fix issues with type0

app/proprietary/src/main/java/stirling/software/SPDF/service/PdfJsonConversionService.java

@@ -311,13 +311,17 @@ public class PdfJsonConversionService {
 
             try (PDDocument document = pdfDocumentFactory.load(workingPath, true)) {
                 int totalPages = document.getNumberOfPages();
-                boolean useLazyImages = totalPages > 5 && jobId != null;
+                // Only use lazy images for real async jobs where client can access the cache
+                // Synchronous calls with synthetic jobId should do full extraction
+                boolean useLazyImages = totalPages > 5 && isRealJobId;
                 Map<COSBase, FontModelCacheEntry> fontCache = new IdentityHashMap<>();
                 Map<COSBase, EncodedImage> imageCache = new IdentityHashMap<>();
                 log.debug(
-                        "Converting PDF to JSON ({} pages) - {} mode",
+                        "Converting PDF to JSON ({} pages) - {} mode (jobId: {}, isRealJobId: {})",
                         totalPages,
-                        useLazyImages ? "lazy image" : "standard");
+                        useLazyImages ? "lazy image" : "standard",
+                        jobId,
+                        isRealJobId);
                 Map<String, PdfJsonFont> fonts = new LinkedHashMap<>();
                 Map<Integer, List<PdfJsonTextElement>> textByPage = new LinkedHashMap<>();
                 Map<Integer, Map<PDFont, String>> pageFontResources = new HashMap<>();
@@ -327,7 +331,8 @@ public class PdfJsonConversionService {
                 int pageNumber = 1;
                 for (PDPage page : document.getPages()) {
                     Map<PDFont, String> resourceMap =
-                            collectFontsForPage(document, page, pageNumber, fonts, fontCache, jobId);
+                            collectFontsForPage(
+                                    document, page, pageNumber, fonts, fontCache, jobId);
                     pageFontResources.put(pageNumber, resourceMap);
                     log.debug(
                             "PDF→JSON: collected {} font resources on page {}",
@@ -444,8 +449,9 @@ public class PdfJsonConversionService {
                 byte[] result = objectMapper.writeValueAsBytes(pdfJson);
                 progress.accept(PdfJsonConversionProgress.complete());
 
-                // If document wasn't cached, clear Type3 cache entries immediately
+                // Clear Type3 cache entries immediately for non-cached conversions
-                // (jobId is always set now, either from request context or synthetic)
+                // Cached conversions (useLazyImages=true) are cleaned when cache expires
+                // Synchronous conversions always clear immediately since they don't use lazy mode
                 if (!useLazyImages) {
                     clearType3CacheEntriesForJob(jobId);
                 }
@@ -718,7 +724,8 @@ public class PdfJsonConversionService {
             mapping.put(font, fontId);
             String key = buildFontKey(jobId, pageNumber, fontId);
             if (!fonts.containsKey(key)) {
-                fonts.put(key, buildFontModel(document, font, fontId, pageNumber, fontCache, jobId));
+                fonts.put(
+                        key, buildFontModel(document, font, fontId, pageNumber, fontCache, jobId));
             }
         }
 
@@ -873,15 +880,13 @@ public class PdfJsonConversionService {
                             || hasPayload(font.getWebProgram())
                             || hasPayload(font.getProgram());
 
-            // Keep cosDictionary for TrueType and Type0 fonts even with usable program
+            // Only clear cosDictionary for Type3 fonts (which have inline content streams)
-            // Subsetted fonts need the ToUnicode CMap from the original dictionary
+            // All other font types may need ToUnicode CMap or encoding from the dictionary
+            // Conservative approach: better to keep extra data than lose encoding info
             String subtype = font.getSubtype();
-            boolean needsCosDictionary =
+            boolean isType3 = subtype != null && subtype.equalsIgnoreCase("Type3");
-                    subtype != null
-                            && (subtype.equalsIgnoreCase("TrueType")
-                                    || subtype.equalsIgnoreCase("Type0"));
 
