"""SQLAlchemy filter constructs for advanced query operations.
This module provides a comprehensive collection of filter datastructures designed to
enhance SQLAlchemy query construction. It implements type-safe, reusable filter patterns
for common database query operations.
Features:
Type-safe filter construction, datetime range filtering, collection-based filtering,
pagination support, search operations, and customizable ordering.
Note:
All filter classes implement the :class:`StatementFilter` ABC, ensuring consistent
interface across different filter types.
See Also:
- :class:`sqlalchemy.sql.expression.Select`: Core SQLAlchemy select expression
- :class:`sqlalchemy.orm.Query`: SQLAlchemy ORM query interface
- :mod:`advanced_alchemy.base`: Base model definitions
"""
import datetime
import logging
from abc import ABC, abstractmethod
from collections.abc import Collection
from dataclasses import dataclass
from operator import attrgetter
from typing import (
TYPE_CHECKING,
Any,
Callable,
ClassVar,
Generic,
Literal,
Optional,
Union,
cast,
)
from sqlalchemy import (
BinaryExpression,
ColumnElement,
Date,
Delete,
Integer,
Select,
Update,
and_,
any_,
case,
exists,
false,
literal_column,
not_,
or_,
select,
text,
true,
)
from sqlalchemy.ext.compiler import compiles
from sqlalchemy.sql import operators as op
from sqlalchemy.sql.dml import ReturningDelete, ReturningUpdate
from sqlalchemy.sql.elements import UnaryExpression
from typing_extensions import TypeAlias, TypedDict, TypeVar
from advanced_alchemy.base import ModelProtocol
if TYPE_CHECKING:
from sqlalchemy.orm import InstrumentedAttribute
FilterFieldName: TypeAlias = Union[str, ColumnElement[Any], InstrumentedAttribute[Any]]
InstrumentedField: TypeAlias = Union[ColumnElement[Any], InstrumentedAttribute[Any]]
else:
FilterFieldName: TypeAlias = Any
InstrumentedField: TypeAlias = Any
__all__ = (
"LIKE_ESCAPE_CHAR",
"BeforeAfter",
"BooleanFilter",
"ChoicesFilter",
"CollectionFilter",
"ComparisonFilter",
"ExistsFilter",
"FilterGroup",
"FilterMap",
"FilterTypes",
"InAnyFilter",
"LimitOffset",
"LogicalOperatorMap",
"MultiFilter",
"NotExistsFilter",
"NotInCollectionFilter",
"NotInSearchFilter",
"NotNullFilter",
"NullFilter",
"OnBeforeAfter",
"OrderBy",
"PaginationFilter",
"SearchFilter",
"StatementFilter",
"StatementFilterT",
"StatementTypeT",
"escape_like_value",
)
T = TypeVar("T")
ModelT = TypeVar("ModelT", bound=ModelProtocol)
StatementFilterT = TypeVar("StatementFilterT", bound="StatementFilter")
StatementTypeT = TypeVar(
"StatementTypeT",
bound=Union[
ReturningDelete[tuple[Any]], ReturningUpdate[tuple[Any]], Select[tuple[Any]], Select[Any], Update, Delete
],
)
logger = logging.getLogger("advanced_alchemy")
# Define TypedDicts for filter and logical maps
[docs]
class FilterMap(TypedDict):
before_after: "type[BeforeAfter]"
on_before_after: "type[OnBeforeAfter]"
boolean: "type[BooleanFilter]"
choices: "type[ChoicesFilter[Any]]"
collection: "type[CollectionFilter[Any]]"
not_in_collection: "type[NotInCollectionFilter[Any]]"
limit_offset: "type[LimitOffset]"
null: "type[NullFilter]"
not_null: "type[NotNullFilter]"
order_by: "type[OrderBy]"
search: "type[SearchFilter]"
not_in_search: "type[NotInSearchFilter]"
comparison: "type[ComparisonFilter]"
exists: "type[ExistsFilter]"
not_exists: "type[NotExistsFilter]"
filter_group: "type[FilterGroup]"
[docs]
class LogicalOperatorMap(TypedDict):
and_: Callable[..., ColumnElement[bool]]
or_: Callable[..., ColumnElement[bool]]
[docs]
class StatementFilter(ABC):
"""Abstract base class for SQLAlchemy statement filters.
This class defines the interface for all filter types in the system. Each filter
implementation must provide a method to append its filtering logic to an existing
SQLAlchemy statement.
"""
[docs]
@abstractmethod
def append_to_statement(
self, statement: StatementTypeT, model: type[ModelT], *args: Any, **kwargs: Any
) -> StatementTypeT:
"""Append filter conditions to a SQLAlchemy statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
*args: Additional positional arguments
**kwargs: Additional keyword arguments
Returns:
StatementTypeT: Modified SQLAlchemy statement with filter conditions applied
Raises:
NotImplementedError: If the concrete class doesn't implement this method
Note:
This method must be implemented by all concrete filter classes.
