[python] Remove default deletion vectors and dedicated vector files for multimodal tables

docs/docs/pypaimon/multimodal-tables.mdx

@@ -35,12 +35,14 @@ primary keys. Each `add`, `overwrite`, `update`, or `delete` call handles its ow
 commit; use the [Python API](./python-api) for explicit writer and commit control.
 
 Creating a multimodal table enables `row-tracking.enabled`,
-`data-evolution.enabled`, `deletion-vectors.enabled`, and
-`blob-as-descriptor` by default. Global index scans use full fallback coverage
+`data-evolution.enabled`, and `blob-as-descriptor` by default.
+Global index scans use full fallback coverage
 through `global-index.search-mode` set to `FULL`. Data files use Paimon's
-default Parquet format, and vector columns are stored in dedicated Parquet
-files through `vector.file.format` set to `parquet`. These defaults require no
-file-format extra:
+default Parquet format, including vector columns. To store vectors in dedicated
+files, explicitly set `vector.file.format`, for example to `parquet`.
+Deletion vectors are disabled by default; enable them with
+`deletion-vectors.enabled` set to `true` in the table creation options.
+These defaults require no file-format extra:
 
 ```shell
 pip install pypaimon
@@ -365,9 +367,10 @@ or delete prepared files after a commit error.
 ## Delete
 
 `delete` removes rows matched by a SQL-like predicate. It uses the same
-predicate syntax as `scan().where(...)` and `update(...)`. Multimodal tables
-enable deletion vectors by default; on partitioned tables, a predicate that
-references only partition columns uses the partition drop path.
+predicate syntax as `scan().where(...)` and `update(...)`. Row-level deletes
+require `deletion-vectors.enabled` set to `true` when creating the table.
+On partitioned tables, a predicate that references only partition columns uses
+the partition drop path.
 
 ```python
 docs.delete(where="id = 2")
@@ -405,7 +408,8 @@ docs.merge("id") \
     ])
 ```
 
-Use `when_matched_delete()` for matched rows that should be removed:
+Use `when_matched_delete()` for matched rows that should be removed. The table
+must have `deletion-vectors.enabled` set to `true`:
 
 ```python
 docs.merge("id") \

paimon-python/pypaimon/multimodal/connection.py

@@ -31,10 +31,8 @@
 _DEFAULT_OPTIONS = {
     "row-tracking.enabled": "true",
     "data-evolution.enabled": "true",
-    "deletion-vectors.enabled": "true",
     "blob-as-descriptor": "true",
     "global-index.search-mode": "full",
-    "vector.file.format": "parquet",
 }
 
 _DEFAULT_DATABASE = Catalog.DEFAULT_DATABASE

paimon-python/pypaimon/tests/batch_vector_lookup_test.py

@@ -36,6 +36,7 @@ def docs(tmp_path, request):
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "docs", schema=schema, partitioned=["category"], options={
             "file.format": "parquet", "vector.file.format": "parquet",
+            "deletion-vectors.enabled": "true",
             "vector-index.search-mode": "full", "read.parallelism": str(request.param),
             "read.batch-size": "2", "blob-as-descriptor": "false",
         })

paimon-python/pypaimon/tests/full_text_scalar_filter_test.py

@@ -33,6 +33,7 @@ def docs(tmp_path):
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "docs", schema=schema, partitioned=["pt"], options={
             "file.format": "parquet", "vector.file.format": "parquet", "read.batch-size": "1",
+            "deletion-vectors.enabled": "true",
             "full-text-index.search-mode": "full", "vector-index.search-mode": "full"})
     table.add(pa.table({"id": [0, 1, 2], "text": ["paimon", "paimon paimon", "paimon long document"],
                         "label": ["other", "other", "target"], "pt": ["a"] * 3,

paimon-python/pypaimon/tests/global_index_live_row_filter_test.py

@@ -27,7 +27,8 @@
 
 def test_scoped_live_rows_only_load_relevant_deletions(tmp_path):
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
-        "rows", schema=pa.schema([("id", pa.int64())]), options={"file.format": "parquet"})
+        "rows", schema=pa.schema([("id", pa.int64())]), options={
+            "file.format": "parquet", "deletion-vectors.enabled": "true"})
     table.add(pa.table({"id": [0, 1, 2]}))
     table.add(pa.table({"id": [3, 4, 5]}))
     snapshot = table.raw_table.snapshot_manager().get_latest_snapshot()