-            if (hasUsableProgram && !needsCosDictionary) {
+            if (hasUsableProgram && isType3) {
                 font.setCosDictionary(null);
             }
         }
@@ -1380,9 +1385,20 @@ public class PdfJsonConversionService {
         String key = buildFontKey(null, -1, effectiveId);
         PDFont font = fontMap.get(key);
         if (font != null) {
+            log.debug(
+                    "[FALLBACK-DEBUG] Reusing cached fallback font {} (key: {})", effectiveId, key);
             return font;
         }
+        log.info(
+                "[FALLBACK-DEBUG] Loading fallback font {} (key: {}) via fallbackFontService",
+                effectiveId,
+                key);
         PDFont loaded = fallbackFontService.loadFallbackPdfFont(document, effectiveId);
+        log.info(
+                "[FALLBACK-DEBUG] Loaded fallback font {} - PDFont class: {}, name: {}",
+                effectiveId,
+                loaded.getClass().getSimpleName(),
+                loaded.getName());
         fontMap.put(key, loaded);
         if (fontModels != null
                 && fontModels.stream().noneMatch(f -> effectiveId.equals(f.getId()))) {
@@ -2561,10 +2577,16 @@ public class PdfJsonConversionService {
                                             && runFontModel.getType3Glyphs() != null
                                             && !runFontModel.getType3Glyphs().isEmpty();
 
-                            if (isNormalizedType3) {
+                            // For fonts with proper Unicode mappings, let PDFBox handle encoding
-                                // For normalized Type3 fonts, use original text directly
+                            // This includes: normalized Type3 fonts, PDType0Font (composite fonts)
-                                // The font has proper Unicode mappings, so PDFBox can encode it
+                            boolean useDirectText =
-                                // correctly
+                                    isNormalizedType3
+                                            || run.font()
+                                                    instanceof
+                                                    org.apache.pdfbox.pdmodel.font.PDType0Font;
+
+                            if (useDirectText) {
+                                // Pass text directly - PDFBox handles encoding internally
                                 contentStream.showText(run.text());
                             } else {
                                 // For actual Type3 fonts and other fonts, encode manually
@@ -2582,6 +2604,14 @@ public class PdfJsonConversionService {
                                     }
                                 } else {
                                     try {
+                                        log.debug(
+                                                "[ENCODE-DEBUG] Encoding text '{}' with font {} (fontId={}, runFontModel={})",
+                                                run.text(),
+                                                run.font().getName(),
+                                                run.fontId(),
+                                                runFontModel != null
+                                                        ? runFontModel.getId()
+                                                        : "null");
                                         encoded =
                                                 encodeTextWithFont(
                                                         run.font(),
@@ -2590,9 +2620,13 @@ public class PdfJsonConversionService {
                                                         run.charCodes());
                                     } catch (IOException ex) {
                                         log.warn(
-                                                "Failed to encode text '{}' with font {} on page {}: {}",
+                                                "Failed to encode text '{}' with font {} (fontId={}, runFontModel={}) on page {}: {}",
                                                 run.text(),
                                                 run.font().getName(),
+                                                run.fontId(),
+                                                runFontModel != null
+                                                        ? runFontModel.getId()
+                                                        : "null",
                                                 pageNumber,
                                                 ex.getMessage());
                                         continue;
@@ -2725,7 +2759,11 @@ public class PdfJsonConversionService {
             }
             if (targetFont == null || !fallbackFontService.canEncode(targetFont, glyph)) {
                 fallbackApplied = true;
-                String fallbackId = fallbackFontService.resolveFallbackFontId(codePoint);
+                // Try to match fallback font to original font family for visual consistency
+                String originalFontName =
+                        baseFontModel != null ? baseFontModel.getBaseName() : null;
+                String fallbackId =
+                        fallbackFontService.resolveFallbackFontId(originalFontName, codePoint);
                 targetFont = ensureFallbackFont(document, fontMap, fontModels, fallbackId);
                 targetFontId = fallbackId != null ? fallbackId : FALLBACK_FONT_ID;
                 if (targetFont == null || !fallbackFontService.canEncode(targetFont, glyph)) {
@@ -3335,7 +3373,8 @@ public class PdfJsonConversionService {
             // or return null to trigger fallback font
         } else if (!isType3Font || fontModel == null) {
             // For non-Type3 fonts without Type3 metadata, use standard encoding
-            return sanitizeEncoded(font.encode(text));
+            byte[] encoded = font.encode(text);
+            return sanitizeEncoded(encoded);
         }
 
