Evaluate/projects/iceberg/python/pyiceberg/expressions/visitors.py

# Licensed to the Apache Software Foundation (ASF) under one
# or more contributor license agreements.  See the NOTICE file
# distributed with this work for additional information
# regarding copyright ownership.  The ASF licenses this file
# to you under the Apache License, Version 2.0 (the
# "License"); you may not use this file except in compliance
# with the License.  You may obtain a copy of the License at
#
#   http://www.apache.org/licenses/LICENSE-2.0
#
# Unless required by applicable law or agreed to in writing,
# software distributed under the License is distributed on an
# "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
# KIND, either express or implied.  See the License for the
# specific language governing permissions and limitations
# under the License.
import math
from abc import ABC, abstractmethod
from functools import singledispatch
from typing import (
    Any,
    Callable,
    Dict,
    Generic,
    List,
    Set,
    Tuple,
    TypeVar,
    Union,
)

from pyiceberg.conversions import from_bytes
from pyiceberg.expressions import (
    AlwaysFalse,
    AlwaysTrue,
    And,
    BooleanExpression,
    BoundEqualTo,
    BoundGreaterThan,
    BoundGreaterThanOrEqual,
    BoundIn,
    BoundIsNaN,
    BoundIsNull,
    BoundLessThan,
    BoundLessThanOrEqual,
    BoundLiteralPredicate,
    BoundNotEqualTo,
    BoundNotIn,
    BoundNotNaN,
    BoundNotNull,
    BoundNotStartsWith,
    BoundPredicate,
    BoundSetPredicate,
    BoundStartsWith,
    BoundTerm,
    BoundUnaryPredicate,
    Not,
    Or,
    UnboundPredicate,
)
from pyiceberg.expressions.literals import Literal
from pyiceberg.manifest import DataFile, ManifestFile, PartitionFieldSummary
from pyiceberg.partitioning import PartitionSpec
from pyiceberg.schema import Schema
from pyiceberg.typedef import EMPTY_DICT, L, StructProtocol
from pyiceberg.types import (
    DoubleType,
    FloatType,
    IcebergType,
    NestedField,
    PrimitiveType,
    StructType,
    TimestampType,
    TimestamptzType,
)
from pyiceberg.utils.datetime import micros_to_timestamp, micros_to_timestamptz

T = TypeVar("T")


class BooleanExpressionVisitor(Generic[T], ABC):
    @abstractmethod
    def visit_true(self) -> T:
        """Visit method for an AlwaysTrue boolean expression.

        Note: This visit method has no arguments since AlwaysTrue instances have no context.
        """

    @abstractmethod
    def visit_false(self) -> T:
        """Visit method for an AlwaysFalse boolean expression.

        Note: This visit method has no arguments since AlwaysFalse instances have no context.
        """

    @abstractmethod
    def visit_not(self, child_result: T) -> T:
        """Visit method for a Not boolean expression.

        Args:
            child_result (T): The result of visiting the child of the Not boolean expression.
        """

    @abstractmethod
    def visit_and(self, left_result: T, right_result: T) -> T:
        """Visit method for an And boolean expression.

        Args:
            left_result (T): The result of visiting the left side of the expression.
            right_result (T): The result of visiting the right side of the expression.
        """

    @abstractmethod
    def visit_or(self, left_result: T, right_result: T) -> T:
        """Visit method for an Or boolean expression.

        Args:
            left_result (T): The result of visiting the left side of the expression.
            right_result (T): The result of visiting the right side of the expression.
        """

    @abstractmethod
    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> T:
        """Visit method for an unbound predicate in an expression tree.

        Args:
            predicate (UnboundPredicate[L): An instance of an UnboundPredicate.
        """

    @abstractmethod
    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> T:
        """Visit method for a bound predicate in an expression tree.

