Optimize _calc_initial_entry_offset by introducing cache

Author: danny0838

Lib/test/test_zipfile/test_core.py

@@ -1787,6 +1787,8 @@ def test_repack_file_entry_before_first_file(self):
                 with open(TESTFN, 'wb') as fh:
                     with zipfile.ZipFile(fh, 'w') as zh:
                         zh.writestr('file.txt', b'dummy')
+                        zh.writestr('file2.txt', b'dummy')
+                        zh.writestr('file3.txt', b'dummy')
                     fh.write(b' ')
                     expected_zinfos = self._prepare_zip_from_test_files(fh, test_files)
                 expected_size = os.path.getsize(TESTFN)
@@ -1795,6 +1797,8 @@ def test_repack_file_entry_before_first_file(self):
                 with open(TESTFN, 'wb') as fh:
                     with zipfile.ZipFile(fh, 'w') as zh:
                         zh.writestr('file.txt', b'dummy')
+                        zh.writestr('file2.txt', b'dummy')
+                        zh.writestr('file3.txt', b'dummy')
                     fh.write(b' ')
                     zinfos = self._prepare_zip_from_test_files(fh, self.test_files)
                 with zipfile.ZipFile(TESTFN, 'a', self.compression) as zh:

Lib/zipfile/__init__.py

@@ -1495,11 +1495,12 @@ def _repack(self, zfile, *, chunk_size=2**20):
         zfile._didModify = True
 
     def _calc_initial_entry_offset(self, fp, data_offset):
+        checked_offsets = set()
         if data_offset > 0:
             if self.debug > 2:
                 print('scanning file signatures before:', data_offset)
             for pos in self._iter_scan_signature(fp, stringFileHeader, 0, data_offset):
-                if self._starts_consecutive_file_entries(fp, pos, data_offset):
+                if self._starts_consecutive_file_entries(fp, pos, data_offset, checked_offsets):
                     return data_offset - pos
         return 0
 
@@ -1528,13 +1529,23 @@ def _iter_scan_signature(self, fp, signature, start_offset, end_offset, chunk_si
             remainder = chunk[-(sig_len - 1):]
             pos += read_size
 
-    def _starts_consecutive_file_entries(self, fp, start_offset, end_offset):
+    def _starts_consecutive_file_entries(self, fp, start_offset, end_offset, checked_offsets):
         offset = start_offset
 
         while offset < end_offset:
             if self.debug > 2:
                 print('checking local file entry:', offset)
 
+            # Cache checked offsets to improve performance by failing
+            # subsequent (possible) file entry offsets early. They are
+            # rechecked only when proven false eventually.
+            if offset in checked_offsets:
+                if self.debug > 2:
+                    print('skipping checked:', offset)
+                return False
+            else:
+                checked_offsets.add(offset)
+
             fp.seek(offset)
             try:
                 fheader = self._read_local_file_header(fp)