         // Type3 glyph mapping logic (for actual Type3 fonts AND normalized Type3 fonts)
@@ -3750,20 +3789,43 @@ public class PdfJsonConversionService {
             }
             // Fall through to Standard14 fallback below if nothing else succeeded.
         } else {
+            // For TrueType and Type0 fonts, prioritize cosDictionary restoration
+            // These fonts often use ToUnicode CMap which is preserved in the dictionary
+            String subtype = fontModel.getSubtype();
+            boolean preferDictionary =
+                    subtype != null
+                            && (subtype.equalsIgnoreCase("TrueType")
+                                    || subtype.equalsIgnoreCase("Type0"));
+
+            if (preferDictionary) {
+                PDFont restored = restoreFontFromDictionary(document, fontModel);
+                if (restored != null) {
+                    log.debug(
+                            "Font {} restored from cosDictionary (preferred for subsetted {})",
+                            fontModel.getId(),
+                            subtype);
+                    return restored;
+                }
+                // If dictionary restoration fails, fall back to font program bytes
+                log.debug(
+                        "Font {} cosDictionary restoration failed, trying font program bytes",
+                        fontModel.getId());
+            }
+
             PDFont loaded =
                     loadFirstAvailableFont(document, fontModel, orderedCandidates, originalFormat);
             if (loaded != null) {
                 return loaded;
             }
-        }
 
-        // Try to restore from COS dictionary if font programs failed
+            // Try to restore from COS dictionary if font programs failed and we haven't tried yet
-        if (!isType3Font) {
+            if (!preferDictionary) {
                 PDFont restored = restoreFontFromDictionary(document, fontModel);
                 if (restored != null) {
                     return restored;
                 }
             }
+        }
 
         for (FontByteSource source : orderedCandidates) {
             byte[] fontBytes = source.bytes();
@@ -3972,34 +4034,74 @@ public class PdfJsonConversionService {
             log.debug("[FONT-RESTORE] Font {} has no cosDictionary", fontModel.getId());
             return null;
         }
-        COSBase restored = cosMapper.deserializeCosValue(fontModel.getCosDictionary(), document);
+
+        // Deserialize the cosDictionary - cosMapper handles validation internally
+        COSBase restored;
+        try {
+            restored = cosMapper.deserializeCosValue(fontModel.getCosDictionary(), document);
+        } catch (Exception ex) {
+            log.warn(
+                    "[FONT-RESTORE] Font {} cosDictionary deserialization failed: {}",
+                    fontModel.getId(),
+                    ex.getMessage());
+            return null;
+        }
+
         if (!(restored instanceof COSDictionary cosDictionary)) {
-            log.debug(
+            log.warn(
                     "[FONT-RESTORE] Font {} cosDictionary deserialized to {} instead of COSDictionary",
                     fontModel.getId(),
                     restored != null ? restored.getClass().getSimpleName() : "null");
             return null;
         }
+
+        // Validate that dictionary contains required font keys
+        if (!cosDictionary.containsKey(org.apache.pdfbox.cos.COSName.TYPE)
+                || !cosDictionary.containsKey(org.apache.pdfbox.cos.COSName.SUBTYPE)) {
+            log.warn(
+                    "[FONT-RESTORE] Font {} cosDictionary missing required Type or Subtype keys",
+                    fontModel.getId());
+            return null;
+        }
+
         try {
             PDFont font = PDFontFactory.createFont(cosDictionary);
-            if (font != null && font.isEmbedded()) {
+            if (font == null) {
+                log.warn(
+                        "[FONT-RESTORE] Font {} PDFontFactory returned null for valid dictionary",
+                        fontModel.getId());
+                return null;
+            }
+
+            if (!font.isEmbedded()) {
+                log.warn(
+                        "[FONT-RESTORE] Font {} restored from dictionary but is not embedded; rejecting to avoid system font substitution",
+                        fontModel.getId());
+                return null;
+            }
+
             applyAdditionalFontMetadata(document, font, fontModel);
             log.info(
-                        "[FONT-RESTORE] Successfully restored embedded font {} from original dictionary",
+                    "[FONT-RESTORE] Successfully restored embedded font {} (subtype={}) from original dictionary",
-                        fontModel.getId());
-                return font;
-            }
-            log.warn(
-                    "[FONT-RESTORE] Restored font {} from dictionary but font was {}embedded; rejecting",
                     fontModel.getId(),
-                    font != null && font.isEmbedded() ? "" : "not ");
+                    font.getSubType());
+            return font;
+
         } catch (IOException ex) {
             log.warn(
-                    "[FONT-RESTORE] Failed to restore font {} from stored dictionary: {}",
+                    "[FONT-RESTORE] Failed to restore font {} from dictionary ({}): {}",
                     fontModel.getId(),
+                    fontModel.getSubtype(),
                     ex.getMessage());
-        }
             return null;
+        } catch (Exception ex) {
+            log.error(
+                    "[FONT-RESTORE] Unexpected error restoring font {} from dictionary: {}",
+                    fontModel.getId(),
+                    ex.getMessage(),
+                    ex);
+            return null;
+        }
     }
 
     private boolean isType1Format(String format) {
@@ -4948,7 +5050,8 @@ public class PdfJsonConversionService {
             }
             String key = buildFontKey(jobId, currentPage, fontId);
             if (!fonts.containsKey(key)) {
-                fonts.put(key, buildFontModel(document, font, fontId, currentPage, fontCache, jobId));
+                fonts.put(
+                        key, buildFontModel(document, font, fontId, currentPage, fontCache, jobId));
             }
             return fontId;
         }
@@ -5514,8 +5617,8 @@ public class PdfJsonConversionService {
     }
 
     /**
-     * Clear job-specific entries from Type3 font caches. Font UIDs include jobId prefix, so we
+     * Clear job-specific entries from Type3 font caches. Font UIDs include jobId prefix, so we can
-     * can identify and remove them.
+     * identify and remove them.
      */
     private void clearType3CacheEntriesForJob(String jobId) {
         if (jobId == null || jobId.isEmpty()) {

app/proprietary/src/main/java/stirling/software/SPDF/service/PdfJsonFallbackFontService.java

@@ -36,6 +36,21 @@ public class PdfJsonFallbackFontService {
     public static final String FALLBACK_FONT_AR_ID = "fallback-noto-arabic";
     public static final String FALLBACK_FONT_TH_ID = "fallback-noto-thai";
 
+    // Font name aliases map PDF font names to available fallback fonts
+    // This provides better visual consistency when editing PDFs
+    private static final Map<String, String> FONT_NAME_ALIASES =
+            Map.ofEntries(
+                    // Liberation fonts are metric-compatible with Microsoft core fonts
+                    Map.entry("arial", "fallback-liberation-sans"),
+                    Map.entry("helvetica", "fallback-liberation-sans"),
+                    Map.entry("arimo", "fallback-liberation-sans"),
+                    Map.entry("times", "fallback-liberation-serif"),
+                    Map.entry("timesnewroman", "fallback-liberation-serif"),
+                    Map.entry("tinos", "fallback-liberation-serif"),
+                    Map.entry("courier", "fallback-liberation-mono"),
+                    Map.entry("couriernew", "fallback-liberation-mono"),
+                    Map.entry("cousine", "fallback-liberation-mono"));
+
     private static final Map<String, FallbackFontSpec> BUILT_IN_FALLBACK_FONTS =
             Map.ofEntries(
                     Map.entry(
@@ -65,6 +80,45 @@ public class PdfJsonFallbackFontService {
                             new FallbackFontSpec(
                                     "classpath:/static/fonts/NotoSansThai-Regular.ttf",
                                     "NotoSansThai-Regular",
+                                    "ttf")),
+                    // Liberation Sans family
+                    Map.entry(
+                            "fallback-liberation-sans",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationSans-Regular.ttf",
+                                    "LiberationSans-Regular",
+                                    "ttf")),
+                    Map.entry(
+                            "fallback-liberation-sans-bold",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationSans-Bold.ttf",
+                                    "LiberationSans-Bold",
+                                    "ttf")),
+                    Map.entry(
+                            "fallback-liberation-sans-italic",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationSans-Italic.ttf",
+                                    "LiberationSans-Italic",
+                                    "ttf")),
+                    Map.entry(
+                            "fallback-liberation-sans-bolditalic",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationSans-BoldItalic.ttf",
+                                    "LiberationSans-BoldItalic",
+                                    "ttf")),
+                    // Liberation Serif family
+                    Map.entry(
+                            "fallback-liberation-serif",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationSerif-Regular.ttf",
+                                    "LiberationSerif-Regular",
+                                    "ttf")),
+                    // Liberation Mono family
+                    Map.entry(
+                            "fallback-liberation-mono",
+                            new FallbackFontSpec(
+                                    "classpath:/static/pdfjs-legacy/standard_fonts/LiberationMono-Regular.ttf",
+                                    "LiberationMono-Regular",
                                     "ttf")));
 
     private final ResourceLoader resourceLoader;
@@ -107,7 +161,9 @@ public class PdfJsonFallbackFontService {
         }
         byte[] bytes = loadFallbackFontBytes(fallbackId, spec);
         try (InputStream stream = new ByteArrayInputStream(bytes)) {
-            return PDType0Font.load(document, stream, true);
+            // Load with embedSubset=false to ensure full glyph coverage
+            // Fallback fonts need all glyphs available for substituting missing characters
+            return PDType0Font.load(document, stream, false);
         }
     }
 
@@ -140,6 +196,53 @@ public class PdfJsonFallbackFontService {
         }
     }
 
+    /**
+     * Resolve fallback font ID based on the original font name and code point. Attempts to match
+     * font family for visual consistency.
+     *
+     * @param originalFontName the name of the original font (may be null)
+     * @param codePoint the Unicode code point that needs to be rendered
+     * @return fallback font ID
+     */
+    public String resolveFallbackFontId(String originalFontName, int codePoint) {
+        // First try to match based on original font name for visual consistency
+        if (originalFontName != null && !originalFontName.isEmpty()) {
+            // Normalize font name: remove subset prefix (e.g. "PXAAAC+"), convert to lowercase,
+            // remove spaces
+            String normalized =
+                    originalFontName
+                            .replaceAll("^[A-Z]{6}\\+", "") // Remove subset prefix
+                            .toLowerCase()
+                            .replaceAll("\\s+", ""); // Remove spaces (e.g. "Times New Roman" ->
+            // "timesnewroman")
+
+            // Extract base name without weight/style suffixes
+            // Split on common delimiters: hyphen, underscore, comma, plus
+            // Handles: "Arimo_700wght" -> "arimo", "Arial-Bold" -> "arial", "Arial,Bold" -> "arial"
+            String baseName = normalized.split("[-_,+]")[0];
+
+            String aliasedFontId = FONT_NAME_ALIASES.get(baseName);
+            if (aliasedFontId != null) {
+                log.debug(
+                        "Matched font '{}' (normalized: '{}', base: '{}') to fallback '{}'",
+                        originalFontName,
+                        normalized,
+                        baseName,
+                        aliasedFontId);
+                return aliasedFontId;
+            }
+        }
+
+        // Fall back to Unicode-based selection
+        return resolveFallbackFontId(codePoint);
+    }
+
+    /**
+     * Resolve fallback font ID based on Unicode code point properties.
+     *
+     * @param codePoint the Unicode code point
+     * @return fallback font ID
+     */
     public String resolveFallbackFontId(int codePoint) {
         Character.UnicodeBlock block = Character.UnicodeBlock.of(codePoint);
         if (block == Character.UnicodeBlock.CJK_UNIFIED_IDEOGRAPHS