        Args:
            predicate (BoundPredicate[L]): An instance of a BoundPredicate.
        """


@singledispatch
def visit(obj: BooleanExpression, visitor: BooleanExpressionVisitor[T]) -> T:
    """Apply a boolean expression visitor to any point within an expression.

    The function traverses the expression in post-order fashion.

    Args:
        obj (BooleanExpression): An instance of a BooleanExpression.
        visitor (BooleanExpressionVisitor[T]): An instance of an implementation of the generic BooleanExpressionVisitor base class.

    Raises:
        NotImplementedError: If attempting to visit an unsupported expression.
    """
    raise NotImplementedError(f"Cannot visit unsupported expression: {obj}")


@visit.register(AlwaysTrue)
def _(_: AlwaysTrue, visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit an AlwaysTrue boolean expression with a concrete BooleanExpressionVisitor."""
    return visitor.visit_true()


@visit.register(AlwaysFalse)
def _(_: AlwaysFalse, visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit an AlwaysFalse boolean expression with a concrete BooleanExpressionVisitor."""
    return visitor.visit_false()


@visit.register(Not)
def _(obj: Not, visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit a Not boolean expression with a concrete BooleanExpressionVisitor."""
    child_result: T = visit(obj.child, visitor=visitor)
    return visitor.visit_not(child_result=child_result)


@visit.register(And)
def _(obj: And, visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit an And boolean expression with a concrete BooleanExpressionVisitor."""
    left_result: T = visit(obj.left, visitor=visitor)
    right_result: T = visit(obj.right, visitor=visitor)
    return visitor.visit_and(left_result=left_result, right_result=right_result)


@visit.register(UnboundPredicate)
def _(obj: UnboundPredicate[L], visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit an unbound boolean expression with a concrete BooleanExpressionVisitor."""
    return visitor.visit_unbound_predicate(predicate=obj)


@visit.register(BoundPredicate)
def _(obj: BoundPredicate[L], visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit a bound boolean expression with a concrete BooleanExpressionVisitor."""
    return visitor.visit_bound_predicate(predicate=obj)


@visit.register(Or)
def _(obj: Or, visitor: BooleanExpressionVisitor[T]) -> T:
    """Visit an Or boolean expression with a concrete BooleanExpressionVisitor."""
    left_result: T = visit(obj.left, visitor=visitor)
    right_result: T = visit(obj.right, visitor=visitor)
    return visitor.visit_or(left_result=left_result, right_result=right_result)


def bind(schema: Schema, expression: BooleanExpression, case_sensitive: bool) -> BooleanExpression:
    """Travers over an expression to bind the predicates to the schema.

    Args:
      schema (Schema): A schema to use when binding the expression.
      expression (BooleanExpression): An expression containing UnboundPredicates that can be bound.
      case_sensitive (bool): Whether to consider case when binding a reference to a field in a schema, defaults to True.

    Raises:
        TypeError: In the case a predicate is already bound.
    """
    return visit(expression, BindVisitor(schema, case_sensitive))


class BindVisitor(BooleanExpressionVisitor[BooleanExpression]):
    """Rewrites a boolean expression by replacing unbound references with references to fields in a struct schema.

    Args:
      schema (Schema): A schema to use when binding the expression.
      case_sensitive (bool): Whether to consider case when binding a reference to a field in a schema, defaults to True.

    Raises:
        TypeError: In the case a predicate is already bound.
    """

    schema: Schema
    case_sensitive: bool

    def __init__(self, schema: Schema, case_sensitive: bool) -> None:
        self.schema = schema
        self.case_sensitive = case_sensitive

    def visit_true(self) -> BooleanExpression:
        return AlwaysTrue()

    def visit_false(self) -> BooleanExpression:
        return AlwaysFalse()

    def visit_not(self, child_result: BooleanExpression) -> BooleanExpression:
        return Not(child=child_result)

    def visit_and(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return And(left=left_result, right=right_result)

    def visit_or(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return Or(left=left_result, right=right_result)

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> BooleanExpression:
        return predicate.bind(self.schema, case_sensitive=self.case_sensitive)

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> BooleanExpression:
        raise TypeError(f"Found already bound predicate: {predicate}")


class BoundBooleanExpressionVisitor(BooleanExpressionVisitor[T], ABC):
    @abstractmethod
    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> T:
        """Visit a bound In predicate."""

    @abstractmethod
    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> T:
        """Visit a bound NotIn predicate."""

    @abstractmethod
    def visit_is_nan(self, term: BoundTerm[L]) -> T:
        """Visit a bound IsNan predicate."""

    @abstractmethod
    def visit_not_nan(self, term: BoundTerm[L]) -> T:
        """Visit a bound NotNan predicate."""

    @abstractmethod
    def visit_is_null(self, term: BoundTerm[L]) -> T:
        """Visit a bound IsNull predicate."""

    @abstractmethod
    def visit_not_null(self, term: BoundTerm[L]) -> T:
        """Visit a bound NotNull predicate."""

    @abstractmethod
    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound Equal predicate."""

    @abstractmethod
    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound NotEqual predicate."""

    @abstractmethod
    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound GreaterThanOrEqual predicate."""

    @abstractmethod
    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound GreaterThan predicate."""

    @abstractmethod
    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound LessThan predicate."""

    @abstractmethod
    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit a bound LessThanOrEqual predicate."""

    @abstractmethod
    def visit_true(self) -> T:
        """Visit a bound True predicate."""

    @abstractmethod
    def visit_false(self) -> T:
        """Visit a bound False predicate."""

    @abstractmethod
    def visit_not(self, child_result: T) -> T:
        """Visit a bound Not predicate."""

    @abstractmethod
    def visit_and(self, left_result: T, right_result: T) -> T:
        """Visit a bound And predicate."""

    @abstractmethod
    def visit_or(self, left_result: T, right_result: T) -> T:
        """Visit a bound Or predicate."""

    @abstractmethod
    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit bound StartsWith predicate."""

    @abstractmethod
    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> T:
        """Visit bound NotStartsWith predicate."""

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> T:
        """Visit an unbound predicate.

        Args:
            predicate (UnboundPredicate[L]): An unbound predicate.
        Raises:
            TypeError: This always raises since an unbound predicate is not expected in a bound boolean expression.
        """
        raise TypeError(f"Not a bound predicate: {predicate}")

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> T:
        """Visit a bound predicate.

        Args:
            predicate (BoundPredicate[L]): A bound predicate.
        """
        return visit_bound_predicate(predicate, self)


@singledispatch
def visit_bound_predicate(expr: BoundPredicate[L], _: BooleanExpressionVisitor[T]) -> T:
    raise TypeError(f"Unknown predicate: {expr}")


@visit_bound_predicate.register(BoundIn)
def _(expr: BoundIn[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_in(term=expr.term, literals=expr.value_set)


@visit_bound_predicate.register(BoundNotIn)
def _(expr: BoundNotIn[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_not_in(term=expr.term, literals=expr.value_set)


@visit_bound_predicate.register(BoundIsNaN)
def _(expr: BoundIsNaN[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_is_nan(term=expr.term)


@visit_bound_predicate.register(BoundNotNaN)
def _(expr: BoundNotNaN[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_not_nan(term=expr.term)


@visit_bound_predicate.register(BoundIsNull)
def _(expr: BoundIsNull[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_is_null(term=expr.term)


@visit_bound_predicate.register(BoundNotNull)
def _(expr: BoundNotNull[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_not_null(term=expr.term)


@visit_bound_predicate.register(BoundEqualTo)
def _(expr: BoundEqualTo[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_equal(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundNotEqualTo)
def _(expr: BoundNotEqualTo[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_not_equal(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundGreaterThanOrEqual)
def _(expr: BoundGreaterThanOrEqual[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    """Visit a bound GreaterThanOrEqual predicate."""
    return visitor.visit_greater_than_or_equal(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundGreaterThan)
def _(expr: BoundGreaterThan[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_greater_than(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundLessThan)
def _(expr: BoundLessThan[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_less_than(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundLessThanOrEqual)
def _(expr: BoundLessThanOrEqual[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_less_than_or_equal(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundStartsWith)
def _(expr: BoundStartsWith[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_starts_with(term=expr.term, literal=expr.literal)


@visit_bound_predicate.register(BoundNotStartsWith)
def _(expr: BoundNotStartsWith[L], visitor: BoundBooleanExpressionVisitor[T]) -> T:
    return visitor.visit_not_starts_with(term=expr.term, literal=expr.literal)


def rewrite_not(expr: BooleanExpression) -> BooleanExpression:
    return visit(expr, _RewriteNotVisitor())


class _RewriteNotVisitor(BooleanExpressionVisitor[BooleanExpression]):
    """Inverts the negations."""

    def visit_true(self) -> BooleanExpression:
        return AlwaysTrue()

    def visit_false(self) -> BooleanExpression:
        return AlwaysFalse()

    def visit_not(self, child_result: BooleanExpression) -> BooleanExpression:
        return ~child_result

    def visit_and(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return And(left=left_result, right=right_result)

    def visit_or(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return Or(left=left_result, right=right_result)

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> BooleanExpression:
        return predicate

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> BooleanExpression:
        return predicate


def expression_evaluator(schema: Schema, unbound: BooleanExpression, case_sensitive: bool) -> Callable[[StructProtocol], bool]:
    return _ExpressionEvaluator(schema, unbound, case_sensitive).eval


class _ExpressionEvaluator(BoundBooleanExpressionVisitor[bool]):
    bound: BooleanExpression
    struct: StructProtocol

    def __init__(self, schema: Schema, unbound: BooleanExpression, case_sensitive: bool):
        self.bound = bind(schema, unbound, case_sensitive)

    def eval(self, struct: StructProtocol) -> bool:
        self.struct = struct
        return visit(self.bound, self)

    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        return term.eval(self.struct) in literals

    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        return term.eval(self.struct) not in literals

    def visit_is_nan(self, term: BoundTerm[L]) -> bool:
        val = term.eval(self.struct)
        return val != val

    def visit_not_nan(self, term: BoundTerm[L]) -> bool:
        val = term.eval(self.struct)
        return val == val

    def visit_is_null(self, term: BoundTerm[L]) -> bool:
        return term.eval(self.struct) is None

    def visit_not_null(self, term: BoundTerm[L]) -> bool:
        return term.eval(self.struct) is not None

    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return term.eval(self.struct) == literal.value

    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return term.eval(self.struct) != literal.value

    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        value = term.eval(self.struct)
        return value is not None and value >= literal.value

    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        value = term.eval(self.struct)
        return value is not None and value > literal.value

    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        value = term.eval(self.struct)
        return value is not None and value < literal.value

    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        value = term.eval(self.struct)
        return value is not None and value <= literal.value

    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        eval_res = term.eval(self.struct)
        return eval_res is not None and str(eval_res).startswith(str(literal.value))

    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return not self.visit_starts_with(term, literal)

    def visit_true(self) -> bool:
        return True

    def visit_false(self) -> bool:
        return False

    def visit_not(self, child_result: bool) -> bool:
        return not child_result

    def visit_and(self, left_result: bool, right_result: bool) -> bool:
        return left_result and right_result

    def visit_or(self, left_result: bool, right_result: bool) -> bool:
        return left_result or right_result


