kumoai 2.12.0.dev202510231830__cp311-cp311-win_amd64.whl → 2.14.0.dev202512311733__cp311-cp311-win_amd64.whl

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

@@ -0,0 +1,32 @@
+from pathlib import Path
+from typing import Any, TypeAlias
+
+try:
+    import adbc_driver_sqlite.dbapi as adbc
+except ImportError:
+    raise ImportError("No module named 'adbc_driver_sqlite'. Please install "
+                      "Kumo SDK with the 'sqlite' extension via "
+                      "`pip install kumoai[sqlite]`.")
+
+Connection: TypeAlias = adbc.AdbcSqliteConnection
+
+
+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.
+    kwargs: Additional connection arguments, following the
+        :class:`adbc_driver_sqlite` protocol.
+    """
+    return adbc.connect(uri, **kwargs)
+
+
+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 +1,398 @@
+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,
+        pkey: 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(pkey)], 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
+
+    # Helper Methods ##########################################################
+
+    def _by_time(
+        self,
+        table_name: str,
+        fkey: str,
+        pkey: 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(pkey)], 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}_{fkey}_{id(tmp)}'
+
+        key_ref = self.table_column_ref_dict[table_name][fkey]
+        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]} fact\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, fkey, _ = edge_type
+                feat_dict[table_name], batch_dict[table_name] = self._by_time(
+                    table_name=table_name,
+                    fkey=fkey,
+                    pkey=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

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

@@ -0,0 +1,184 @@
+import re
+from collections import Counter
+from collections.abc import Sequence
+from typing import cast
+
+import pandas as pd
+from kumoapi.model_plan import MissingType
+from kumoapi.typing import Dtype
+
+from kumoai.experimental.rfm.backend.sqlite import Connection
+from kumoai.experimental.rfm.base import (
+    ColumnSpec,
+    ColumnSpecType,
+    DataBackend,
+    SourceColumn,
+    SourceForeignKey,
+    Table,
+)
+from kumoai.utils import quote_ident
+
+
+class SQLiteTable(Table):
+    r"""A table backed by a :class:`sqlite` database.
+
+    Args:
+        connection: The connection to a :class:`sqlite` database.
+        name: The name of this table.
+        source_name: The source name of this table. If set to ``None``,
+            ``name`` is being used.
+        columns: The selected columns of this table.
+        primary_key: The name of the primary key of this table, if it exists.
+        time_column: The name of the time column of this table, if it exists.
+        end_time_column: The name of the end time column of this table, if it
+            exists.
+    """
+    def __init__(
+        self,
+        connection: Connection,
+        name: str,
+        source_name: str | None = None,
+        columns: Sequence[ColumnSpecType] | None = None,
+        primary_key: MissingType | str | None = MissingType.VALUE,
+        time_column: str | None = None,
+        end_time_column: str | None = None,
+    ) -> None:
+
+        self._connection = connection
+
+        super().__init__(
+            name=name,
+            source_name=source_name,
+            columns=columns,
+            primary_key=primary_key,
+            time_column=time_column,
+            end_time_column=end_time_column,
+        )
+
+    @property
+    def backend(self) -> DataBackend:
+        return cast(DataBackend, DataBackend.SQLITE)
+
+    def _get_source_columns(self) -> list[SourceColumn]:
+        source_columns: list[SourceColumn] = []
+        with self._connection.cursor() as cursor:
+            sql = f"PRAGMA table_info({self._quoted_source_name})"
+            cursor.execute(sql)
+            columns = cursor.fetchall()
+
+            if len(columns) == 0:
+                raise ValueError(f"Table '{self.source_name}' does not exist "
+                                 f"in the SQLite database")
+
+            unique_keys: set[str] = set()
+            sql = f"PRAGMA index_list({self._quoted_source_name})"
+            cursor.execute(sql)
+            for _, index_name, is_unique, *_ in cursor.fetchall():
+                if bool(is_unique):
+                    sql = f"PRAGMA index_info({quote_ident(index_name)})"
+                    cursor.execute(sql)
+                    index = cursor.fetchall()
+                    if len(index) == 1:
+                        unique_keys.add(index[0][2])
+
+            # Special SQLite case that creates a rowid alias for
+            # `INTEGER PRIMARY KEY` annotated columns:
+            rowid_candidates = [
+                column for _, column, dtype, _, _, is_pkey in columns
+                if bool(is_pkey) and dtype.strip().upper() == 'INTEGER'
+            ]
+            if len(rowid_candidates) == 1:
+                unique_keys.add(rowid_candidates[0])
+
+            for _, column, dtype, notnull, _, is_pkey in columns:
+                source_column = SourceColumn(
+                    name=column,
+                    dtype=self._to_dtype(dtype),
+                    is_primary_key=bool(is_pkey),
+                    is_unique_key=column in unique_keys,
+                    is_nullable=not bool(is_pkey) and not bool(notnull),
+                )
+                source_columns.append(source_column)
+
+        return source_columns
+
+    def _get_source_foreign_keys(self) -> list[SourceForeignKey]:
+        source_foreign_keys: list[SourceForeignKey] = []
+        with self._connection.cursor() as cursor:
+            sql = f"PRAGMA foreign_key_list({self._quoted_source_name})"
+            cursor.execute(sql)
+            rows = cursor.fetchall()
+            counts = Counter(row[0] for row in rows)
+            for idx, _, dst_table, foreign_key, primary_key, *_ in rows:
+                if counts[idx] == 1:
+                    source_foreign_key = SourceForeignKey(
+                        name=foreign_key,
+                        dst_table=dst_table,
+                        primary_key=primary_key,
+                    )
+                    source_foreign_keys.append(source_foreign_key)
+        return source_foreign_keys
+
+    def _get_source_sample_df(self) -> pd.DataFrame:
+        with self._connection.cursor() as cursor:
+            columns = [quote_ident(col) for col in self._source_column_dict]
+            sql = (f"SELECT {', '.join(columns)} "
+                   f"FROM {self._quoted_source_name} "
+                   f"ORDER BY rowid "
+                   f"LIMIT {self._NUM_SAMPLE_ROWS}")
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        if len(table) == 0:
+            raise RuntimeError(f"Table '{self.source_name}' is empty")
+
+        return self._sanitize(
+            df=table.to_pandas(types_mapper=pd.ArrowDtype),
+            dtype_dict={
+                column.name: column.dtype
+                for column in self._source_column_dict.values()
+            },
+            stype_dict=None,
+        )
+
+    def _get_num_rows(self) -> int | None:
+        return None
+
+    def _get_expr_sample_df(
+        self,
+        columns: Sequence[ColumnSpec],
+    ) -> pd.DataFrame:
+        with self._connection.cursor() as cursor:
+            projections = [
+                f"{column.expr} AS {quote_ident(column.name)}"
+                for column in columns
+            ]
+            sql = (f"SELECT {', '.join(projections)} "
+                   f"FROM {self._quoted_source_name} "
+                   f"ORDER BY rowid "
+                   f"LIMIT {self._NUM_SAMPLE_ROWS}")
+            cursor.execute(sql)
+            table = cursor.fetch_arrow_table()
+
+        if len(table) == 0:
+            raise RuntimeError(f"Table '{self.source_name}' is empty")
+
+        return self._sanitize(
+            df=table.to_pandas(types_mapper=pd.ArrowDtype),
+            dtype_dict={column.name: column.dtype
+                        for column in columns},
+            stype_dict=None,
+        )
+
+    @staticmethod
+    def _to_dtype(dtype: str | None) -> Dtype | None:
+        if dtype is None:
+            return None
+        dtype = dtype.strip().upper()
+        if re.search('INT', dtype):
+            return Dtype.int
+        if re.search('TEXT|CHAR|CLOB', dtype):
+            return Dtype.string
+        if re.search('REAL|FLOA|DOUB', dtype):
+            return Dtype.float
+        return None  # NUMERIC affinity.

kumoai/experimental/rfm/base/__init__.py

@@ -0,0 +1,30 @@
+from kumoapi.common import StrEnum
+
+
+class DataBackend(StrEnum):
+    LOCAL = 'local'
+    SQLITE = 'sqlite'
+    SNOWFLAKE = 'snowflake'
+
+
+from .source import SourceColumn, SourceForeignKey  # noqa: E402
+from .expression import Expression, LocalExpression  # noqa: E402
+from .column import ColumnSpec, ColumnSpecType, Column  # noqa: E402
+from .table import Table  # noqa: E402
+from .sampler import SamplerOutput, Sampler  # noqa: E402
+from .sql_sampler import SQLSampler  # noqa: E402
+
+__all__ = [
+    'DataBackend',
+    'SourceColumn',
+    'SourceForeignKey',
+    'Expression',
+    'LocalExpression',
+    'ColumnSpec',
+    'ColumnSpecType',
+    'Column',
+    'Table',
+    'SamplerOutput',
+    'Sampler',
+    'SQLSampler',
+]