docs/pdf-json-editor-backlog.md

@@ -13,11 +13,25 @@
   - Reuse the existing job cache (`documentCache`) to serve these on demand and clean up after timeouts (`PdfJsonConversionService.java:3608-3687`).
 
 - **Editor UX Safeguards**
-  - Respect `fallbackFontService` indicators; mark groups using fallback glyphs so the UI can warn about possible appearance shifts (`frontend/src/proprietary/components/tools/pdfJsonEditor/PdfJsonEditorView.tsx:1260-1287`).
+  - Mark groups using fallback glyphs so the UI can warn about possible appearance shifts. Font family matching is now implemented (Liberation fonts), but weight matching is still TODO, so bold/italic text using fallbacks may appear lighter than original.
   - Surface when Type3 conversion was downgraded (e.g., rasterized glyphs) and limit editing to operations that keep the PDF stable.
+  - Reference: `frontend/src/proprietary/components/tools/pdfJsonEditor/PdfJsonEditorView.tsx:1260-1287`
 
 - **Canonical Font Sharing**
   - Emit fonts once per unique embedded program. Add a `canonicalFonts` array containing the full payload (program, ToUnicode, metadata) and a compact `fontAliases` mapping `{pageNumber, fontId, canonicalUid}` so text elements can still reference per-page IDs.
-  - Store COS dictionaries only on canonical entries; aliases should keep light fields (e.g., size adjustments) if they differ.
+  - Note: COS dictionaries are currently preserved for TrueType/Type0 fonts (needed for ToUnicode CMap). The canonical approach should maintain this preservation while deduplicating font programs.
   - Update `buildFontMap` to resolve aliases when recreating PDFBox fonts, and adjust the front end to load programs via the canonical UID.
   - Optional: expose a lazy endpoint for the original COS dictionary if the canonical record strips it, so export still reconstructs untouched fonts.
+
+- **Font Weight Matching for Fallback Fonts**
+  - Font family matching is now implemented (Arial→LiberationSans, Times→LiberationSerif, Courier→LiberationMono).
+  - However, fallback fonts still use Regular weight for all missing glyphs, regardless of the original font weight (e.g., bold text falls back to regular weight).
+  - TODO: Parse weight from font names (e.g., `Arimo_700wght`, `Arial-Bold`, `TimesNewRoman,SemiBold`) and map to corresponding Liberation font variants:
+    - Regular/Normal → LiberationSans-Regular, LiberationSerif-Regular, LiberationMono-Regular
+    - Bold/700 → LiberationSans-Bold, LiberationSerif-Bold, LiberationMono-Bold
+    - Italic/Oblique → LiberationSans-Italic, LiberationSerif-Italic, LiberationMono-Italic
+    - BoldItalic → LiberationSans-BoldItalic, LiberationSerif-BoldItalic, LiberationMono-BoldItalic
+  - Add all Liberation font variants to `BUILT_IN_FALLBACK_FONTS` map with appropriate IDs (e.g., `fallback-liberation-sans-bold`).
+  - Update `resolveFallbackFontId(String originalFontName, int codePoint)` in `PdfJsonFallbackFontService.java` to detect weight/style and return the matching variant ID.
+  - Benefits: Better visual consistency when editing text in bold/italic fonts, as missing characters will match the original weight.
+  - Implementation reference: `app/proprietary/src/main/java/stirling/software/SPDF/service/PdfJsonFallbackFontService.java:186-213`