ROWS_MIGHT_MATCH = True
ROWS_MUST_MATCH = True
ROWS_CANNOT_MATCH = False
ROWS_MIGHT_NOT_MATCH = False
IN_PREDICATE_LIMIT = 200


def _from_byte_buffer(field_type: IcebergType, val: bytes) -> Any:
    if not isinstance(field_type, PrimitiveType):
        raise ValueError(f"Expected a PrimitiveType, got: {type(field_type)}")
    return from_bytes(field_type, val)


class _ManifestEvalVisitor(BoundBooleanExpressionVisitor[bool]):
    partition_fields: List[PartitionFieldSummary]
    partition_filter: BooleanExpression

    def __init__(self, partition_struct_schema: Schema, partition_filter: BooleanExpression, case_sensitive: bool) -> None:
        self.partition_filter = bind(partition_struct_schema, rewrite_not(partition_filter), case_sensitive)

    def eval(self, manifest: ManifestFile) -> bool:
        if partitions := manifest.partitions:
            self.partition_fields = partitions
            return visit(self.partition_filter, self)

        # No partition information
        return ROWS_MIGHT_MATCH

    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.lower_bound is None:
            return ROWS_CANNOT_MATCH

        if len(literals) > IN_PREDICATE_LIMIT:
            return ROWS_MIGHT_MATCH

        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)

        if all(lower > val for val in literals):
            return ROWS_CANNOT_MATCH

        if field.upper_bound is not None:
            upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)
            if all(upper < val for val in literals):
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        # because the bounds are not necessarily a min or max value, this cannot be answered using
        # them. notIn(col, {X, ...}) with (X, Y) doesn't guarantee that X is a value in col.
        return ROWS_MIGHT_MATCH

    def visit_is_nan(self, term: BoundTerm[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.contains_nan is False:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_nan(self, term: BoundTerm[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.contains_nan is True and field.contains_null is False and field.lower_bound is None:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_is_null(self, term: BoundTerm[L]) -> bool:
        pos = term.ref().accessor.position

        if self.partition_fields[pos].contains_null is False:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_null(self, term: BoundTerm[L]) -> bool:
        pos = term.ref().accessor.position

        # contains_null encodes whether at least one partition value is null,
        # lowerBound is null if all partition values are null
        all_null = self.partition_fields[pos].contains_null is True and self.partition_fields[pos].lower_bound is None

        if all_null and isinstance(term.ref().field.field_type, (DoubleType, FloatType)):
            # floating point types may include NaN values, which we check separately.
            # In case bounds don't include NaN value, contains_nan needs to be checked against.
            all_null = self.partition_fields[pos].contains_nan is False

        if all_null:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.lower_bound is None or field.upper_bound is None:
            # values are all null and literal cannot contain null
            return ROWS_CANNOT_MATCH

        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)

        if lower > literal.value:
            return ROWS_CANNOT_MATCH

        upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)

        if literal.value > upper:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # because the bounds are not necessarily a min or max value, this cannot be answered using
        # them. notEq(col, X) with (X, Y) doesn't guarantee that X is a value in col.
        return ROWS_MIGHT_MATCH

    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.upper_bound is None:
            return ROWS_CANNOT_MATCH

        upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)

        if literal.value > upper:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.upper_bound is None:
            return ROWS_CANNOT_MATCH

        upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)

        if literal.value >= upper:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.lower_bound is None:
            return ROWS_CANNOT_MATCH

        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)

        if literal.value <= lower:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]

        if field.lower_bound is None:
            return ROWS_CANNOT_MATCH

        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)

        if literal.value < lower:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]
        prefix = str(literal.value)
        len_prefix = len(prefix)

        if field.lower_bound is None:
            return ROWS_CANNOT_MATCH

        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)
        # truncate lower bound so that its length is not greater than the length of prefix
        if lower is not None and lower[:len_prefix] > prefix:
            return ROWS_CANNOT_MATCH

        if field.upper_bound is None:
            return ROWS_CANNOT_MATCH

        upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)
        # truncate upper bound so that its length is not greater than the length of prefix
        if upper is not None and upper[:len_prefix] < prefix:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        pos = term.ref().accessor.position
        field = self.partition_fields[pos]
        prefix = str(literal.value)
        len_prefix = len(prefix)

        if field.contains_null or field.lower_bound is None or field.upper_bound is None:
            return ROWS_MIGHT_MATCH

        # not_starts_with will match unless all values must start with the prefix. This happens when
        # the lower and upper bounds both start with the prefix.
        lower = _from_byte_buffer(term.ref().field.field_type, field.lower_bound)
        upper = _from_byte_buffer(term.ref().field.field_type, field.upper_bound)

        if lower is not None and upper is not None:
            # if lower is shorter than the prefix then lower doesn't start with the prefix
            if len(lower) < len_prefix:
                return ROWS_MIGHT_MATCH

            if lower[:len_prefix] == prefix:
                # if upper is shorter than the prefix then upper can't start with the prefix
                if len(upper) < len_prefix:
                    return ROWS_MIGHT_MATCH

                if upper[:len_prefix] == prefix:
                    return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_true(self) -> bool:
        return ROWS_MIGHT_MATCH

    def visit_false(self) -> bool:
        return ROWS_CANNOT_MATCH

    def visit_not(self, child_result: bool) -> bool:
        return not child_result

    def visit_and(self, left_result: bool, right_result: bool) -> bool:
        return left_result and right_result

    def visit_or(self, left_result: bool, right_result: bool) -> bool:
        return left_result or right_result


def manifest_evaluator(
    partition_spec: PartitionSpec, schema: Schema, partition_filter: BooleanExpression, case_sensitive: bool = True
) -> Callable[[ManifestFile], bool]:
    partition_type = partition_spec.partition_type(schema)
    partition_schema = Schema(*partition_type.fields)
    evaluator = _ManifestEvalVisitor(partition_schema, partition_filter, case_sensitive)
    return evaluator.eval


class ProjectionEvaluator(BooleanExpressionVisitor[BooleanExpression], ABC):
    schema: Schema
    spec: PartitionSpec
    case_sensitive: bool

    def __init__(self, schema: Schema, spec: PartitionSpec, case_sensitive: bool):
        self.schema = schema
        self.spec = spec
        self.case_sensitive = case_sensitive

    def project(self, expr: BooleanExpression) -> BooleanExpression:
        #  projections assume that there are no NOT nodes in the expression tree. to ensure that this
        #  is the case, the expression is rewritten to push all NOT nodes down to the expression
        #  leaf nodes.
        #  this is necessary to ensure that the default expression returned when a predicate can't be
        #  projected is correct.
        return visit(bind(self.schema, rewrite_not(expr), self.case_sensitive), self)

    def visit_true(self) -> BooleanExpression:
        return AlwaysTrue()

    def visit_false(self) -> BooleanExpression:
        return AlwaysFalse()

    def visit_not(self, child_result: BooleanExpression) -> BooleanExpression:
        raise ValueError(f"Cannot project not expression, should be rewritten: {child_result}")

    def visit_and(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return And(left_result, right_result)

    def visit_or(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return Or(left_result, right_result)

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> BooleanExpression:
        raise ValueError(f"Cannot project unbound predicate: {predicate}")


class InclusiveProjection(ProjectionEvaluator):
    def visit_bound_predicate(self, predicate: BoundPredicate[Any]) -> BooleanExpression:
        parts = self.spec.fields_by_source_id(predicate.term.ref().field.field_id)

        result: BooleanExpression = AlwaysTrue()
        for part in parts:
            # consider (d = 2019-01-01) with bucket(7, d) and bucket(5, d)
            # projections: b1 = bucket(7, '2019-01-01') = 5, b2 = bucket(5, '2019-01-01') = 0
            # any value where b1 != 5 or any value where b2 != 0 cannot be the '2019-01-01'
            #
            # similarly, if partitioning by day(ts) and hour(ts), the more restrictive
            # projection should be used. ts = 2019-01-01T01:00:00 produces day=2019-01-01 and
            # hour=2019-01-01-01. the value will be in 2019-01-01-01 and not in 2019-01-01-02.
            incl_projection = part.transform.project(name=part.name, pred=predicate)
            if incl_projection is not None:
                result = And(result, incl_projection)

