kumoai 2.13.0.dev202512040649__cp313-cp313-win_amd64.whl → 2.14.0.dev202601081732__cp313-cp313-win_amd64.whl

kumoai/experimental/rfm/backend/sqlite/__init__.py

@@ -1,5 +1,5 @@
 from pathlib import Path
-from typing import Any, TypeAlias, Union
+from typing import Any, TypeAlias
 
 try:
     import adbc_driver_sqlite.dbapi as adbc
@@ -11,7 +11,7 @@ except ImportError:
 Connection: TypeAlias = adbc.AdbcSqliteConnection
 
 
-def connect(uri: Union[str, Path, None] = None, **kwargs: Any) -> Connection:
+def connect(uri: str | Path | None = None, **kwargs: Any) -> Connection:
     r"""Opens a connection to a :class:`sqlite` database.
 
     uri: The path to the database file to be opened.
@@ -22,9 +22,11 @@ def connect(uri: Union[str, Path, None] = None, **kwargs: Any) -> Connection:
 
 
 from .table import SQLiteTable  # noqa: E402
+from .sampler import SQLiteSampler  # noqa: E402
 
 __all__ = [
     'connect',
     'Connection',
     'SQLiteTable',
+    'SQLiteSampler',
 ]

kumoai/experimental/rfm/backend/sqlite/sampler.py

@@ -0,0 +1,454 @@
+import warnings
+from collections import defaultdict
+from typing import TYPE_CHECKING
+
+import numpy as np
+import pandas as pd
+import pyarrow as pa
+from kumoapi.pquery import ValidatedPredictiveQuery
+
+from kumoai.experimental.rfm.backend.sqlite import SQLiteTable
+from kumoai.experimental.rfm.base import SQLSampler, Table
+from kumoai.experimental.rfm.pquery import PQueryPandasExecutor
+from kumoai.utils import ProgressLogger, quote_ident
+
+if TYPE_CHECKING:
+    from kumoai.experimental.rfm import Graph
+
+
+class SQLiteSampler(SQLSampler):
+    def __init__(
+        self,
+        graph: 'Graph',
+        verbose: bool | ProgressLogger = True,
+        optimize: bool = False,
+    ) -> None:
+        super().__init__(graph=graph, verbose=verbose)
+
+        for table in graph.tables.values():
+            assert isinstance(table, SQLiteTable)
+            self._connection = table._connection
+
+        if optimize:
+            with self._connection.cursor() as cursor:
+                cursor.execute("PRAGMA temp_store = MEMORY")
+                cursor.execute("PRAGMA cache_size = -2000000")  # 2 GB
+
+        # Collect database indices for speeding sampling:
+        index_dict: dict[str, set[tuple[str, ...]]] = defaultdict(set)
+        for table_name, primary_key in self.primary_key_dict.items():
+            source_table = self.source_table_dict[table_name]
+            if primary_key not in source_table:
+                continue  # No physical column.
+            if source_table[primary_key].is_unique_key:
+                continue
+            index_dict[table_name].add((primary_key, ))
+        for src_table_name, foreign_key, _ in graph.edges:
+            source_table = self.source_table_dict[src_table_name]
+            if foreign_key not in source_table:
+                continue  # No physical column.
+            if source_table[foreign_key].is_unique_key:
+                continue
+            time_column = self.time_column_dict.get(src_table_name)
+            if time_column is not None and time_column in source_table:
+                index_dict[src_table_name].add((foreign_key, time_column))
+            else:
+                index_dict[src_table_name].add((foreign_key, ))
+
+        # Only maintain missing indices:
+        with self._connection.cursor() as cursor:
+            for table_name in list(index_dict.keys()):
+                indices = index_dict[table_name]
+                source_name = self.source_name_dict[table_name]
+                sql = f"PRAGMA index_list({source_name})"
+                cursor.execute(sql)
+                for _, index_name, *_ in cursor.fetchall():
+                    sql = f"PRAGMA index_info({quote_ident(index_name)})"
+                    cursor.execute(sql)
+                    # Fetch index information and sort by `seqno`:
+                    index_info = tuple(info[2] for info in sorted(