paimon-python/pypaimon/tests/multimodal_columns_test.py

@@ -36,7 +36,8 @@ def frames(tmp_path):
         ]),
         data=[{"id": i, "label": "old", "image": bytes([i]) * 100}
               for i in range(5)],
-        options={"file.format": "parquet", "read.batch-size": "2"},
+        options={"file.format": "parquet", "read.batch-size": "2",
+                 "deletion-vectors.enabled": "true"},
     )
 
 

paimon-python/pypaimon/tests/multimodal_lance_test.py

@@ -92,7 +92,7 @@ def test_export_lance_exports_logical_updates_deletes_and_historical_schema(conn
     }, [
         {"id": 1, "text": "before", "image": b"before"},
         {"id": 2, "text": "deleted", "image": b"deleted"},
-    ])
+    ], options={"deletion-vectors.enabled": "true"})
     snapshot_id = table.raw_table.snapshot_manager().get_latest_snapshot().id
     table.raw_table.create_tag("training-v1", snapshot_id=snapshot_id)
     conn.catalog.alter_table("default.history", [

paimon-python/pypaimon/tests/multimodal_table_test.py

@@ -137,11 +137,11 @@ def test_create_table_defaults_data_evolution_options(self):
         options = table.raw_table.table_schema.options
         self.assertEqual("true", options["row-tracking.enabled"])
         self.assertEqual("true", options["data-evolution.enabled"])
-        self.assertEqual("true", options["deletion-vectors.enabled"])
+        self.assertNotIn("deletion-vectors.enabled", options)
         self.assertEqual("true", options["blob-as-descriptor"])
         self.assertNotIn("data-evolution.row-sidecar.enabled", options)
         self.assertEqual("full", options["global-index.search-mode"])
-        self.assertEqual("parquet", options["vector.file.format"])
+        self.assertNotIn("vector.file.format", options)
         self.assertEqual("default.docs", self.conn.get_table("docs").identifier)
         self.assertEqual(["id", "content", "embedding", "payload"],
                          [field.name for field in table.raw_table.fields])

paimon-python/pypaimon/tests/ray_vector_filter_exactness_test.py

@@ -108,6 +108,7 @@ def test_worker_filters_overlapping_scalar_shards_at_planned_snapshot(tmp_path,
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "vectors", schema=schema, options={
             "file.format": "parquet", "vector.file.format": "parquet",
+            "deletion-vectors.enabled": "true",
             "global-index.filter.refine-from-data": "true"})
     for start in range(0, 12, 3):
         row_ids = list(range(start, start + 3))

paimon-python/pypaimon/tests/ray_vector_search_test.py

@@ -57,6 +57,7 @@ def table(tmp_path):
     return pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "vectors", schema=SCHEMA,
         options={"file.format": "parquet", "vector.file.format": "parquet",
+                 "deletion-vectors.enabled": "true",
                  "vector-index.search-mode": "full", "read.batch-size": "2",
                  "source.split.target-size": "1 b", "source.split.open-file-cost": "1 b"})
 

paimon-python/pypaimon/tests/search_result_metadata_test.py

@@ -30,7 +30,8 @@ def docs(tmp_path):
                         ("embedding", pa.list_(pa.float32(), 2))])
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "docs", schema=schema, options={"file.format": "parquet", "vector.file.format": "parquet",
-                                        "vector-index.search-mode": "full", "read.batch-size": "1"})
+                                        "vector-index.search-mode": "full", "read.batch-size": "1",
+                                        "deletion-vectors.enabled": "true"})
     table.add(pa.table({"id": [0, 1, 2, 3], "text": ["paimon long document", "paimon", "other", "paimon"],
                         "embedding": [[2., 1.], [0., 1.], [1., 1.], [0., 1.]]}, schema=schema))
     return table