        return result


def inclusive_projection(
    schema: Schema, spec: PartitionSpec, case_sensitive: bool = True
) -> Callable[[BooleanExpression], BooleanExpression]:
    return InclusiveProjection(schema, spec, case_sensitive).project


class _ColumnNameTranslator(BooleanExpressionVisitor[BooleanExpression]):
    """Converts the column names with the ones in the actual file.

    Args:
      file_schema (Schema): The schema of the file.
      case_sensitive (bool): Whether to consider case when binding a reference to a field in a schema, defaults to True.

    Raises:
        TypeError: In the case of an UnboundPredicate.
        ValueError: When a column name cannot be found.
    """

    file_schema: Schema
    case_sensitive: bool

    def __init__(self, file_schema: Schema, case_sensitive: bool) -> None:
        self.file_schema = file_schema
        self.case_sensitive = case_sensitive

    def visit_true(self) -> BooleanExpression:
        return AlwaysTrue()

    def visit_false(self) -> BooleanExpression:
        return AlwaysFalse()

    def visit_not(self, child_result: BooleanExpression) -> BooleanExpression:
        return Not(child=child_result)

    def visit_and(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return And(left=left_result, right=right_result)

    def visit_or(self, left_result: BooleanExpression, right_result: BooleanExpression) -> BooleanExpression:
        return Or(left=left_result, right=right_result)

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> BooleanExpression:
        raise TypeError(f"Expected Bound Predicate, got: {predicate.term}")

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> BooleanExpression:
        file_column_name = self.file_schema.find_column_name(predicate.term.ref().field.field_id)

        if file_column_name is None:
            # In the case of schema evolution, the column might not be present
            # in the file schema when reading older data
            if isinstance(predicate, BoundIsNull):
                return AlwaysTrue()
            else:
                return AlwaysFalse()

        if isinstance(predicate, BoundUnaryPredicate):
            return predicate.as_unbound(file_column_name)
        elif isinstance(predicate, BoundLiteralPredicate):
            return predicate.as_unbound(file_column_name, predicate.literal)
        elif isinstance(predicate, BoundSetPredicate):
            return predicate.as_unbound(file_column_name, predicate.literals)
        else:
            raise ValueError(f"Unsupported predicate: {predicate}")


def translate_column_names(expr: BooleanExpression, file_schema: Schema, case_sensitive: bool) -> BooleanExpression:
    return visit(expr, _ColumnNameTranslator(file_schema, case_sensitive))


class _ExpressionFieldIDs(BooleanExpressionVisitor[Set[int]]):
    """Extracts the field IDs used in the BooleanExpression."""

    def visit_true(self) -> Set[int]:
        return set()

    def visit_false(self) -> Set[int]:
        return set()

    def visit_not(self, child_result: Set[int]) -> Set[int]:
        return child_result

    def visit_and(self, left_result: Set[int], right_result: Set[int]) -> Set[int]:
        return left_result.union(right_result)

    def visit_or(self, left_result: Set[int], right_result: Set[int]) -> Set[int]:
        return left_result.union(right_result)

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> Set[int]:
        raise ValueError("Only works on bound records")

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> Set[int]:
        return {predicate.term.ref().field.field_id}


def extract_field_ids(expr: BooleanExpression) -> Set[int]:
    return visit(expr, _ExpressionFieldIDs())


class _RewriteToDNF(BooleanExpressionVisitor[Tuple[BooleanExpression, ...]]):
    def visit_true(self) -> Tuple[BooleanExpression, ...]:
        return (AlwaysTrue(),)

    def visit_false(self) -> Tuple[BooleanExpression, ...]:
        return (AlwaysFalse(),)

    def visit_not(self, child_result: Tuple[BooleanExpression, ...]) -> Tuple[BooleanExpression, ...]:
        raise ValueError(f"Not expressions are not allowed: {child_result}")

    def visit_and(
        self, left_result: Tuple[BooleanExpression, ...], right_result: Tuple[BooleanExpression, ...]
    ) -> Tuple[BooleanExpression, ...]:
        # Distributive law:
        # ((P OR Q) AND (R OR S)) AND (((P AND R) OR (P AND S)) OR ((Q AND R) OR ((Q AND S)))
        # A AND (B OR C) = (A AND B) OR (A AND C)
        # (A OR B) AND C = (A AND C) OR (B AND C)
        return tuple(And(le, re) for le in left_result for re in right_result)

    def visit_or(
        self, left_result: Tuple[BooleanExpression, ...], right_result: Tuple[BooleanExpression, ...]
    ) -> Tuple[BooleanExpression, ...]:
        return left_result + right_result

    def visit_unbound_predicate(self, predicate: UnboundPredicate[L]) -> Tuple[BooleanExpression, ...]:
        return (predicate,)

    def visit_bound_predicate(self, predicate: BoundPredicate[L]) -> Tuple[BooleanExpression, ...]:
        return (predicate,)


def rewrite_to_dnf(expr: BooleanExpression) -> Tuple[BooleanExpression, ...]:
    # Rewrites an arbitrary boolean expression to disjunctive normal form (DNF):
    # (A AND NOT(B) AND C) OR (NOT(D) AND E AND F) OR (G)
    expr_without_not = rewrite_not(expr)
    return visit(expr_without_not, _RewriteToDNF())


class ExpressionToPlainFormat(BoundBooleanExpressionVisitor[List[Tuple[str, str, Any]]]):
    cast_int_to_date: bool

    def __init__(self, cast_int_to_date: bool = False) -> None:
        self.cast_int_to_date = cast_int_to_date

    def _cast_if_necessary(self, iceberg_type: IcebergType, literal: Union[L, Set[L]]) -> Union[L, Set[L]]:
        if self.cast_int_to_date:
            iceberg_type_class = type(iceberg_type)
            conversions = {TimestampType: micros_to_timestamp, TimestamptzType: micros_to_timestamptz}
            if iceberg_type_class in conversions:
                conversion_function = conversions[iceberg_type_class]
                if isinstance(literal, set):
                    return {conversion_function(lit) for lit in literal}  # type: ignore
                else:
                    return conversion_function(literal)  # type: ignore
        return literal

    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> List[Tuple[str, str, Any]]:
        field = term.ref().field
        return [(term.ref().field.name, "in", self._cast_if_necessary(field.field_type, literals))]

    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> List[Tuple[str, str, Any]]:
        field = term.ref().field
        return [(field.name, "not in", self._cast_if_necessary(field.field_type, literals))]

    def visit_is_nan(self, term: BoundTerm[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "==", float("nan"))]

    def visit_not_nan(self, term: BoundTerm[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "!=", float("nan"))]

    def visit_is_null(self, term: BoundTerm[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "==", None)]

    def visit_not_null(self, term: BoundTerm[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "!=", None)]

    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "==", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "!=", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, ">=", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, ">", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "<", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return [(term.ref().field.name, "<=", self._cast_if_necessary(term.ref().field.field_type, literal.value))]

    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return []

    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> List[Tuple[str, str, Any]]:
        return []

    def visit_true(self) -> List[Tuple[str, str, Any]]:
        return []  # Not supported

    def visit_false(self) -> List[Tuple[str, str, Any]]:
        raise ValueError("Not supported: AlwaysFalse")

    def visit_not(self, child_result: List[Tuple[str, str, Any]]) -> List[Tuple[str, str, