Attempt at writing tar headers directly

tar.py

@@ -1,26 +1,65 @@
 from flask import Response
-import io
+import functools
 import os
-import tarfile
 
-# 4MiB chuns
+# 4MiB chunks
 CHUNK_SIZE = 4_194_304
 
+# Specification de l'entête du fichier
+# Numéro Nom      Début   Taille   Description
+#      1 name         0      100   Nom du fichier
+#      2 mode       100        8   Permissions
+#      3 uid        108        8   Propriétaire (inutilisé si format étendu)
+#      4 gid        116        8   Groupe (inutilisé si format étendu)
+#      5 size       124       12   Taille du fichier en octets.
+#      6 mtime      136       12   Dernière modification en temps Unix.
+#      7 chksum     148        8   Somme de contrôle de l'en-tête où ce champ est considéré comme rempli d'espaces (32)
+#      8 type flag  156        1   Type de fichier
+#      9 linkname   157      100   Nom du fichier pointé par ce lien symbolique (Si le type indique un lien symbolique)
+
 
 def header_chunk(filename: str, filepath: str) -> bytes:
-    bytes = io.BytesIO()
     stat = os.stat(filepath)
+    buffer = bytearray()
 
-    # Create dummy tar to extract tar header for file
-    with tarfile.open(fileobj=bytes, mode='w') as buffer:
-        tar_info = tarfile.TarInfo(filepath)
-        tar_info.name = filename
-        tar_info.size = stat.st_size
-        buffer.addfile(tar_info)
+    # Field 1: filename on 100 bytes
+    buffer += filename.encode('ascii')
+    buffer += b'\x00' * (100 - len(filename))
 
-    # TODO if we were able to build this chunk without tarfile, it would avoid
-    # whole file copy in memory
-    return bytes.getvalue()[:512]
+    # Field 2: mode, on 8 bytes
+    # TODO we put 777 for test
+    buffer += '0000777'.encode('ascii') + b'\x00'
+
+    # Field 3: owner, we put 1000, default user
+    buffer += '0001000'.encode('ascii') + b'\x00'
+
+    # Field 4: group, we put 1000
+    buffer += '0001000'.encode('ascii') + b'\x00'
+
+    # Field 5: file size in bytes in ascii
+    buffer += str(stat.st_size).rjust(11, '0').encode('ascii') + b'\x00'
+
+    # Field 6: last modified, zeros for now
+    # buffer += '00000000000'.encode('ascii') + b'\x00'
+    buffer += '1721728914'.rjust(11, '0').encode('ascii') + b'\x00'
+
+    # Field 7: checksum, we put spaces and we will edit it at the end
+    buffer += '        '.encode('ascii')
+
+    # Field 8: type flag, 0 because we only have regular files
+    buffer += b'0'
+
+    # Field 9: linkname, \x00s because we only have regular files
+    buffer += b'\x00' * 100
+
+    # POSIX 1003.1-1990: 255 empty bytes
+    buffer += b'\x00' * 255
+
+    # Compute the checksum
+    checksum = str(functools.reduce(lambda x, y: x + y, buffer)).rjust(6, '0').encode('ascii') + b'\x00 '
+    buffer[148:156] = checksum
+
+    return bytes(buffer)
 
 
 class TarSender: