datachain 0.34.1__tar.gz → 0.34.3__tar.gz

{datachain-0.34.1 → datachain-0.34.3}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: datachain
-Version: 0.34.1
+Version: 0.34.3
 Summary: Wrangle unstructured AI data at scale
 Author-email: Dmitry Petrov <support@dvc.org>
 License-Expression: Apache-2.0

{datachain-0.34.1 → datachain-0.34.3}/src/datachain/catalog/loader.py

@@ -3,6 +3,7 @@ import sys
 from importlib import import_module
 from typing import TYPE_CHECKING, Any, Optional
 
+from datachain.plugins import ensure_plugins_loaded
 from datachain.utils import get_envs_by_prefix
 
 if TYPE_CHECKING:
@@ -24,6 +25,8 @@ IN_MEMORY_ERROR_MESSAGE = "In-memory is only supported on SQLite"
 
 
 def get_metastore(in_memory: bool = False) -> "AbstractMetastore":
+    ensure_plugins_loaded()
+
     from datachain.data_storage import AbstractMetastore
     from datachain.data_storage.serializer import deserialize
 
@@ -64,6 +67,8 @@ def get_metastore(in_memory: bool = False) -> "AbstractMetastore":
 
 
 def get_warehouse(in_memory: bool = False) -> "AbstractWarehouse":
+    ensure_plugins_loaded()
+
     from datachain.data_storage import AbstractWarehouse
     from datachain.data_storage.serializer import deserialize
 

datachain-0.34.3/src/datachain/data_storage/serializer.py

@@ -0,0 +1,119 @@
+import base64
+import json
+from abc import abstractmethod
+from collections.abc import Callable
+from typing import Any, ClassVar
+
+from datachain.plugins import ensure_plugins_loaded
+
+
+class CallableRegistry:
+    _registry: ClassVar[dict[str, Callable]] = {}
+
+    @classmethod
+    def register(cls, callable_obj: Callable, name: str) -> str:
+        cls._registry[name] = callable_obj
+        return name
+
+    @classmethod
+    def get(cls, name: str) -> Callable:
+        return cls._registry[name]
+
+
+class Serializable:
+    @classmethod
+    @abstractmethod
+    def serialize_callable_name(cls) -> str:
+        """Return the registered name used for this class' factory callable."""
+
+    @abstractmethod
+    def clone_params(self) -> tuple[Callable[..., Any], list[Any], dict[str, Any]]:
+        """Return (callable, args, kwargs) necessary to recreate this object."""
+
+    def _prepare(self, params: tuple) -> dict:
+        callable, args, kwargs = params
+        callable_name = callable.__self__.serialize_callable_name()
+        return {
+            "callable": callable_name,
+            "args": args,
+            "kwargs": {
+                k: self._prepare(v) if isinstance(v, tuple) else v
+                for k, v in kwargs.items()
+            },
+        }
+
+    def serialize(self) -> str:
+        """Return a base64-encoded JSON string with registered callable + params."""
+        _ensure_default_callables_registered()
+        data = self.clone_params()
+        return base64.b64encode(json.dumps(self._prepare(data)).encode()).decode()
+
+
+def deserialize(s: str) -> Serializable:
+    """Deserialize from base64-encoded JSON using only registered callables.
+
+    Nested serialized objects are instantiated automatically except for those
+    passed via clone parameter tuples (keys ending with ``_clone_params``),
+    which must remain as (callable, args, kwargs) for later factory usage.
+    """
+    ensure_plugins_loaded()
+    _ensure_default_callables_registered()
+    decoded = base64.b64decode(s.encode())
+    data = json.loads(decoded.decode())
+
+    def _is_serialized(obj: Any) -> bool:
+        return isinstance(obj, dict) and {"callable", "args", "kwargs"}.issubset(
+            obj.keys()
+        )
+
+    def _reconstruct(obj: Any, nested: bool = False) -> Any:
+        if not _is_serialized(obj):
+            return obj
+        callable_name: str = obj["callable"]
+        args: list[Any] = obj["args"]
+        kwargs: dict[str, Any] = obj["kwargs"]
+        # Recurse only inside kwargs because serialize() only nests through kwargs
+        for k, v in list(kwargs.items()):
+            if _is_serialized(v):
+                kwargs[k] = _reconstruct(v, True)
+        callable_obj = CallableRegistry.get(callable_name)
+        if nested:
+            return (callable_obj, args, kwargs)
+        # Otherwise instantiate
+        return callable_obj(*args, **kwargs)
+
+    if not _is_serialized(data):
+        raise ValueError("Invalid serialized data format")
+    return _reconstruct(data, False)
+
+
+class _DefaultsState:
+    registered = False
+
+
+def _ensure_default_callables_registered() -> None:
+    if _DefaultsState.registered:
+        return
+
+    from datachain.data_storage.sqlite import (
+        SQLiteDatabaseEngine,
+        SQLiteMetastore,
+        SQLiteWarehouse,
+    )
+
+    # Register (idempotent by name overwrite is fine) using class-level
+    # serialization names to avoid hard-coded literals here.
+    CallableRegistry.register(
+        SQLiteDatabaseEngine.from_db_file,
+        SQLiteDatabaseEngine.serialize_callable_name(),
+    )
+    CallableRegistry.register(
+        SQLiteMetastore.init_after_clone,
+        SQLiteMetastore.serialize_callable_name(),
+    )
+    CallableRegistry.register(
+        SQLiteWarehouse.init_after_clone,
+        SQLiteWarehouse.serialize_callable_name(),
+    )
+
+    _DefaultsState.registered = True

{datachain-0.34.1 → datachain-0.34.3}/src/datachain/data_storage/sqlite.py

@@ -201,10 +201,14 @@ class SQLiteDatabaseEngine(DatabaseEngine):
         """
         return (
             SQLiteDatabaseEngine.from_db_file,
-            [self.db_file],
+            [str(self.db_file)],
             {},
         )
 
+    @classmethod
+    def serialize_callable_name(cls) -> str:
+        return "sqlite.from_db_file"
+
     def _reconnect(self) -> None:
         if not self.is_closed:
             raise RuntimeError("Cannot reconnect on still-open DB!")
@@ -403,6 +407,10 @@ class SQLiteMetastore(AbstractDBMetastore):
             },
         )
 
+    @classmethod
+    def serialize_callable_name(cls) -> str:
+        return "sqlite.metastore.init_after_clone"
+
     @classmethod
     def init_after_clone(
         cls,
@@ -610,6 +618,10 @@ class SQLiteWarehouse(AbstractWarehouse):
             {"db_clone_params": self.db.clone_params()},
         )
 
+    @classmethod
+    def serialize_callable_name(cls) -> str:
+        return "sqlite.warehouse.init_after_clone"
+
     @classmethod
     def init_after_clone(
         cls,
@@ -863,8 +875,17 @@ class SQLiteWarehouse(AbstractWarehouse):
                 if isinstance(c, BinaryExpression):
                     right_left_join = add_left_rows_filter(c)
 
-        union = sqlalchemy.union(left_right_join, right_left_join).subquery()
-        return sqlalchemy.select(*union.c).select_from(union)
+        # Use CTE instead of subquery to force SQLite to materialize the result
+        # This breaks deep nesting and prevents parser stack overflow.
+        union_cte = sqlalchemy.union(left_right_join, right_left_join).cte()
+
+        return self._regenerate_system_columns(union_cte)
+
+    def _system_row_number_expr(self):
+        return func.row_number().over()
+
+    def _system_random_expr(self):
+        return self._system_row_number_expr() * 1103515245 + 12345
 
     def create_pre_udf_table(self, query: "Select") -> "Table":
         """

{datachain-0.34.1 → datachain-0.34.3}/src/datachain/data_storage/warehouse.py

@@ -246,6 +246,44 @@ class AbstractWarehouse(ABC, Serializable):
                     break  # no more results
                 offset += page_size
 
+    def _regenerate_system_columns(self, selectable):
+        """Return a SELECT that regenerates sys__id and sys__rand deterministically."""
+
+        base = selectable.subquery() if hasattr(selectable, "subquery") else selectable
+
+        system_types: dict[str, sa.types.TypeEngine] = {
+            sys_col.name: sys_col.type
+            for sys_col in self.schema.dataset_row_cls.sys_columns()
+        }
+
+        result_columns = []
+        for col in base.c:
+            if col.name == "sys__id":
+                expr = self._system_row_number_expr()
+                expr = sa.cast(expr, system_types["sys__id"])
+                result_columns.append(expr.label("sys__id"))
+            elif col.name == "sys__rand":
+                expr = self._system_random_expr()
+                expr = sa.cast(expr, system_types["sys__rand"])
+                result_columns.append(expr.label("sys__rand"))
+            else:
+                result_columns.append(col)
+
+        # Wrap in subquery to materialize window functions, then wrap again in SELECT
+        # This ensures window functions are computed before INSERT...FROM SELECT
+        inner = sa.select(*result_columns).select_from(base).subquery()
+        return sa.select(*inner.c).select_from(inner)
+
+    def _system_row_number_expr(self):
+        """Return an expression that produces deterministic row numbers."""
+
+        raise NotImplementedError
+
+    def _system_random_expr(self):
+        """Return an expression that produces deterministic random values."""
+
+        raise NotImplementedError
+
     #
     # Table Name Internal Functions
     #
@@ -923,6 +961,8 @@ class AbstractWarehouse(ABC, Serializable):
         right: "_FromClauseArgument",
         onclause: "_OnClauseArgument",
         inner: bool = True,
+        full: bool = False,
+        columns=None,
     ) -> sa.Select:
         """
         Join two tables together.

{datachain-0.34.1 → datachain-0.34.3}/src/datachain/lib/dc/datachain.py

@@ -1701,7 +1701,11 @@ class DataChain:
             )
 
         query = self._query.join(
-            right_ds._query, sqlalchemy.and_(*ops), inner, full, rname + "{name}"
+            right_ds._query,
+            sqlalchemy.and_(*ops),
+            inner,
+            full,
+            rname + "{name}",
         )
         query.feature_schema = None
         ds = self._evolve(query=query)
@@ -1989,7 +1993,8 @@ class DataChain:
         results = self.results(include_hidden=include_hidden)
         if as_object:
             df = pd.DataFrame(results, columns=columns, dtype=object)
-            return df.where(pd.notna(df), None)
+            df.where(pd.notna(df), None, inplace=True)
+            return df
         return pd.DataFrame.from_records(results, columns=columns)
 
     def show(

datachain-0.34.3/src/datachain/plugins.py

@@ -0,0 +1,30 @@
+"""Plugin loader for DataChain callables.
+
+Discovers and invokes entry points in the group "datachain.callables" once
+per process. This enables external packages (e.g., Studio) to register
+their callables with the serializer registry without explicit imports.
+"""
+
+from importlib import metadata as importlib_metadata
+
+_plugins_loaded = False
+
+
+def ensure_plugins_loaded() -> None:
+    global _plugins_loaded  # noqa: PLW0603
+    if _plugins_loaded:
+        return
+
+    # Compatible across importlib.metadata versions
+    eps_obj = importlib_metadata.entry_points()
+    if hasattr(eps_obj, "select"):
+        eps_list = eps_obj.select(group="datachain.callables")
+    else:
+        # Compatibility for older versions of importlib_metadata, Python 3.9
+        eps_list = eps_obj.get("datachain.callables", [])  # type: ignore[attr-defined]
+
+    for ep in eps_list:
+        func = ep.load()
+        func()
+
+    _plugins_loaded = True

{datachain-0.34.1 → datachain-0.34.3}/src/datachain.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: datachain
-Version: 0.34.1
+Version: 0.34.3
 Summary: Wrangle unstructured AI data at scale
 Author-email: Dmitry Petrov <support@dvc.org>
 License-Expression: Apache-2.0

{datachain-0.34.1 → datachain-0.34.3}/src/datachain.egg-info/SOURCES.txt

@@ -118,6 +118,7 @@ src/datachain/namespace.py
 src/datachain/node.py
 src/datachain/nodes_fetcher.py
 src/datachain/nodes_thread_pool.py
+src/datachain/plugins.py
 src/datachain/progress.py
 src/datachain/project.py
 src/datachain/py.typed

{datachain-0.34.1 → datachain-0.34.3}/tests/conftest.py

@@ -126,6 +126,7 @@ def clean_environment(
     working_dir = str(tmp_path_factory.mktemp("default_working_dir"))
     monkeypatch_session.chdir(working_dir)
     monkeypatch_session.delenv(DataChainDir.ENV_VAR, raising=False)
+    monkeypatch_session.delenv(DataChainDir.ENV_VAR_DATACHAIN_ROOT, raising=False)
 
 
 @pytest.fixture

{datachain-0.34.1 → datachain-0.34.3}/tests/func/test_datachain_merge.py

@@ -1,6 +1,7 @@
 import pytest
 
 import datachain as dc
+from datachain import File
 from datachain.sql.types import Int
 
 
@@ -99,3 +100,46 @@ def test_merge_multiple(cloud_test_catalog, inner1, inner2, inner3):
             ("dogs/dog3", 1, signal_default_value),
             ("dogs/others/dog4", 1, signal_default_value),
         ]
+
+
+def test_full_outer_join_preserves_all_rows(test_session):
+    """Test that full outer join correctly saves all rows including right-only rows.
+    This test verifies the fix for the NULL sys__id bug where right-only rows
+    from a full outer join had NULL sys__id values
+    """
+    # Create two datasets with no overlapping file paths
+    ds1 = dc.read_values(
+        id=[1, 2, 3],
+        file=[File(path=str(i)) for i in [1, 2, 3]],
+        session=test_session,
+    ).persist()
+
+    ds2 = dc.read_values(
+        id=[5, 6, 7],
+        file=[File(path=str(i)) for i in [5, 6, 7]],
+        session=test_session,
+    ).persist()
+
+    merged = ds1.merge(ds2, on="file.path", full=True)
+
+    # Use internal method to get all records including sys columns
+    records = merged._query.to_db_records()
+
+    assert len(records) == 6
+
+    # Extract sys__id and sys__rand from records
+    sys_ids = [r["sys__id"] for r in records]
+    sys_rands = [r["sys__rand"] for r in records]
+
+    # All sys__id values should be non-NULL and unique
+    assert all(sid is not None for sid in sys_ids)
+    assert len(set(sys_ids)) == 6
+    assert all(rand is not None for rand in sys_rands)
+
+    count_before = merged.count()
+
+    # Save and verify all rows are persisted
+    merged.save("test_merge")
+    count_after = dc.read_dataset("test_merge", session=test_session).count()
+
+    assert count_before == count_after == 6

{datachain-0.34.1 → datachain-0.34.3}/tests/unit/test_database_engine.py

@@ -1,12 +1,15 @@
 import base64
+import json
 import os
-import pickle
 
 import pytest
 from sqlalchemy import Column, Integer, Table
 
 from datachain.data_storage.serializer import deserialize
-from datachain.data_storage.sqlite import SQLiteDatabaseEngine, get_db_file_in_memory
+from datachain.data_storage.sqlite import (
+    SQLiteDatabaseEngine,
+    get_db_file_in_memory,
+)
 from tests.utils import skip_if_not_sqlite
 
 
@@ -24,6 +27,7 @@ def test_init_clone(tmp_dir, db_file, expected_db_file):
         expected_db_file = os.fspath(tmp_dir / expected_db_file)
 
     with SQLiteDatabaseEngine.from_db_file(db_file) as db:
+        assert isinstance(db, SQLiteDatabaseEngine)
         assert db.db_file == expected_db_file
 
         # Test clone
@@ -53,17 +57,15 @@ def test_get_db_file_in_memory(db_file, in_memory, expected):
 
 
 def test_serialize(sqlite_db):
-    # Test serialization
+    # JSON serialization format
     serialized = sqlite_db.serialize()
     assert serialized
-    serialized_pickled = base64.b64decode(serialized.encode())
-    assert serialized_pickled
-    (f, args, kwargs) = pickle.loads(serialized_pickled)  # noqa: S301
-    assert str(f) == str(SQLiteDatabaseEngine.from_db_file)
-    assert args == [":memory:"]
-    assert kwargs == {}
-
-    # Test deserialization
+    raw = base64.b64decode(serialized.encode())
+    data = json.loads(raw.decode())
+    assert data["callable"] == "sqlite.from_db_file"
+    assert data["args"] == [":memory:"]
+    assert data["kwargs"] == {}
+
     obj3 = deserialize(serialized)
     assert isinstance(obj3, SQLiteDatabaseEngine)
     assert obj3.db_file == ":memory:"

{datachain-0.34.1 → datachain-0.34.3}/tests/unit/test_metastore.py

@@ -1,5 +1,5 @@
 import base64
-import pickle
+import json
 
 import pytest
 
@@ -24,18 +24,19 @@ def test_sqlite_metastore(sqlite_db):
     assert obj2.db.db_file == sqlite_db.db_file
     assert obj2.clone_params() == obj.clone_params()
 
-    # Test serialization
+    # Test serialization JSON format
     serialized = obj.serialize()
     assert serialized
-    serialized_pickled = base64.b64decode(serialized.encode())
-    assert serialized_pickled
-    (f, args, kwargs) = pickle.loads(serialized_pickled)  # noqa: S301
-    assert str(f) == str(SQLiteMetastore.init_after_clone)
-    assert args == []
-    assert kwargs["uri"] == uri
-    assert str(kwargs["db_clone_params"]) == str(sqlite_db.clone_params())
-
-    # Test deserialization
+    raw = base64.b64decode(serialized.encode())
+    data = json.loads(raw.decode())
+    assert data["callable"] == "sqlite.metastore.init_after_clone"
+    assert data["args"] == []
+    assert data["kwargs"]["uri"] == uri
+    nested = data["kwargs"]["db_clone_params"]
+    assert nested["callable"] == "sqlite.from_db_file"
+    assert nested["args"] == [":memory:"]
+    assert nested["kwargs"] == {}
+
     obj3 = deserialize(serialized)
     assert isinstance(obj3, SQLiteMetastore)
     assert obj3.uri == uri