ocr improved

analyze_pdf_fonts.py

@@ -0,0 +1,75 @@
+import fitz  # PyMuPDF
+import sys
+
+def analyze_pdf_fonts(pdf_path):
+    doc = fitz.open(pdf_path)
+    page = doc[0]
+    
+    # Get page text with flags
+    text = page.get_text("dict")
+    print("Page structure keys:", text.keys())
+    
+    # Look for blocks
+    blocks = text.get("blocks", [])
+    print(f"Number of blocks: {len(blocks)}")
+    
+    for i, block in enumerate(blocks[:5]):  # first few blocks
+        print(f"\nBlock {i}: type={block.get('type')}")
+        if block.get('type') == 0:  # text block
+            lines = block.get('lines', [])
+            print(f"  Lines: {len(lines)}")
+            for line in lines[:2]:
+                spans = line.get('spans', [])
+                print(f"    Spans: {len(spans)}")
+                for span in spans[:2]:
+                    font = span.get('font', '')
+                    flags = span.get('flags', 0)
+                    color = span.get('color', 0)
+                    size = span.get('size', 0)
+                    text = span.get('text', '')
+                    print(f"      Font: {font[:30]}...")
+                    print(f"      Flags: {flags}")
+                    print(f"      Size: {size}")
+                    print(f"      Text sample: {repr(text[:50])}")
+    
+    # Try to extract text with different encodings
+    print("\n--- Text extraction methods ---")
+    # 1. Raw text
+    raw = page.get_text()
+    print(f"Raw text length: {len(raw)}")
+    print(f"First 200 chars: {repr(raw[:200])}")
+    
+    # 2. HTML output (includes font info)
+    html = page.get_text("html")
+    print(f"HTML length: {len(html)}")
+    # find font-family in html
+    import re
+    fonts = re.findall(r'font-family:([^;]+);', html)
+    if fonts:
+        print(f"Font families used: {set(fonts)}")
+    
+    # 3. Check if there's a ToUnicode map
+    xref = page.get_contents()[0]  # first content stream xref
+    print(f"Content stream xref: {xref}")
+    
+    # Get the page object
+    page_obj = doc._getPageObject(page.number)
+    print(f"Page object keys: {page_obj.keys()}")
+    
+    # Look for Resources -> Font
+    resources = page_obj.get('Resources', {})
+    fonts = resources.get('Font', {})
+    print(f"Number of fonts: {len(fonts)}")
+    for font_name, font_obj in fonts.items():
+        print(f"  Font {font_name}: {font_obj}")
+        # Check if font has ToUnicode
+        if '/ToUnicode' in font_obj:
+            print("    Has ToUnicode map")
+        else:
+            print("    No ToUnicode map")
+    
+    doc.close()
+
+if __name__ == "__main__":
+    pdf_path = "test/safedistance.pdf"
+    analyze_pdf_fonts(pdf_path)