feat: implement Hamming 8/4 error correction and auto-alignment for T42 files

src/teletext/io.py

@@ -1,22 +1,46 @@
 import os
 from typing import List, Callable, Optional
-from .models import Packet, Page, TeletextService
+from .models import Packet, Page, TeletextService, decode_hamming_8_4
 
 def load_t42(file_path: str, progress_callback: Optional[Callable[[int, int], None]] = None) -> TeletextService:
     service = TeletextService()
     
     total_bytes = os.path.getsize(file_path)
+    
+    # Auto-detect alignment
+    # Scan first 4096 bytes for the offset that yields the most Row 0 headers
+    best_offset = 0
+    max_headers = 0
+    with open(file_path, 'rb') as f:
+        head_data = f.read(4096 + 42)
+        for offset in range(42):
+            headers = 0
+            for i in range(offset, len(head_data) - 42, 42):
+                chunk = head_data[i:i+42]
+                try:
+                    # We check if Mag/Row decode to something sensible
+                    # AND if it's Row 0, we check if Page Num decodes without errors (ideally)
+                    # But for now, just counting Row 0s is good enough.
+                    mag, row = decode_packet_header(chunk[0], chunk[1])
+                    if row == 0 and mag != 8: # Mag 8 is often just nulls
+                        headers += 1
+                except:
+                    pass
+            if headers > max_headers:
+                max_headers = headers
+                best_offset = offset
+    
     # Each packet is 42 bytes
-    total_packets = total_bytes // 42
+    total_packets = (total_bytes - best_offset) // 42
     processed_packets = 0
     
     with open(file_path, 'rb') as f:
+        f.seek(best_offset)
         while True:
             chunk = f.read(42)
             if not chunk:
                 break
             if len(chunk) < 42:
-                # Should not happen in a valid T42 stream, or we just ignore incomplete tail
                 break
             
             processed_packets += 1
@@ -27,9 +51,6 @@ def load_t42(file_path: str, progress_callback: Optional[Callable[[int, int], No
             service.all_packets.append(packet)
             
             # Logic to group into pages.
-            # This is non-trivial because packets for a page might be interleaved or sequential.
-            # Standard implementation: Packets arrive in order. Row 0 starts a new page/subpage.
-            
             if packet.row == 0:
                 # Start of a new page header.
                 # Byte 2-9 of header contain Page Number, Subcode, Control bits etc.
@@ -175,12 +196,6 @@ def save_t42(file_path: str, service: TeletextService, progress_callback: Option
                 
             f.write(header + packet.data)
 
-def decode_hamming_8_4(byte_val):
-    return ((byte_val >> 1) & 1) | \
-           (((byte_val >> 3) & 1) << 1) | \
-           (((byte_val >> 5) & 1) << 2) | \
-           (((byte_val >> 7) & 1) << 3)
-
 def parse_header(data: bytearray):
     # Data is 40 bytes.
     # Bytes 0-7 are Page Num (2), Subcode (4), Control (2) - ALL Hamming encoded.