See Also:
:meth:`sqlalchemy.sql.expression.Select.where`: SQLAlchemy where clause
"""
return statement
@staticmethod
def _get_instrumented_attr(model: Any, key: FilterFieldName) -> InstrumentedField:
"""Get SQLAlchemy instrumented attribute from model.
Args:
model: SQLAlchemy model class or instance
key: Attribute name or instrumented attribute
Returns:
InstrumentedAttribute[Any]: SQLAlchemy instrumented attribute
See Also:
:class:`sqlalchemy.orm.attributes.InstrumentedAttribute`: SQLAlchemy attribute
"""
return cast("InstrumentedAttribute[Any]", getattr(model, key)) if isinstance(key, str) else key
[docs]
@dataclass
class BeforeAfter(StatementFilter):
"""DateTime range filter with exclusive bounds.
This filter creates date/time range conditions using < and > operators,
excluding the boundary values.
If either `before` or `after` is None, that boundary condition is not applied.
See Also:
---------
:class:`OnBeforeAfter` : Inclusive datetime range filtering
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
before: Optional[datetime.datetime]
"""Filter results where field is earlier than this value."""
after: Optional[datetime.datetime]
"""Filter results where field is later than this value."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply datetime range conditions to statement.
Parameters
----------
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
Returns:
--------
StatementTypeT
Modified statement with datetime range conditions
"""
field = self._get_instrumented_attr(model, self.field_name)
if self.before is not None:
statement = cast("StatementTypeT", statement.where(field < self.before))
if self.after is not None:
statement = cast("StatementTypeT", statement.where(field > self.after))
return statement
[docs]
@dataclass
class OnBeforeAfter(StatementFilter):
"""DateTime range filter with inclusive bounds.
This filter creates date/time range conditions using <= and >= operators,
including the boundary values.
If either `on_or_before` or `on_or_after` is None, that boundary condition
is not applied.
See Also:
---------
:class:`BeforeAfter` : Exclusive datetime range filtering
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
on_or_before: Optional[datetime.datetime]
"""Filter results where field is on or earlier than this value."""
on_or_after: Optional[datetime.datetime]
"""Filter results where field is on or later than this value."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply inclusive datetime range conditions to statement.
Parameters
----------
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
Returns:
--------
StatementTypeT
Modified statement with inclusive datetime range conditions
"""
field = self._get_instrumented_attr(model, self.field_name)
if self.on_or_before is not None:
statement = cast("StatementTypeT", statement.where(field <= self.on_or_before))
if self.on_or_after is not None:
statement = cast("StatementTypeT", statement.where(field >= self.on_or_after))
return statement
[docs]
class InAnyFilter(StatementFilter, ABC):
"""Base class for filters using IN or ANY operators.
This abstract class provides common functionality for filters that check
membership in a collection using either the SQL IN operator or the ANY operator.
"""
[docs]
@dataclass
class CollectionFilter(InAnyFilter, Generic[T]):
"""Data required to construct a WHERE ... IN (...) clause.
This filter restricts records based on a field's presence in a collection of values.
The filter supports both ``IN`` and ``ANY`` operators for collection membership testing.
Use ``prefer_any=True`` in ``append_to_statement`` to use the ``ANY`` operator.
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
values: Union[Collection[T], None]
"""Values for the ``IN`` clause. If this is None, no filter is applied.
An empty list will force an empty result set (WHERE 1=-1)"""
[docs]
def append_to_statement(
self,
statement: StatementTypeT,
model: type[ModelT],
prefer_any: bool = False,
) -> StatementTypeT:
"""Apply a WHERE ... IN or WHERE ... ANY (...) clause to the statement.
Parameters
----------
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
prefer_any : bool, optional
If True, uses the SQLAlchemy :func:`any_` operator instead of
:func:`in_` for the filter condition
Returns:
--------
StatementTypeT
Modified statement with the appropriate IN conditions
"""
field = self._get_instrumented_attr(model, self.field_name)
if self.values is None:
return statement
if not self.values:
# Return empty result set by forcing a false condition
return cast("StatementTypeT", statement.where(text("1=-1")))
if prefer_any:
return cast("StatementTypeT", statement.where(any_(self.values) == field)) # type: ignore[arg-type]
return cast("StatementTypeT", statement.where(field.in_(self.values)))
[docs]
@dataclass
class ChoicesFilter(CollectionFilter[T]):
"""Filter records whose field matches one of a fixed set of choices."""
[docs]
@dataclass
class BooleanFilter(StatementFilter):
"""Filter records by a boolean field value.
If ``value`` is ``None``, the statement is returned unchanged.
"""
field_name: "Union[str, ColumnElement[Any], InstrumentedAttribute[Any]]"
"""Field name, model attribute, or func expression."""
value: Optional[bool]
"""Boolean value to match. If None, no filter is applied."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply a boolean equality condition to the statement."""
if self.value is None:
return statement
field = self._get_instrumented_attr(model, self.field_name)
return cast("StatementTypeT", statement.where(field == self.value))
[docs]
@dataclass
class NotInCollectionFilter(InAnyFilter, Generic[T]):
"""Data required to construct a WHERE ... NOT IN (...) clause.
This filter restricts records based on a field's absence in a collection of values.
The filter supports both ``NOT IN`` and ``!= ANY`` operators for collection exclusion.
Use ``prefer_any=True`` in ``append_to_statement`` to use the ``ANY`` operator.
Parameters
----------
field_name : str
Name of the model attribute to filter on
values : abc.Collection[T] | None
Values for the ``NOT IN`` clause. If this is None or empty,
the filter is not applied.
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
values: Union[Collection[T], None]
"""Values for the ``NOT IN`` clause. If None or empty, no filter is applied."""
[docs]
def append_to_statement(
self,
statement: StatementTypeT,
model: type[ModelT],
prefer_any: bool = False,
) -> StatementTypeT:
"""Apply a WHERE ... NOT IN or WHERE ... != ANY(...) clause to the statement.
Parameters
----------
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
prefer_any : bool, optional
If True, uses the SQLAlchemy :func:`any_` operator instead of
:func:`notin_` for the filter condition
Returns:
--------
StatementTypeT
Modified statement with the appropriate NOT IN conditions
"""
field = self._get_instrumented_attr(model, self.field_name)
if not self.values:
# If None or empty, we do not modify the statement
return statement
if prefer_any:
return cast("StatementTypeT", statement.where(any_(self.values) != field)) # type: ignore[arg-type]
return cast("StatementTypeT", statement.where(field.notin_(self.values)))
[docs]
@dataclass
class NullFilter(StatementFilter):
"""Filter for NULL values (IS NULL).
This filter creates IS NULL conditions for database fields.
Use this to find records where a field has no value.
Example:
Basic NULL filtering::
from advanced_alchemy.filters import NullFilter
# Find records where email_verified_at is NULL
null_filter = NullFilter("email_verified_at")
unverified = await repo.get_many(null_filter)
With multiple filters::
from advanced_alchemy.filters import (
NullFilter,
CollectionFilter,
)
# Find unverified users in specific roles
filters = [
NullFilter("email_verified_at"),
CollectionFilter("role", ["admin", "moderator"]),
]
results = await repo.get_many(*filters)
See Also:
- :class:`NotNullFilter`: Filter for NOT NULL values
- :class:`CollectionFilter`: Filter by collection membership
- :meth:`sqlalchemy.sql.expression.ColumnOperators.is_`: IS NULL operator
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply IS NULL condition to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with IS NULL condition applied
"""
field = self._get_instrumented_attr(model, self.field_name)
return cast("StatementTypeT", statement.where(field.is_(None)))
[docs]
@dataclass
class NotNullFilter(StatementFilter):
"""Filter for NOT NULL values (IS NOT NULL).
This filter creates IS NOT NULL conditions for database fields.
Use this to find records where a field has a value.
Example:
Basic NOT NULL filtering::
from advanced_alchemy.filters import NotNullFilter
# Find records where email_verified_at is NOT NULL
not_null_filter = NotNullFilter("email_verified_at")
verified = await repo.get_many(not_null_filter)
With multiple filters::
from advanced_alchemy.filters import (
NotNullFilter,
CollectionFilter,
)
# Find verified users in specific roles
filters = [
NotNullFilter("email_verified_at"),
CollectionFilter("role", ["admin", "moderator"]),
]
results = await repo.get_many(*filters)
See Also:
- :class:`NullFilter`: Filter for NULL values
- :class:`CollectionFilter`: Filter by collection membership
- :meth:`sqlalchemy.sql.expression.ColumnOperators.is_not`: IS NOT NULL operator
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply IS NOT NULL condition to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with IS NOT NULL condition applied
"""
field = self._get_instrumented_attr(model, self.field_name)
return cast("StatementTypeT", statement.where(field.is_not(None)))
[docs]
@dataclass
class LimitOffset(PaginationFilter):
"""Limit and offset pagination filter.
Implements traditional pagination using SQL LIMIT and OFFSET clauses.
Only applies to SELECT statements; other statement types are returned unmodified.
Note:
This filter only modifies SELECT statements. For other statement types
(UPDATE, DELETE), the statement is returned unchanged.
See Also:
- :meth:`sqlalchemy.sql.expression.Select.limit`: SQLAlchemy LIMIT clause
- :meth:`sqlalchemy.sql.expression.Select.offset`: SQLAlchemy OFFSET clause
"""
limit: int
"""Maximum number of rows to return."""
offset: int
"""Number of rows to skip before returning results."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply LIMIT/OFFSET pagination to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with limit and offset applied
Note:
Only modifies SELECT statements. Other statement types are returned as-is.
See Also:
:class:`sqlalchemy.sql.expression.Select`: SQLAlchemy SELECT statement
"""
if isinstance(statement, Select):
statement = cast("StatementTypeT", statement.limit(self.limit).offset(self.offset))
return statement
[docs]
@dataclass
class OrderBy(StatementFilter):
"""Order by a specific field.
Appends an ORDER BY clause to SELECT statements, sorting records by the
specified field in ascending or descending order.
Note:
This filter only modifies SELECT statements. For other statement types,
the statement is returned unchanged.
