| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 1 | package mapper |
| 2 | |
| 3 | import ( |
| 4 | "fmt" |
| 5 | "strings" |
| 6 | |
| Akron | 2ef703c | 2025-07-03 15:57:42 +0200 | [diff] [blame] | 7 | "github.com/KorAP/Koral-Mapper/ast" |
| 8 | "github.com/KorAP/Koral-Mapper/matcher" |
| 9 | "github.com/KorAP/Koral-Mapper/parser" |
| Akron | a1337ef | 2025-07-01 12:28:03 +0200 | [diff] [blame] | 10 | "github.com/rs/zerolog/log" |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 11 | ) |
| 12 | |
| 13 | // ApplyResponseMappings applies the specified mapping rules to a JSON object |
| 14 | func (m *Mapper) ApplyResponseMappings(mappingID string, opts MappingOptions, jsonData any) (any, error) { |
| 15 | // Validate mapping ID |
| 16 | if _, exists := m.mappingLists[mappingID]; !exists { |
| 17 | return nil, fmt.Errorf("mapping list with ID %s not found", mappingID) |
| 18 | } |
| 19 | |
| Akron | 2f93c58 | 2026-02-19 16:49:13 +0100 | [diff] [blame] | 20 | if m.mappingLists[mappingID].IsCorpus() { |
| 21 | return m.applyCorpusResponseMappings(mappingID, opts, jsonData) |
| 22 | } |
| 23 | |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 24 | // Get the parsed rules |
| Akron | 2f93c58 | 2026-02-19 16:49:13 +0100 | [diff] [blame] | 25 | rules := m.parsedQueryRules[mappingID] |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 26 | |
| 27 | // Check if we have a snippet to process |
| 28 | jsonMap, ok := jsonData.(map[string]any) |
| 29 | if !ok { |
| 30 | return jsonData, nil |
| 31 | } |
| 32 | |
| 33 | snippetValue, exists := jsonMap["snippet"] |
| 34 | if !exists { |
| 35 | return jsonData, nil |
| 36 | } |
| 37 | |
| 38 | snippet, ok := snippetValue.(string) |
| 39 | if !ok { |
| 40 | return jsonData, nil |
| 41 | } |
| 42 | |
| 43 | // Process the snippet with each rule |
| 44 | processedSnippet := snippet |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 45 | for ruleIndex, rule := range rules { |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 46 | // Create pattern and replacement based on direction |
| 47 | var pattern, replacement ast.Node |
| 48 | if opts.Direction { // true means AtoB |
| 49 | pattern = rule.Upper |
| 50 | replacement = rule.Lower |
| 51 | } else { |
| 52 | pattern = rule.Lower |
| 53 | replacement = rule.Upper |
| 54 | } |
| 55 | |
| 56 | // Extract the inner nodes from the pattern and replacement tokens |
| 57 | if token, ok := pattern.(*ast.Token); ok { |
| 58 | pattern = token.Wrap |
| 59 | } |
| 60 | if token, ok := replacement.(*ast.Token); ok { |
| 61 | replacement = token.Wrap |
| 62 | } |
| 63 | |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 64 | // Apply foundry and layer overrides with proper precedence |
| 65 | mappingList := m.mappingLists[mappingID] |
| 66 | |
| 67 | // Determine foundry and layer values based on direction |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 68 | var patternFoundry, patternLayer, replacementFoundry, replacementLayer string |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 69 | if opts.Direction { // AtoB |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 70 | patternFoundry, patternLayer = opts.FoundryA, opts.LayerA |
| 71 | replacementFoundry, replacementLayer = opts.FoundryB, opts.LayerB |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 72 | // Apply mapping list defaults if not specified |
| 73 | if replacementFoundry == "" { |
| 74 | replacementFoundry = mappingList.FoundryB |
| 75 | } |
| 76 | if replacementLayer == "" { |
| 77 | replacementLayer = mappingList.LayerB |
| 78 | } |
| 79 | } else { // BtoA |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 80 | patternFoundry, patternLayer = opts.FoundryB, opts.LayerB |
| 81 | replacementFoundry, replacementLayer = opts.FoundryA, opts.LayerA |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 82 | // Apply mapping list defaults if not specified |
| 83 | if replacementFoundry == "" { |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 84 | replacementFoundry = mappingList.FoundryA |
| Akron | 4de47a9 | 2025-06-27 11:58:11 +0200 | [diff] [blame] | 85 | } |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 86 | if replacementLayer == "" { |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 87 | replacementLayer = mappingList.LayerA |
| 88 | } |
| 89 | } |
| 90 | |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 91 | // Clone pattern and apply foundry and layer overrides |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 92 | processedPattern := pattern.Clone() |
| 93 | if patternFoundry != "" || patternLayer != "" { |
| 94 | ast.ApplyFoundryAndLayerOverrides(processedPattern, patternFoundry, patternLayer) |
| 95 | } |
| 96 | |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 97 | // Create snippet matcher for this rule |
| 98 | snippetMatcher, err := matcher.NewSnippetMatcher( |
| 99 | ast.Pattern{Root: processedPattern}, |
| 100 | ast.Replacement{Root: replacement}, |
| 101 | ) |
| 102 | if err != nil { |
| 103 | continue // Skip this rule if we can't create a matcher |
| 104 | } |
| 105 | |
| 106 | // Find matching tokens in the snippet |
| 107 | matchingTokens, err := snippetMatcher.FindMatchingTokens(processedSnippet) |
| 108 | if err != nil { |
| 109 | continue // Skip this rule if parsing fails |
| 110 | } |
| 111 | |
| 112 | if len(matchingTokens) == 0 { |
| 113 | continue // No matches, try next rule |
| 114 | } |
| 115 | |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 116 | // Apply RestrictToObligatory with layer precedence logic |
| 117 | restrictedReplacement := m.applyReplacementWithLayerPrecedence( |
| 118 | replacement, replacementFoundry, replacementLayer, |
| 119 | mappingID, ruleIndex, bool(opts.Direction)) |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 120 | if restrictedReplacement == nil { |
| 121 | continue // Nothing obligatory to add |
| 122 | } |
| 123 | |
| 124 | // Generate annotation strings from the restricted replacement |
| 125 | annotationStrings, err := m.generateAnnotationStrings(restrictedReplacement) |
| 126 | if err != nil { |
| 127 | continue // Skip if we can't generate annotations |
| 128 | } |
| 129 | |
| 130 | if len(annotationStrings) == 0 { |
| 131 | continue // Nothing to add |
| 132 | } |
| 133 | |
| 134 | // Apply annotations to matching tokens in the snippet |
| 135 | processedSnippet, err = m.addAnnotationsToSnippet(processedSnippet, matchingTokens, annotationStrings) |
| 136 | if err != nil { |
| 137 | continue // Skip if we can't apply annotations |
| 138 | } |
| 139 | } |
| 140 | |
| Akron | a1337ef | 2025-07-01 12:28:03 +0200 | [diff] [blame] | 141 | log.Debug().Str("snippet", processedSnippet).Msg("Processed snippet") |
| 142 | |
| Akron | a3675e9 | 2025-06-26 17:46:59 +0200 | [diff] [blame] | 143 | // Create a copy of the input data and update the snippet |
| 144 | result := make(map[string]any) |
| 145 | for k, v := range jsonMap { |
| 146 | result[k] = v |
| 147 | } |
| 148 | result["snippet"] = processedSnippet |
| 149 | |
| 150 | return result, nil |
| 151 | } |
| 152 | |
| 153 | // generateAnnotationStrings converts a replacement AST node into annotation strings |
| 154 | func (m *Mapper) generateAnnotationStrings(node ast.Node) ([]string, error) { |
| 155 | if node == nil { |
| 156 | return nil, nil |
| 157 | } |
| 158 | |
| 159 | switch n := node.(type) { |
| 160 | case *ast.Term: |
| 161 | // Create annotation string in format "foundry/layer:key" or "foundry/layer:key:value" |
| 162 | annotation := n.Foundry + "/" + n.Layer + ":" + n.Key |
| 163 | if n.Value != "" { |
| 164 | annotation += ":" + n.Value |
| 165 | } |
| 166 | return []string{annotation}, nil |
| 167 | |
| 168 | case *ast.TermGroup: |
| 169 | if n.Relation == ast.AndRelation { |
| 170 | // For AND groups, collect all annotations |
| 171 | var allAnnotations []string |
| 172 | for _, operand := range n.Operands { |
| 173 | annotations, err := m.generateAnnotationStrings(operand) |
| 174 | if err != nil { |
| 175 | return nil, err |
| 176 | } |
| 177 | allAnnotations = append(allAnnotations, annotations...) |
| 178 | } |
| 179 | return allAnnotations, nil |
| 180 | } else { |
| 181 | // For OR groups (should not happen with RestrictToObligatory, but handle gracefully) |
| 182 | return nil, nil |
| 183 | } |
| 184 | |
| 185 | case *ast.Token: |
| 186 | // Handle wrapped tokens |
| 187 | if n.Wrap != nil { |
| 188 | return m.generateAnnotationStrings(n.Wrap) |
| 189 | } |
| 190 | return nil, nil |
| 191 | |
| 192 | default: |
| 193 | return nil, nil |
| 194 | } |
| 195 | } |
| 196 | |
| 197 | // addAnnotationsToSnippet adds new annotations to matching tokens in the snippet |
| 198 | func (m *Mapper) addAnnotationsToSnippet(snippet string, matchingTokens []matcher.TokenSpan, annotationStrings []string) (string, error) { |
| 199 | if len(matchingTokens) == 0 || len(annotationStrings) == 0 { |
| 200 | return snippet, nil |
| 201 | } |
| 202 | |
| 203 | result := snippet |
| 204 | |
| 205 | // Process each matching token |
| 206 | for _, token := range matchingTokens { |
| 207 | // For nested span structure, we need to find the innermost text and wrap it |
| 208 | // Look for the actual token text within span tags |
| 209 | tokenText := token.Text |
| 210 | |
| 211 | // Find all occurrences of the token text in the current snippet |
| 212 | // We need to be careful about which occurrence to replace |
| 213 | startPos := 0 |
| 214 | for { |
| 215 | tokenStart := strings.Index(result[startPos:], tokenText) |
| 216 | if tokenStart == -1 { |
| 217 | break // No more occurrences |
| 218 | } |
| 219 | tokenStart += startPos |
| 220 | tokenEnd := tokenStart + len(tokenText) |
| 221 | |
| 222 | // Check if this token text is within the expected context |
| 223 | // Look backwards and forwards to see if we're in the right span context |
| 224 | beforeContext := result[:tokenStart] |
| 225 | afterContext := result[tokenEnd:] |
| 226 | |
| 227 | // Simple heuristic: if we're immediately preceded by a > and followed by a < |
| 228 | // then we're likely at the innermost text node |
| 229 | if strings.HasSuffix(beforeContext, ">") && (strings.HasPrefix(afterContext, "<") || len(afterContext) == 0 || afterContext[0] == ' ') { |
| 230 | // Build the replacement with nested spans for each annotation |
| 231 | replacement := tokenText |
| 232 | for i := len(annotationStrings) - 1; i >= 0; i-- { |
| 233 | replacement = fmt.Sprintf(`<span title="%s" class="notinindex">%s</span>`, annotationStrings[i], replacement) |
| 234 | } |
| 235 | |
| 236 | // Replace this occurrence |
| 237 | result = result[:tokenStart] + replacement + result[tokenEnd:] |
| 238 | break // Only replace the first appropriate occurrence for this token |
| 239 | } |
| 240 | |
| 241 | // Move past this occurrence |
| 242 | startPos = tokenEnd |
| 243 | } |
| 244 | } |
| 245 | |
| 246 | return result, nil |
| 247 | } |
| Akron | 497cfe8 | 2025-07-03 13:26:54 +0200 | [diff] [blame] | 248 | |
| 249 | // applyReplacementWithLayerPrecedence applies RestrictToObligatory with proper layer precedence |
| 250 | func (m *Mapper) applyReplacementWithLayerPrecedence( |
| 251 | replacement ast.Node, foundry, layerOverride string, |
| 252 | mappingID string, ruleIndex int, direction bool) ast.Node { |
| 253 | |
| 254 | // First, apply RestrictToObligatory without layer override to preserve explicit layers |
| 255 | restricted := ast.RestrictToObligatory(replacement, foundry, "") |
| 256 | if restricted == nil { |
| 257 | return nil |
| 258 | } |
| 259 | |
| 260 | // If no layer override is specified, we're done |
| 261 | if layerOverride == "" { |
| 262 | return restricted |
| 263 | } |
| 264 | |
| 265 | // Apply layer override only to terms that didn't have explicit layers in the original rule |