+                        cursor.fetchall(), key=lambda x: x[0]))
+                    # Remove all indices in case primary index already exists:
+                    for index in list(indices):
+                        if index_info[0] == index[0]:
+                            indices.discard(index)
+                if len(indices) == 0:
+                    del index_dict[table_name]
+
+        if optimize and len(index_dict) > 0:
+            if not isinstance(verbose, ProgressLogger):
+                verbose = ProgressLogger.default(
+                    msg="Optimizing SQLite database",
+                    verbose=verbose,
+                )
+
+            with verbose as logger, self._connection.cursor() as cursor:
+                for table_name, indices in index_dict.items():
+                    for index in indices:
+                        name = f"kumo_index_{table_name}_{'_'.join(index)}"
+                        name = quote_ident(name)
+                        columns = ', '.join(quote_ident(v) for v in index)
+                        columns += ' DESC' if len(index) > 1 else ''
+                        source_name = self.source_name_dict[table_name]
+                        sql = (f"CREATE INDEX IF NOT EXISTS {name}\n"
+                               f"ON {source_name}({columns})")
+                        cursor.execute(sql)
+                        self._connection.commit()
+                        if len(index) > 1:
+                            logger.log(f"Created index on {index} in table "
+                                       f"'{table_name}'")
+                        else:
+                            logger.log(f"Created index on '{index[0]}' in "
+                                       f"table '{table_name}'")
+
+        elif len(index_dict) > 0:
+            num = sum(len(indices) for indices in index_dict.values())
+            index_repr = '1 index' if num == 1 else f'{num} indices'
+            num = len(index_dict)
+            table_repr = '1 table' if num == 1 else f'{num} tables'
+            warnings.warn(f"Missing {index_repr} in {table_repr} for optimal "
+                          f"database querying. For improving runtime, we "
+                          f"strongly suggest to create indices for primary "
+                          f"and foreign keys, e.g., automatically by "
+                          f"instantiating KumoRFM via "
+                          f"`KumoRFM(graph, optimize=True)`.")
+
+    def _get_min_max_time_dict(
+        self,
+        table_names: list[str],
+    ) -> dict[str, tuple[pd.Timestamp, pd.Timestamp]]:
+        selects: list[str] = []
+        for table_name in table_names:
+            column = self.time_column_dict[table_name]
+            column_ref = self.table_column_ref_dict[table_name][column]
+            select = (f"SELECT\n"
+                      f"  ? as table_name,\n"
+                      f"  MIN({column_ref}) as min_date,\n"
+                      f"  MAX({column_ref}) as max_date\n"
+                      f"FROM {self.source_name_dict[table_name]}")
+            selects.append(select)
+        sql = "\nUNION ALL\n".join(selects)
+
+        out_dict: dict[str, tuple[pd.Timestamp, pd.Timestamp]] = {}
+        with self._connection.cursor() as cursor:
+            cursor.execute(sql, table_names)
+            for table_name, _min, _max in cursor.fetchall():
+                out_dict[table_name] = (
+                    pd.Timestamp.max if _min is None else pd.Timestamp(_min),
+                    pd.Timestamp.min if _max is None else pd.Timestamp(_max),
+                )
+        return out_dict
+
+    def _sample_entity_table(
+        self,
+        table_name: str,
+        columns: set[str],
+        num_rows: int,
+        random_seed: int | None = None,
+    ) -> pd.DataFrame:
+        # NOTE SQLite does not natively support passing a `random_seed`.
+
+        source_table = self.source_table_dict[table_name]
+        filters: list[str] = []
+
+        key = self.primary_key_dict[table_name]
+        if key not in source_table or source_table[key].is_nullable:
+            key_ref = self.table_column_ref_dict[table_name][key]
+            filters.append(f" {key_ref} IS NOT NULL")
+
+        column = self.time_column_dict.get(table_name)
+        if column is None:
+            pass
+        elif column not in source_table or source_table[column].is_nullable:
+            column_ref = self.table_column_ref_dict[table_name][column]
+            filters.append(f" {column_ref} IS NOT NULL")
+
+        # TODO Make this query more efficient - it does full table scan.
+        projections = [
+            self.table_column_proj_dict[table_name][column]
+            for column in columns
+        ]
+        sql = (f"SELECT {', '.join(projections)}\n"
+               f"FROM {self.source_name_dict[table_name]}")
+        if len(filters) > 0:
+            sql += f"\nWHERE{' AND'.join(filters)}"
+        sql += f"\nORDER BY RANDOM() LIMIT {num_rows}"
+
+        with self._connection.cursor() as cursor:
+            # NOTE This may return duplicate primary keys. This is okay.
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        return Table._sanitize(
+            df=table.to_pandas(types_mapper=pd.ArrowDtype),
+            dtype_dict=self.table_dtype_dict[table_name],
+            stype_dict=self.table_stype_dict[table_name],
+        )
+
+    def _sample_target(
+        self,
+        query: ValidatedPredictiveQuery,
+        entity_df: pd.DataFrame,
+        train_index: np.ndarray,
+        train_time: pd.Series,
+        num_train_examples: int,
+        test_index: np.ndarray,
+        test_time: pd.Series,
+        num_test_examples: int,
+        columns_dict: dict[str, set[str]],
+        time_offset_dict: dict[
+            tuple[str, str, str],
+            tuple[pd.DateOffset | None, pd.DateOffset],
+        ],
+    ) -> tuple[pd.Series, np.ndarray, pd.Series, np.ndarray]:
+        train_y, train_mask = self._sample_target_set(
+            query=query,
+            entity_df=entity_df,
+            index=train_index,
+            anchor_time=train_time,
+            num_examples=num_train_examples,
+            columns_dict=columns_dict,
+            time_offset_dict=time_offset_dict,
+        )
+
+        test_y, test_mask = self._sample_target_set(
+            query=query,
+            entity_df=entity_df,
+            index=test_index,
+            anchor_time=test_time,
+            num_examples=num_test_examples,
+            columns_dict=columns_dict,
+            time_offset_dict=time_offset_dict,
+        )
+
+        return train_y, train_mask, test_y, test_mask
+
+    def _by_pkey(
+        self,
+        table_name: str,
+        index: pd.Series,
+        columns: set[str],
+    ) -> tuple[pd.DataFrame, np.ndarray]:
+        source_table = self.source_table_dict[table_name]
+        key = self.primary_key_dict[table_name]
+        key_ref = self.table_column_ref_dict[table_name][key]
+        projections = [
+            self.table_column_proj_dict[table_name][column]
+            for column in columns
+        ]
+
+        tmp = pa.table([pa.array(index)], names=['__kumo_id__'])
+        tmp_name = f'tmp_{table_name}_{key}_{id(tmp)}'
+
+        sql = (f"SELECT "
+               f"tmp.rowid - 1 as __kumo_batch__, "
+               f"{', '.join(projections)}\n"
+               f"FROM {quote_ident(tmp_name)} tmp\n"
+               f"JOIN {self.source_name_dict[table_name]} ent\n")
+        if key in source_table and source_table[key].is_unique_key:
+            sql += (f"  ON {key_ref} = tmp.__kumo_id__")
+        else:
+            sql += (f"  ON ent.rowid = (\n"
+                    f"    SELECT rowid\n"
+                    f"    FROM {self.source_name_dict[table_name]}\n"
+                    f"    WHERE {key_ref} == tmp.__kumo_id__\n"
+                    f"    LIMIT 1\n"
+                    f")")
+
+        with self._connection.cursor() as cursor:
+            cursor.adbc_ingest(tmp_name, tmp, mode='replace')
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        batch = table['__kumo_batch__'].to_numpy()
+        batch_index = table.schema.get_field_index('__kumo_batch__')
+        table = table.remove_column(batch_index)
+
+        return Table._sanitize(
+            df=table.to_pandas(),
+            dtype_dict=self.table_dtype_dict[table_name],
+            stype_dict=self.table_stype_dict[table_name],
+        ), batch
+
+    def _by_fkey(
+        self,
+        table_name: str,
+        foreign_key: str,
+        index: pd.Series,
+        num_neighbors: int,
+        anchor_time: pd.Series | None,
+        columns: set[str],
+    ) -> tuple[pd.DataFrame, np.ndarray]:
+        time_column = self.time_column_dict.get(table_name)
+
+        # NOTE SQLite does not have a native datetime format. Currently, we
+        # assume timestamps are given as `TEXT` in `ISO-8601 UTC`:
+        tmp = pa.table([pa.array(index)], names=['__kumo_id__'])
+        if time_column is not None and anchor_time is not None:
+            anchor_time = anchor_time.dt.strftime("%Y-%m-%d %H:%M:%S")
+            tmp = tmp.append_column('__kumo_time__', pa.array(anchor_time))
+        tmp_name = f'tmp_{table_name}_{foreign_key}_{id(tmp)}'
+
+        key_ref = self.table_column_ref_dict[table_name][foreign_key]
+        projections = [
+            self.table_column_proj_dict[table_name][column]
+            for column in columns
+        ]
+        sql = (f"SELECT "
+               f"tmp.rowid - 1 as __kumo_batch__, "
+               f"{', '.join(projections)}\n"
+               f"FROM {quote_ident(tmp_name)} tmp\n"
+               f"JOIN {self.source_name_dict[table_name]} fact\n"
+               f"ON fact.rowid IN (\n"
+               f"  SELECT rowid\n"
+               f"  FROM {self.source_name_dict[table_name]}\n"
+               f"  WHERE {key_ref} = tmp.__kumo_id__\n")
+        if time_column is not None and anchor_time is not None:
+            time_ref = self.table_column_ref_dict[table_name][time_column]
+            sql += f"  AND {time_ref} <= tmp.__kumo_time__\n"
+        if time_column is not None:
+            time_ref = self.table_column_ref_dict[table_name][time_column]
+            sql += f"  ORDER BY {time_ref} DESC\n"
+        sql += (f"  LIMIT {num_neighbors}\n"
+                f")")
+
+        with self._connection.cursor() as cursor:
+            cursor.adbc_ingest(tmp_name, tmp, mode='replace')
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        batch = table['__kumo_batch__'].to_numpy()
+        batch_index = table.schema.get_field_index('__kumo_batch__')
+        table = table.remove_column(batch_index)
+
+        return Table._sanitize(
+            df=table.to_pandas(),
+            dtype_dict=self.table_dtype_dict[table_name],
+            stype_dict=self.table_stype_dict[table_name],
+        ), batch
+
+    # Helper Methods ##########################################################
+
+    def _by_time(
+        self,
+        table_name: str,
+        foreign_key: str,
+        index: pd.Series,
+        anchor_time: pd.Series,
+        min_offset: pd.DateOffset | None,
+        max_offset: pd.DateOffset,
+        columns: set[str],
+    ) -> tuple[pd.DataFrame, np.ndarray]:
+        time_column = self.time_column_dict[table_name]
+
+        # NOTE SQLite does not have a native datetime format. Currently, we
+        # assume timestamps are given as `TEXT` in `ISO-8601 UTC`:
+        tmp = pa.table([pa.array(index)], names=['__kumo_id__'])
+        end_time = anchor_time + max_offset
+        end_time = end_time.dt.strftime("%Y-%m-%d %H:%M:%S")
+        tmp = tmp.append_column('__kumo_end__', pa.array(end_time))
+        if min_offset is not None:
+            start_time = anchor_time + min_offset
+            start_time = start_time.dt.strftime("%Y-%m-%d %H:%M:%S")
+            tmp = tmp.append_column('__kumo_start__', pa.array(start_time))
+        tmp_name = f'tmp_{table_name}_{foreign_key}_{id(tmp)}'