Note:
``nulls`` controls where NULLs land. Left as ``None`` the database decides, and the
databases disagree: PostgreSQL and Oracle sort NULLs last ascending and first descending,
while MySQL and SQLite do the opposite. Set it explicitly for a stable order across
backends — ``nulls="last"`` is usually what a user reading a descending list expects, since
rows with no value otherwise fill the first page.
MySQL, MariaDB and SQL Server have no ``NULLS FIRST``/``NULLS LAST`` syntax. They sort
NULL as the lowest value, so ``asc`` with ``nulls="first"`` and ``desc`` with
``nulls="last"`` compile to the plain ordering term. The other two placements are emulated
with a leading ``CASE WHEN ... IS NULL`` sort key. SQL Server requires every ORDER BY term
to appear in the select list under ``SELECT DISTINCT``, so combine ``nulls`` with a
``DISTINCT`` statement there only for those two native placements.
See Also:
- :meth:`sqlalchemy.sql.expression.Select.order_by`: SQLAlchemy ORDER BY clause
- :meth:`sqlalchemy.sql.expression.ColumnElement.asc`: Ascending order
- :meth:`sqlalchemy.sql.expression.ColumnElement.desc`: Descending order
- :func:`sqlalchemy.sql.expression.nulls_first`: NULLS FIRST
- :func:`sqlalchemy.sql.expression.nulls_last`: NULLS LAST
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression (e.g., ``func.random()``)."""
sort_order: Literal["asc", "desc"] = "asc"
"""Sort direction ("asc" or "desc")."""
nulls: Optional[Literal["first", "last"]] = None
"""Where NULLs sort. ``None`` leaves the placement to the database."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Append an ORDER BY clause to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with an ORDER BY clause
Note:
Only modifies SELECT statements. Other statement types are returned as-is.
See Also:
:meth:`sqlalchemy.sql.expression.Select.order_by`: SQLAlchemy ORDER BY
"""
if isinstance(statement, Select):
field = self._get_instrumented_attr(model, self.field_name)
ordering: ColumnElement[Any] = field.desc() if self.sort_order == "desc" else field.asc()
if self.nulls is not None:
ordering = _NullsPlacement(
ordering, modifier=op.nulls_first_op if self.nulls == "first" else op.nulls_last_op
)
statement = cast("StatementTypeT", statement.order_by(ordering))
return statement
LIKE_ESCAPE_CHAR = "/"
"""Escape character used when :class:`SearchFilter` escapes LIKE wildcards.
Not a backslash: SQLAlchemy renders one into the ``ESCAPE`` clause as ``'\\\\'``, which DuckDB reads
as two characters and rejects with "Escape string must be empty or one character". ``/`` needs no
such doubling on dialects that support an explicit ``ESCAPE`` clause.
"""
[docs]
def escape_like_value(value: str, escape_char: str = LIKE_ESCAPE_CHAR) -> str:
"""Escape the LIKE wildcards ``%`` and ``_`` so they match literally.
The escape character itself is escaped first, otherwise it would consume the
escape characters added for the wildcards.
Args:
value: The raw search term.
escape_char: The character to escape with.
Returns:
str: ``value`` with ``escape_char``, ``%`` and ``_`` escaped.
"""
return value.replace(escape_char, escape_char * 2).replace("%", f"{escape_char}%").replace("_", f"{escape_char}_")
[docs]
@dataclass
class SearchFilter(StatementFilter):
"""Case-sensitive or case-insensitive substring matching filter.
Implements text search using SQL LIKE or ILIKE operators. Can search across
multiple fields using OR conditions.
Note:
The search pattern automatically adds wildcards before and after the search
value, equivalent to SQL pattern '%value%'.
Note:
``%`` and ``_`` in ``value`` remain SQL wildcards by default. Set
``escape_wildcards=True`` to match them literally on databases supporting
the SQL ``ESCAPE`` clause. Spanner has no ``ESCAPE`` clause, so
``escape_wildcards`` must stay ``False`` there.
See Also:
- :class:`.NotInSearchFilter`: Opposite filter using NOT LIKE/ILIKE
- :meth:`sqlalchemy.sql.expression.ColumnOperators.like`: Case-sensitive LIKE
- :meth:`sqlalchemy.sql.expression.ColumnOperators.ilike`: Case-insensitive LIKE
"""
field_name: Union[str, set[str]]
"""Name or set of names of model attributes to search on."""
value: str
"""Text to match within the field(s)."""
ignore_case: Optional[bool] = False
"""Whether to use case-insensitive matching."""
escape_wildcards: bool = False
"""Whether to treat ``%`` and ``_`` in :attr:`value` as literal characters."""
@property
def _operator(self) -> Callable[..., ColumnElement[bool]]:
"""Return the SQL operator for combining multiple search clauses.
Returns:
Callable[..., ColumnElement[bool]]: The `or_` operator for OR conditions
See Also:
:func:`sqlalchemy.sql.expression.or_`: SQLAlchemy OR operator
"""
return or_
@property
def _func(self) -> "attrgetter[Callable[..., BinaryExpression[bool]]]":
"""Return the appropriate LIKE or ILIKE operator as a function.
Returns:
attrgetter: Bound method for LIKE or ILIKE operations
See Also:
- :meth:`sqlalchemy.sql.expression.ColumnOperators.like`: LIKE operator
- :meth:`sqlalchemy.sql.expression.ColumnOperators.ilike`: ILIKE operator
"""
return attrgetter("ilike" if self.ignore_case else "like")
@property
def normalized_field_names(self) -> set[str]:
"""Convert field_name to a set if it's a single string.
Returns:
set[str]: Set of field names to be searched
"""
return {self.field_name} if isinstance(self.field_name, str) else self.field_name
[docs]
def get_search_clauses(self, model: type[ModelT]) -> list[BinaryExpression[bool]]:
"""Generate the LIKE/ILIKE clauses for all specified fields.
Args:
model: The SQLAlchemy model class
Returns:
list[BinaryExpression[bool]]: List of text matching expressions
See Also:
:class:`sqlalchemy.sql.expression.BinaryExpression`: SQLAlchemy expression
"""
search_clause: list[BinaryExpression[bool]] = []
escape_char = LIKE_ESCAPE_CHAR if self.escape_wildcards else None
value = escape_like_value(self.value) if self.escape_wildcards else self.value
for field_name in self.normalized_field_names:
try:
field = self._get_instrumented_attr(model, field_name)
search_text = f"%{value}%"
search_clause.append(self._func(field)(search_text, escape=escape_char))
except AttributeError:
msg = f"Skipping search for field {field_name}. It is not found in model {model.__name__}"
logger.debug(msg)
continue
return search_clause
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Append a LIKE/ILIKE clause to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with text search clauses
See Also:
:meth:`sqlalchemy.sql.expression.Select.where`: SQLAlchemy WHERE clause
"""
search_clauses = self.get_search_clauses(model)
if not search_clauses:
return statement
where_clause = self._operator(*search_clauses)
return cast("StatementTypeT", statement.where(where_clause))
# Regular typed dictionary for operators_map
operators_map: dict[str, Callable[[Any, Any], ColumnElement[bool]]] = {
"eq": op.eq,
"ne": op.ne,
"gt": op.gt,
"ge": op.ge,
"lt": op.lt,
"le": op.le,
"in": op.in_op,
"notin": op.notin_op,
"between": lambda c, v: c.between(v[0], v[1]),
"like": op.like_op,
"ilike": op.ilike_op,
"startswith": op.startswith_op,
"istartswith": lambda c, v: c.ilike(v + "%"),
"endswith": op.endswith_op,
"iendswith": lambda c, v: c.ilike(v + "%"),
"dateeq": lambda c, v: cast("Date", c) == v,
}
VALID_OPERATORS = set(operators_map.keys())
"""Set of valid operators that can be used in ComparisonFilter."""
[docs]
@dataclass
class ComparisonFilter(StatementFilter):
"""Simple comparison filter for equality and inequality operations.
This filter applies basic comparison operators (=, !=, >, >=, <, <=) to a field.
It provides a generic way to perform common comparison operations.
Args:
field_name: Name of the model attribute to filter on
operator: Comparison operator to use (must be one of: 'eq', 'ne', 'gt', 'ge', 'lt', 'le', 'in', 'notin', 'between', 'like', 'ilike', 'startswith', 'istartswith', 'endswith', 'iendswith', 'dateeq')
value: Value to compare against
Raises:
ValueError: If an invalid operator is provided
"""
field_name: FilterFieldName
"""Field name, model attribute, or func expression."""
operator: str
"""Comparison operator to use (one of 'eq', 'ne', 'gt', 'ge', 'lt', 'le')."""
value: Any
"""Value to compare against."""
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Apply a comparison operation to the statement.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with the comparison condition
Raises:
ValueError: If an invalid operator is provided
"""
field = self._get_instrumented_attr(model, self.field_name)
operator_func = operators_map.get(self.operator)
if operator_func is None:
msg = f"Invalid operator '{self.operator}'. Must be one of: {', '.join(sorted(VALID_OPERATORS))}"
raise ValueError(msg)
condition = operator_func(field, self.value)
return cast("StatementTypeT", statement.where(condition))
[docs]
@dataclass
class NotInSearchFilter(SearchFilter):
"""Filter for excluding records that match a substring.
Implements negative text search using SQL NOT LIKE or NOT ILIKE operators.
Can exclude across multiple fields using AND conditions.
Args:
field_name: Name or set of names of model attributes to search on
value: Text to exclude from the field(s)
ignore_case: If True, uses NOT ILIKE for case-insensitive matching
escape_wildcards: If True, matches wildcard characters literally using
SQL ESCAPE. Defaults to False; not supported by Spanner when enabled.
Note:
Uses AND for multiple fields, meaning records matching any field will be excluded.
See Also:
- :class:`.SearchFilter`: Opposite filter using LIKE/ILIKE
- :meth:`sqlalchemy.sql.expression.ColumnOperators.notlike`: NOT LIKE operator
- :meth:`sqlalchemy.sql.expression.ColumnOperators.notilike`: NOT ILIKE operator
"""
@property
def _operator(self) -> Callable[..., ColumnElement[bool]]:
"""Return the SQL operator for combining multiple negated search clauses.
Returns:
Callable[..., ColumnElement[bool]]: The `and_` operator for AND conditions
See Also:
:func:`sqlalchemy.sql.expression.and_`: SQLAlchemy AND operator
"""
return and_
@property
def _func(self) -> "attrgetter[Callable[..., BinaryExpression[bool]]]":
"""Return the appropriate NOT LIKE or NOT ILIKE operator as a function.
Returns:
attrgetter: Bound method for NOT LIKE or NOT ILIKE operations
See Also:
- :meth:`sqlalchemy.sql.expression.ColumnOperators.notlike`: NOT LIKE
- :meth:`sqlalchemy.sql.expression.ColumnOperators.notilike`: NOT ILIKE
"""
return attrgetter("not_ilike" if self.ignore_case else "not_like")
[docs]
@dataclass
class ExistsFilter(StatementFilter):
"""Filter for EXISTS subqueries.
This filter creates an EXISTS condition using a list of column expressions.
The expressions can be combined using either AND or OR logic. The filter applies
a correlated subquery that returns only the rows from the main query that match
the specified conditions.
For example, if searching movies with `Movie.genre == "Action"`, only rows where
the genre is "Action" will be returned.
Parameters
----------
values : list[ColumnElement[bool]]
values: List of SQLAlchemy column expressions to use in the EXISTS clause
operator : Literal["and", "or"], optional
operator: If "and", combines conditions with AND, otherwise uses OR. Defaults to "and".
Example:
--------
Basic usage with AND conditions::
from sqlalchemy import select
from advanced_alchemy.filters import ExistsFilter
filter = ExistsFilter(
values=[User.email.like("%@example.com%")],
)
statement = filter.append_to_statement(
select(Organization), Organization
)
This will return only organizations where the user's email contains "@example.com".
Using OR conditions::
filter = ExistsFilter(
values=[User.role == "admin", User.role == "owner"],
operator="or",
)
This will return organizations where the user's role is either "admin" OR "owner".
See Also:
--------
:class:`NotExistsFilter`: The inverse of this filter
:func:`sqlalchemy.sql.expression.exists`: SQLAlchemy EXISTS expression
"""
values: list[ColumnElement[bool]]
"""List of SQLAlchemy column expressions to use in the EXISTS clause."""
operator: Literal["and", "or"] = "and"
"""If "and", combines conditions with the AND operator, otherwise uses OR."""
@property
def _and(self) -> Callable[..., ColumnElement[bool]]:
"""Access the SQLAlchemy `and_` operator.
Returns:
Callable[..., ColumnElement[bool]]: The `and_` operator for AND conditions
See Also:
:func:`sqlalchemy.sql.expression.and_`: SQLAlchemy AND operator
"""
return and_
@property
def _or(self) -> Callable[..., ColumnElement[bool]]:
"""Access the SQLAlchemy `or_` operator.
Returns:
Callable[..., ColumnElement[bool]]: The `or_` operator for OR conditions
See Also:
:func:`sqlalchemy.sql.expression.or_`: SQLAlchemy OR operator
"""
return or_
def _get_combined_conditions(self) -> ColumnElement[bool]:
"""Combine the filter conditions using the specified operator.
Returns:
ColumnElement[bool]:
A SQLAlchemy column expression combining all conditions with AND or OR
"""
op = self._and if self.operator == "and" else self._or
return op(*self.values)
[docs]
def get_exists_clause(self, model: type[ModelT]) -> ColumnElement[bool]:
"""Generate the EXISTS clause for the statement.
Args:
model : type[ModelT]
The SQLAlchemy model class to correlate with
Returns:
ColumnElement[bool]:
A correlated EXISTS expression for use in a WHERE clause
"""
# Handle empty values list case
if not self.values:
# Use explicitly imported 'false' from sqlalchemy
# Return SQLAlchemy FALSE expression
return false()
# Combine all values with AND or OR (using the operator specified in the filter)
# This creates a single boolean expression from multiple conditions
combined_conditions = self._get_combined_conditions()
# Create a correlated subquery with the combined conditions
try:
subquery = select(1).where(combined_conditions)
correlated_subquery = subquery.correlate(model.__table__)
return exists(correlated_subquery)
except Exception: # noqa: BLE001
return false()
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Append EXISTS condition to the statement.
Args:
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
Returns:
StatementTypeT:
Modified statement with EXISTS condition
"""
# We apply the exists clause regardless of whether self.values is empty,
# as get_exists_clause handles the empty case by returning false().
exists_clause = self.get_exists_clause(model)
return cast("StatementTypeT", statement.where(exists_clause))
[docs]
@dataclass
class NotExistsFilter(StatementFilter):
"""Filter for NOT EXISTS subqueries.
This filter creates a NOT EXISTS condition using a list of column expressions.
The expressions can be combined using either AND or OR logic. The filter applies
a correlated subquery that returns only the rows from the main query that DO NOT
match the specified conditions.
For example, if searching movies with `Movie.genre == "Action"`, only rows where
the genre is NOT "Action" will be returned.
Parameters
----------
values : list[ColumnElement[bool]]
values: List of SQLAlchemy column expressions to use in the NOT EXISTS clause
operator : Literal["and", "or"], optional
operator: If "and", combines conditions with AND, otherwise uses OR. Defaults to "and".
Example:
--------
Basic usage with AND conditions::
from sqlalchemy import select
from advanced_alchemy.filters import NotExistsFilter
filter = NotExistsFilter(
values=[User.email.like("%@example.com%")],
)
statement = filter.append_to_statement(
select(Organization), Organization
)
This will return only organizations where the user's email does NOT contain "@example.com".
Using OR conditions::
filter = NotExistsFilter(
values=[User.role == "admin", User.role == "owner"],
operator="or",
)
This will return organizations where the user's role is NEITHER "admin" NOR "owner".
See Also:
--------
:class:`ExistsFilter`: The inverse of this filter
:func:`sqlalchemy.sql.expression.not_`: SQLAlchemy NOT operator
:func:`sqlalchemy.sql.expression.exists`: SQLAlchemy EXISTS expression
"""
values: list[ColumnElement[bool]]
"""List of SQLAlchemy column expressions to use in the NOT EXISTS clause."""
operator: Literal["and", "or"] = "and"
"""If "and", combines conditions with the AND operator, otherwise uses OR."""
@property
def _and(self) -> Callable[..., ColumnElement[bool]]:
"""Access the SQLAlchemy `and_` operator.
Returns:
Callable[..., ColumnElement[bool]]: The `and_` operator for AND conditions
See Also:
:func:`sqlalchemy.sql.expression.and_`: SQLAlchemy AND operator
"""
return and_
@property
def _or(self) -> Callable[..., ColumnElement[bool]]:
"""Access the SQLAlchemy `or_` operator.
Returns:
Callable[..., ColumnElement[bool]]: The `or_` operator for OR conditions
See Also:
:func:`sqlalchemy.sql.expression.or_`: SQLAlchemy OR operator
"""
return or_
def _get_combined_conditions(self) -> ColumnElement[bool]:
op = self._and if self.operator == "and" else self._or
return op(*self.values)
[docs]
def get_exists_clause(self, model: type[ModelT]) -> ColumnElement[bool]:
"""Generate the NOT EXISTS clause for the statement.
Args:
model : type[ModelT]
The SQLAlchemy model class to correlate with
Returns:
ColumnElement[bool]:
A correlated NOT EXISTS expression for use in a WHERE clause
"""
# Handle empty values list case
if not self.values:
# Return SQLAlchemy TRUE expression
return true()
# Combine conditions and create correlated subquery
combined_conditions = self._get_combined_conditions()
subquery = select(1).where(combined_conditions)
correlated_subquery = subquery.correlate(model.__table__)
return not_(exists(correlated_subquery))
[docs]
def append_to_statement(self, statement: StatementTypeT, model: type[ModelT]) -> StatementTypeT:
"""Append NOT EXISTS condition to the statement.
Args:
statement : StatementTypeT
The SQLAlchemy statement to modify
model : type[ModelT]
The SQLAlchemy model class
Returns:
StatementTypeT:
Modified statement with NOT EXISTS condition
"""
# We apply the exists clause regardless of whether self.values is empty,
# as get_exists_clause handles the empty case by returning true.
exists_clause = self.get_exists_clause(model)
return cast("StatementTypeT", statement.where(exists_clause))
[docs]
@dataclass
class FilterGroup(StatementFilter):
"""A group of filters combined with a logical operator.
This class combines multiple filters with a logical operator (AND/OR).
It provides a way to create complex nested filter conditions.
"""
logical_operator: Callable[..., ColumnElement[bool]]
"""Logical operator to combine the filters."""
filters: list[StatementFilter]
"""List of filters to combine."""
[docs]
def append_to_statement(
self,
statement: StatementTypeT,
model: type[ModelT],
) -> "StatementTypeT":
"""Apply all filters combined with the logical operator.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with combined filters
"""
if not self.filters:
return statement
# Create a list of expressions from each filter
expressions = []
for filter_obj in self.filters:
# Each filter needs to be applied to a clean version of the statement
# to get just its expression
filter_statement = filter_obj.append_to_statement(select(), model)
# Extract the whereclause from the filter's statement
if hasattr(filter_statement, "whereclause") and filter_statement.whereclause is not None:
expressions.append(filter_statement.whereclause) # pyright: ignore
if expressions:
# Combine all expressions with the logical operator
combined = self.logical_operator(*expressions)
return cast("StatementTypeT", statement.where(combined))
return statement
[docs]
@dataclass
class MultiFilter(StatementFilter):
"""Apply multiple filters to a query based on a JSON/dict input.
This filter provides a way to construct complex filter trees from
a structured dictionary input, supporting nested logical groups and
various filter types.
"""
filters: dict[str, Any]
"""JSON/dict structure representing the filters."""
