[performance] Reduce number of files in TAR to verify

Author: mxmlnkn

core/ratarmountcore/SQLiteIndex.py

@@ -147,7 +147,10 @@ class SQLiteIndex:
     __version__ = '0.7.0'
 
     MAGIC_BYTES = b'SQLite format 3\x00'  # If it is encrypted, the first 16 B are the (random) salt.
-    NUMBER_OF_METADATA_TO_VERIFY = 1000  # shouldn't take more than 1 second according to benchmarks
+    # 1000 rows to verify shouldn't take more than 1 second on HDDs according to benchmarks.
+    # For remote/network/global filesystems and/or large block sizes, it can snowball, therefore it was reduced
+    # by more than one magnitude to 100.
+    NUMBER_OF_METADATA_TO_VERIFY = 100
 
     # The maximum blob size configured by SQLite is exactly 1 GB, see https://www.sqlite.org/limits.html
     # Therefore, this should be smaller. Another argument for making it smaller is that this blob size

core/ratarmountcore/mountsource/formats/tar.py

@@ -1336,13 +1336,14 @@ def _check_metadata2(self, metadata: dict[str, Any]) -> None:
     def _check_index_validity(self) -> bool:
         # Check some of the first and last files in the archive and some random selection in between.
         selectFiles = "SELECT * " + SQLiteIndex.FROM_REGULAR_FILES
+        limit = int(max(1, SQLiteIndex.NUMBER_OF_METADATA_TO_VERIFY // 3))
         result = self.index.get_connection().execute(
             f"""
-            SELECT * FROM ( {selectFiles} ORDER BY offset ASC LIMIT 100 )
+            SELECT * FROM ( {selectFiles} ORDER BY offset ASC LIMIT {limit} )
             UNION
-            SELECT * FROM ( {selectFiles} ORDER BY RANDOM() LIMIT {SQLiteIndex.NUMBER_OF_METADATA_TO_VERIFY} )
+            SELECT * FROM ( {selectFiles} ORDER BY RANDOM() LIMIT {limit} )
             UNION
-            SELECT * FROM ( {selectFiles} ORDER BY offset DESC LIMIT 100 )
+            SELECT * FROM ( {selectFiles} ORDER BY offset DESC LIMIT {limit} )
             ORDER BY offset
         """
         )