+
+        key_ref = self.table_column_ref_dict[table_name][foreign_key]
+        time_ref = self.table_column_ref_dict[table_name][time_column]
+        projections = [
+            self.table_column_proj_dict[table_name][column]
+            for column in columns
+        ]
+        sql = (f"SELECT "
+               f"tmp.rowid - 1 as __kumo_batch__, "
+               f"{', '.join(projections)}\n"
+               f"FROM {quote_ident(tmp_name)} tmp\n"
+               f"JOIN {self.source_name_dict[table_name]}\n"
+               f"  ON {key_ref} = tmp.__kumo_id__\n"
+               f" AND {time_ref} <= tmp.__kumo_end__")
+        if min_offset is not None:
+            sql += f"\n AND {time_ref} > tmp.__kumo_start__"
+
+        with self._connection.cursor() as cursor:
+            cursor.adbc_ingest(tmp_name, tmp, mode='replace')
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        batch = table['__kumo_batch__'].to_numpy()
+        batch_index = table.schema.get_field_index('__kumo_batch__')
+        table = table.remove_column(batch_index)
+
+        return Table._sanitize(
+            df=table.to_pandas(types_mapper=pd.ArrowDtype),
+            dtype_dict=self.table_dtype_dict[table_name],
+            stype_dict=self.table_stype_dict[table_name],
+        ), batch
+
+    def _sample_target_set(
+        self,
+        query: ValidatedPredictiveQuery,
+        entity_df: pd.DataFrame,
+        index: np.ndarray,
+        anchor_time: pd.Series,
+        num_examples: int,
+        columns_dict: dict[str, set[str]],
+        time_offset_dict: dict[
+            tuple[str, str, str],
+            tuple[pd.DateOffset | None, pd.DateOffset],
+        ],
+        batch_size: int = 10_000,
+    ) -> tuple[pd.Series, np.ndarray]:
+
+        count = 0
+        ys: list[pd.Series] = []
+        mask = np.full(len(index), False, dtype=bool)
+        for start in range(0, len(index), batch_size):
+            df = entity_df.iloc[index[start:start + batch_size]]
+            time = anchor_time.iloc[start:start + batch_size]
+
+            feat_dict: dict[str, pd.DataFrame] = {query.entity_table: df}
+            time_dict: dict[str, pd.Series] = {}
+            time_column = self.time_column_dict.get(query.entity_table)
+            if time_column in columns_dict[query.entity_table]:
+                time_dict[query.entity_table] = df[time_column]
+            batch_dict: dict[str, np.ndarray] = {
+                query.entity_table: np.arange(len(df)),
+            }
+            for edge_type, (_min, _max) in time_offset_dict.items():
+                table_name, foreign_key, _ = edge_type
+                feat_dict[table_name], batch_dict[table_name] = self._by_time(
+                    table_name=table_name,
+                    foreign_key=foreign_key,
+                    index=df[self.primary_key_dict[query.entity_table]],
+                    anchor_time=time,
+                    min_offset=_min,
+                    max_offset=_max,
+                    columns=columns_dict[table_name],
+                )
+                time_column = self.time_column_dict.get(table_name)
+                if time_column in columns_dict[table_name]:
+                    time_dict[table_name] = feat_dict[table_name][time_column]
+
+            y, _mask = PQueryPandasExecutor().execute(
+                query=query,
+                feat_dict=feat_dict,
+                time_dict=time_dict,
+                batch_dict=batch_dict,
+                anchor_time=anchor_time,
+                num_forecasts=query.num_forecasts,
+            )
+            ys.append(y)
+            mask[start:start + batch_size] = _mask
+
+            count += len(y)
+            if count >= num_examples:
+                break
+
+        if len(ys) == 0:
+            y = pd.Series([], dtype=float)
+        elif len(ys) == 1:
+            y = ys[0]
+        else:
+            y = pd.concat(ys, axis=0, ignore_index=True)
+
+        return y, mask