paimon-python/pypaimon/tests/search_snapshot_test.py

@@ -40,7 +40,7 @@ def table(tmp_path):
     return pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "vectors", schema=SCHEMA,
         options={"file.format": "parquet", "vector.file.format": "parquet",
-                 "vector-index.search-mode": "full"})
+                 "vector-index.search-mode": "full", "deletion-vectors.enabled": "true"})
 
 
 def search(table, kind="vector", **options):

paimon-python/pypaimon/tests/torch_read_test.py

@@ -2045,7 +2045,9 @@ def test_torch_batch_shuffle_vectors_deletions_and_snapshot(self):
 
         schema = pa.schema([('id', pa.int64()), ('vector', pa.list_(pa.float32(), 2))])
         table = pm.connect(options={'warehouse': self.warehouse}).create_table(
-            'batch_shuffle_vectors', schema=schema, options={'file.format': 'parquet'})
+            'batch_shuffle_vectors', schema=schema, options={
+                'file.format': 'parquet', 'vector.file.format': 'parquet',
+                'deletion-vectors.enabled': 'true'})
         table.add(pa.table({'id': list(range(8)), 'vector': [[i, i + 1] for i in range(8)]}, schema=schema))
         snapshot = table.raw_table.snapshot_manager().get_latest_snapshot().id
         table.delete('id = 2')

paimon-python/pypaimon/tests/vector_filter_exactness_test.py

@@ -42,7 +42,7 @@ def table(tmp_path):
     schema = pa.schema([("id", pa.int64()), ("name", pa.string()), ("embedding", pa.list_(pa.float32(), 2))])
     table = pm.connect(options={"warehouse": str(tmp_path)}).create_table(
         "vectors", schema=schema, options={"file.format": "parquet", "vector.file.format": "parquet",
-                                           "read.batch-size": "1"})
+                                           "read.batch-size": "1", "deletion-vectors.enabled": "true"})
     table.add(pa.table({"id": [0, 1, 2], "name": ["alpha", "beta zeta", "gamma"],
                         "embedding": [[0., 1.], [1., 1.], [2., 1.]]}, schema=schema))
     table.raw_table.create_global_index(

paimon-python/pypaimon/tests/vector_index_dv_build_test.py

@@ -46,6 +46,7 @@ def create_table(tmp_path, null_vector=False):
     table = pm.connect(options={'warehouse': str(tmp_path)}).create_table(
         'vectors', schema=schema, options={
             'file.format': 'parquet', 'vector.file.format': 'parquet', 'read.batch-size': '2',
+            'deletion-vectors.enabled': 'true',
             'global-index.row-count-per-shard': '4', 'global-index.build.parallelism': '1'})
     append(table, list(range(7)), null_vector)
     return table
@@ -123,7 +124,8 @@ def test_python_ivfpq_search_keeps_live_null_vectors_out_of_results(tmp_path, ba
     schema = pa.schema([('id', pa.int64()), ('embedding', pa.list_(pa.float32()))])
     table = pm.connect(options={'warehouse': str(tmp_path)}).create_table(
         'vectors', schema=schema, options={
-            'file.format': 'parquet', 'vector.file.format': 'parquet', 'vector-index.search-mode': 'fast'})
+            'file.format': 'parquet', 'vector.file.format': 'parquet', 'vector-index.search-mode': 'fast',
+            'deletion-vectors.enabled': 'true'})
     table.add(pa.table({
         'id': list(range(5)), 'embedding': [[1., 0.], [0., 1.], None, [1., 1.], [2., 2.]]}, schema=schema))
     table.delete('id = 0 OR id = 3')
@@ -178,7 +180,9 @@ def test_native_incremental_build_with_concurrent_delete(table):
 def test_native_partition_scoped_build_after_deletion(tmp_path):
     pytest.importorskip('paimon_vindex')
     table = pm.connect(options={'warehouse': str(tmp_path)}).create_table(
-        'vectors', schema=SCHEMA, partitioned=['category'], options={'file.format': 'parquet'})
+        'vectors', schema=SCHEMA, partitioned=['category'], options={
+            'file.format': 'parquet', 'vector.file.format': 'parquet',
+            'deletion-vectors.enabled': 'true'})
     append(table, [0, 1, 2], category='other')
     append(table, [3, 4, 5])
     table.delete("category = 'yes' AND id = 4")