189 | allAnnotations = append(allAnnotations, token.Annotations...) |
| 190 | } |
| 191 | |
| 192 | // Remove duplicates from combined annotations |
| 193 | annotationMap := make(map[string]bool) |
| 194 | uniqueAnnotations := make([]string, 0) |
| 195 | for _, annotation := range allAnnotations { |
| 196 | if !annotationMap[annotation] { |
| 197 | annotationMap[annotation] = true |
| 198 | uniqueAnnotations = append(uniqueAnnotations, annotation) |
| 199 | } |
| 200 | } |
| 201 | |
| 202 | if len(uniqueAnnotations) == 0 { |
| 203 | return false, nil |
| 204 | } |
| 205 | |
| 206 | // Create a combined token for checking |
| 207 | combinedToken := TokenSpan{ |
| 208 | Text: strings.Join(getTokenTexts(tokens), " "), |
| 209 | StartPos: tokens[0].StartPos, |
| 210 | EndPos: tokens[len(tokens)-1].EndPos, |
| 211 | Annotations: uniqueAnnotations, |
| 212 | } |
| 213 | |
| 214 | return sm.CheckToken(combinedToken) |
| 215 | } |
| 216 | |
| 217 | // FindMatchingTokens finds all tokens in the snippet that match the pattern |
| 218 | func (sm *SnippetMatcher) FindMatchingTokens(snippet string) ([]TokenSpan, error) { |
| 219 | tokens, err := sm.ParseSnippet(snippet) |
| 220 | if err != nil { |
| 221 | return nil, err |
| 222 | } |
| 223 | |
| 224 | matchingTokens := make([]TokenSpan, 0) |
| 225 | |
| 226 | for _, token := range tokens { |
| 227 | matches, err := sm.CheckToken(token) |
| 228 | if err != nil { |
| 229 | return nil, fmt.Errorf("failed to check token '%s': %w", token.Text, err) |
| 230 | } |
| 231 | if matches { |
| 232 | matchingTokens = append(matchingTokens, token) |
| 233 | } |
| 234 | } |
| 235 | |
| 236 | return matchingTokens, nil |
| 237 | } |
| 238 | |
| 239 | // FindMatchingTokenSequences finds all token sequences that match the pattern |
| 240 | func (sm *SnippetMatcher) FindMatchingTokenSequences(snippet string, maxSequenceLength int) ([][]TokenSpan, error) { |
| 241 | tokens, err := sm.ParseSnippet(snippet) |
| 242 | if err != nil { |
| 243 | return nil, err |
| 244 | } |
| 245 | |
| 246 | if maxSequenceLength <= 0 { |
| 247 | maxSequenceLength = len(tokens) |
| 248 | } |
| 249 | |
| 250 | matchingSequences := make([][]TokenSpan, 0) |
| 251 | |
| 252 | // Check all possible token sequences up to maxSequenceLength |
| 253 | for start := 0; start < len(tokens); start++ { |
| 254 | for length := 1; length <= maxSequenceLength && start+length <= len(tokens); length++ { |
| 255 | sequence := tokens[start : start+length] |
| 256 | |
| 257 | matches, err := sm.CheckTokenSequence(sequence) |
| 258 | if err != nil { |
| 259 | return nil, fmt.Errorf("failed to check token sequence: %w", err) |
| 260 | } |
| 261 | if matches { |
| 262 | matchingSequences = append(matchingSequences, sequence) |
| 263 | } |
| 264 | } |
| 265 | } |
| 266 | |
| 267 | return matchingSequences, nil |
| 268 | } |
| 269 | |
| 270 | // GetReplacement returns the replacement node from the matcher |
| 271 | func (sm *SnippetMatcher) GetReplacement() ast.Node { |
| 272 | return sm.matcher.replacement.Root |
| 273 | } |
| 274 | |
| 275 | // Helper function to extract token texts |
| 276 | func getTokenTexts(tokens []TokenSpan) []string { |
| 277 | texts := make([]string, len(tokens)) |
| 278 | for i, token := range tokens { |
| 279 | texts[i] = token.Text |
| 280 | } |
| 281 | return texts |
| 282 | } |