| 266 | mappingList := m.mappingLists[mappingID] |
| 267 | if ruleIndex < len(mappingList.Mappings) { |
| 268 | originalRule := string(mappingList.Mappings[ruleIndex]) |
| 269 | m.applySelectiveLayerOverrides(restricted, layerOverride, originalRule, direction) |
| 270 | } |
| 271 | |
| 272 | return restricted |
| 273 | } |
| 274 | |
| 275 | // applySelectiveLayerOverrides applies layer overrides only to terms without explicit layers |
| 276 | func (m *Mapper) applySelectiveLayerOverrides(node ast.Node, layerOverride, originalRule string, direction bool) { |
| 277 | if node == nil { |
| 278 | return |
| 279 | } |
| 280 | |
| 281 | // Parse the original rule without defaults to detect explicit layers |
| 282 | explicitTerms := m.getExplicitTerms(originalRule, direction) |
| 283 | |
| 284 | // Apply overrides only to terms that weren't explicit in the original rule |
| 285 | termIndex := 0 |
| 286 | m.applyLayerOverrideToImplicitTerms(node, layerOverride, explicitTerms, &termIndex) |
| 287 | } |
| 288 | |
| 289 | // getExplicitTerms parses the original rule without defaults to identify terms with explicit layers |
| 290 | func (m *Mapper) getExplicitTerms(originalRule string, direction bool) map[int]bool { |
| 291 | explicitTerms := make(map[int]bool) |
| 292 | |
| 293 | // Parse without defaults to see what was explicitly specified |
| 294 | parser, err := parser.NewGrammarParser("", "") |
| 295 | if err != nil { |
| 296 | return explicitTerms |
| 297 | } |
| 298 | |
| 299 | result, err := parser.ParseMapping(originalRule) |
| 300 | if err != nil { |
| 301 | return explicitTerms |
| 302 | } |
| 303 | |
| 304 | // Get the replacement side based on direction |
| 305 | var replacement ast.Node |
| 306 | if direction { // AtoB |
| 307 | replacement = result.Lower.Wrap |
| 308 | } else { // BtoA |
| 309 | replacement = result.Upper.Wrap |
| 310 | } |
| 311 | |
| 312 | // Extract terms and check which ones have explicit layers |
| 313 | termIndex := 0 |
| 314 | m.markExplicitTerms(replacement, explicitTerms, &termIndex) |
| 315 | return explicitTerms |
| 316 | } |
| 317 | |
| 318 | // markExplicitTerms recursively marks terms that have explicit layers |
| 319 | func (m *Mapper) markExplicitTerms(node ast.Node, explicitTerms map[int]bool, termIndex *int) { |
| 320 | if node == nil { |
| 321 | return |
| 322 | } |
| 323 | |
| 324 | switch n := node.(type) { |
| 325 | case *ast.Term: |
| 326 | // A term has an explicit layer if it was specified in the original rule |
| 327 | if n.Layer != "" { |
| 328 | explicitTerms[*termIndex] = true |
| 329 | } |
| 330 | *termIndex++ |
| 331 | |
| 332 | case *ast.TermGroup: |
| 333 | for _, operand := range n.Operands { |
| 334 | m.markExplicitTerms(operand, explicitTerms, termIndex) |
| 335 | } |
| 336 | |
| 337 | case *ast.Token: |
| 338 | if n.Wrap != nil { |
| 339 | m.markExplicitTerms(n.Wrap, explicitTerms, termIndex) |
| 340 | } |
| 341 | } |
| 342 | } |
| 343 | |
| 344 | // applyLayerOverrideToImplicitTerms applies layer override only to terms not marked as explicit |
| 345 | func (m *Mapper) applyLayerOverrideToImplicitTerms(node ast.Node, layerOverride string, explicitTerms map[int]bool, termIndex *int) { |
| 346 | if node == nil { |
| 347 | return |
| 348 | } |
| 349 | |
| 350 | switch n := node.(type) { |
| 351 | case *ast.Term: |
| 352 | // Apply override only if this term wasn't explicit in the original rule |
| 353 | if !explicitTerms[*termIndex] && n.Layer != "" { |
| 354 | n.Layer = layerOverride |
| 355 | } |
| 356 | *termIndex++ |
| 357 | |
| 358 | case *ast.TermGroup: |
| 359 | for _, operand := range n.Operands { |
| 360 | m.applyLayerOverrideToImplicitTerms(operand, layerOverride, explicitTerms, termIndex) |
| 361 | } |
| 362 | |
| 363 | case *ast.Token: |
| 364 | if n.Wrap != nil { |
| 365 | m.applyLayerOverrideToImplicitTerms(n.Wrap, layerOverride, explicitTerms, termIndex) |
| 366 | } |
| 367 | } |
| 368 | } |