# TypedDict class variables
_filter_map: ClassVar[FilterMap] = {
"before_after": BeforeAfter,
"on_before_after": OnBeforeAfter,
"boolean": BooleanFilter,
"choices": ChoicesFilter,
"collection": CollectionFilter,
"not_in_collection": NotInCollectionFilter,
"limit_offset": LimitOffset,
"null": NullFilter,
"not_null": NotNullFilter,
"order_by": OrderBy,
"search": SearchFilter,
"not_in_search": NotInSearchFilter,
"filter_group": FilterGroup,
"comparison": ComparisonFilter,
"exists": ExistsFilter,
"not_exists": NotExistsFilter,
}
_logical_map: ClassVar[LogicalOperatorMap] = {
"and_": and_,
"or_": or_,
}
[docs]
def append_to_statement(
self,
statement: StatementTypeT,
model: type[ModelT],
) -> StatementTypeT:
"""Apply the filters to the statement based on the filter definitions.
Args:
statement: The SQLAlchemy statement to modify
model: The SQLAlchemy model class
Returns:
StatementTypeT: Modified statement with all filters applied
"""
for filter_type, conditions in self.filters.items():
operator = self._logical_map.get(filter_type)
if operator and isinstance(conditions, list):
# Create filters from the conditions
valid_filters = []
for cond in conditions: # pyright: ignore
filter_instance = self._create_filter(cond) # pyright: ignore
if filter_instance is not None:
valid_filters.append(filter_instance) # pyright: ignore
# Only create a filter group if we have valid filters
if valid_filters:
filter_group = FilterGroup(
logical_operator=operator, # type: ignore
filters=valid_filters, # pyright: ignore
)
statement = filter_group.append_to_statement(statement, model)
return statement
def _create_filter(self, condition: dict[str, Any]) -> Optional[StatementFilter]:
"""Create a filter instance from a condition dictionary.
Args:
condition: Dictionary defining a filter
Returns:
Optional[StatementFilter]: Filter instance if successfully created, None otherwise
"""
# Check if condition is a nested logical group
logical_keys = set(self._logical_map.keys())
intersect = logical_keys.intersection(condition.keys())
if intersect:
# It's a nested filter group
for key in intersect:
operator = self._logical_map.get(key)
if operator and isinstance(condition.get(key), list):
nested_filters = []
for cond in condition[key]:
filter_instance = self._create_filter(cond)
if filter_instance is not None:
nested_filters.append(filter_instance) # pyright: ignore
if nested_filters:
return FilterGroup(logical_operator=operator, filters=nested_filters) # type: ignore
else:
# Regular filter
filter_type = condition.get("type")
if filter_type is not None and isinstance(filter_type, str):
filter_class = self._filter_map.get(filter_type)
if filter_class is not None:
try:
# Create a copy of the condition without the type key
filter_args = {k: v for k, v in condition.items() if k != "type"}
return filter_class(**filter_args) # type: ignore
except Exception: # noqa: BLE001
return None
return None
# Define FilterTypes using direct class references
FilterTypes: TypeAlias = Union[
BeforeAfter,
OnBeforeAfter,
BooleanFilter,
ChoicesFilter[Any],
CollectionFilter[Any],
LimitOffset,
NullFilter,
NotNullFilter,
OrderBy,
SearchFilter,
NotInCollectionFilter[Any],
NotInSearchFilter,
ExistsFilter,
NotExistsFilter,
ComparisonFilter,
MultiFilter,
FilterGroup,
]
"""Aggregate type alias of the types supported for collection filtering."""
class _NullsPlacement(UnaryExpression[Any]):
"""An ORDER BY term that pins where NULLs sort.
MySQL, MariaDB and SQL Server have no ``NULLS FIRST``/``NULLS LAST`` syntax. Those backends
sort NULL as the lowest value, so ascending-first and descending-last are already native
and compile to the plain ordering term; the other two placements are emulated with a leading
nullity key. Every other backend gets the native clause, which an index on the column can
still satisfy.
UnaryExpression preserves ordering traversal for ORM eager loading and includes
the column, direction and NULL placement in SQL compilation cache keys.
"""
inherit_cache = True
@compiles(_NullsPlacement, "mysql")
@compiles(_NullsPlacement, "mariadb")
@compiles(_NullsPlacement, "mssql")
def _compile_nulls_placement_emulated( # pyright: ignore[reportUnusedFunction]
element: _NullsPlacement, compiler: Any, **kw: Any
) -> str:
ordering = cast("UnaryExpression[Any]", element.element)
nulls_go_last = element.modifier is op.nulls_last_op
if nulls_go_last is (ordering.modifier is op.desc_op):
return cast("str", compiler.process(ordering, **kw))
# Literals, not bound parameters: a placeholder inside ORDER BY is ambiguous on these backends.
key = case(
(ordering.element.is_(None), literal_column("1" if nulls_go_last else "0", Integer)),
else_=literal_column("0" if nulls_go_last else "1", Integer),
)
# SQL Server permits a SELECT alias as an ordering term, but not inside CASE.
key_kw = {**kw, "render_label_as_label": None}
return f"{compiler.process(key, **key_kw)}, {compiler.process(ordering, **kw)}"
