set WITHOUT ROWID for the 2 binding tables

less db size especially when with lots of chunks, 82M -> 77M for f4c on 1.5GB data

Author: Fallen-Breath

prime_backup/db/migrations/migration_3_4.py

@@ -9,6 +9,7 @@
 
 from prime_backup.db.migrations import MigrationImplBase
 from prime_backup.db.values import BlobStorageMethod
+from prime_backup.utils import db_utils
 
 
 class _V4:
@@ -50,13 +51,15 @@ class _V4:
 		Column('chunk_group_id', Integer, ForeignKey('chunk_group.id'), primary_key=True),
 		Column('chunk_offset', BigInteger, primary_key=True),
 		Column('chunk_id', Integer, ForeignKey('chunk.id'), index=True, nullable=False),
+		**({'sqlite_with_rowid': False} if db_utils.check_sqlite_without_rowid() else {}),
 	)
 	BlobChunkGroupBinding = Table(
 		'blob_chunk_group_binding',
 		Base.metadata,
 		Column('blob_id', Integer, ForeignKey('blob.id'), primary_key=True),
 		Column('chunk_group_offset', BigInteger, primary_key=True),
 		Column('chunk_group_id', Integer, ForeignKey('chunk_group.id'), index=True, nullable=False),
+		**({'sqlite_with_rowid': False} if db_utils.check_sqlite_without_rowid() else {}),
 	)
 	File = Table(
 		'file',

prime_backup/db/schema.py

@@ -4,6 +4,7 @@
 from sqlalchemy.orm import DeclarativeBase, Mapped, mapped_column, relationship
 
 from prime_backup.db.values import BackupTagDict
+from prime_backup.utils import db_utils
 
 
 class Base(DeclarativeBase):
@@ -78,6 +79,7 @@ class ChunkGroupChunkBinding(Base):
 	"""
 
 	__tablename__ = 'chunk_group_chunk_binding'
+	__table_args__ = {'sqlite_with_rowid': False} if db_utils.check_sqlite_without_rowid() else {}
 
 	chunk_group_id: Mapped[int] = mapped_column(ForeignKey('chunk_group.id'), primary_key=True)
 	chunk_offset: Mapped[int] = mapped_column(BigInteger, primary_key=True)  # &chunk[0] - &chunk_group[0]
@@ -94,6 +96,7 @@ class BlobChunkGroupBinding(Base):
 	"""
 
 	__tablename__ = 'blob_chunk_group_binding'
+	__table_args__ = {'sqlite_with_rowid': False} if db_utils.check_sqlite_without_rowid() else {}
 
 	blob_id: Mapped[int] = mapped_column(ForeignKey('blob.id'), primary_key=True)
 	chunk_group_offset: Mapped[int] = mapped_column(BigInteger, primary_key=True)

prime_backup/utils/db_utils.py

@@ -28,6 +28,13 @@ def check_sqlite_row_number() -> bool:
 	return sqlite3.sqlite_version_info >= (3, 25, 0)
 
 
+def check_sqlite_without_rowid() -> bool:
+	"""
+	https://sqlite.org/withoutrowid.html#compatibility
+	"""
+	return sqlite3.sqlite_version_info >= (3, 8, 2)
+
+
 def vacuum_into_via_backup_api(src_db_path: PathLike, into_path: PathLike):
 	"""
 	Fallback for VACUUM INTO on old SQLite versions that do not support "VACUUM INTO"
@@ -48,3 +55,4 @@ def vacuum_into_via_backup_api(src_db_path: PathLike, into_path: PathLike):
 	print('json query:', check_sqlite_json_query_support())
 	print('vacuum into:', check_sqlite_vacuum_into_support())
 	print('row number:', check_sqlite_row_number())
+	print('without rowid:', check_sqlite_without_rowid())

tests/data/schema_ddl_v4.yml

@@ -35,6 +35,7 @@ tables:
     	FOREIGN KEY(blob_id) REFERENCES blob (id), 
     	FOREIGN KEY(chunk_group_id) REFERENCES chunk_group (id)
     )
+     WITHOUT ROWID
   chunk: |-
     CREATE TABLE chunk (
     	id INTEGER NOT NULL PRIMARY KEY AUTOINCREMENT, 
@@ -62,6 +63,7 @@ tables:
     	FOREIGN KEY(chunk_group_id) REFERENCES chunk_group (id), 
     	FOREIGN KEY(chunk_id) REFERENCES chunk (id)
     )
+     WITHOUT ROWID
   db_meta: |-
     CREATE TABLE db_meta (
     	magic INTEGER NOT NULL, 

tests/schema_utils.py

@@ -25,7 +25,7 @@ def collect_schema(engine: Engine) -> SchemaDDL:
 	with engine.connect() as conn:
 		def collect(sql: str) -> Dict[str, str]:
 			return {
-				row.name: row.sql
+				row.name: row.sql.strip()
 				for row in conn.execute(text(sql))
 				if not row.name.startswith(_SQLITE_INTERNAL_PREFIX)
 			}

tests/test_schema_ddl.py

@@ -5,11 +5,13 @@
 
 Review the diff of the updated YAML files and commit them as the new golden values
 """
+import re
 import unittest
 from pathlib import Path
 
 from ruamel.yaml import YAML
 
+from prime_backup.utils import db_utils
 from tests import schema_utils
 
 __DATA_DIR = Path(__file__).parent / 'data'
@@ -22,7 +24,9 @@ def __load_expected(filename: str) -> schema_utils.SchemaDDL:
 
 
 def _assert_schema_matches(tc: unittest.TestCase, actual: schema_utils.SchemaDDL, filename: str, exact_match: bool):
-	import re
+	if not db_utils.check_sqlite_without_rowid():
+		tc.fail(f'SQLite without rowid is not supported, version: {db_utils.get_sqlite_version()}')
+
 	frozen_answer = __load_expected(filename)
 	hint = f'Re-run tests/tools/dump_schema_ddl.py to update {filename}.'
 	if exact_match: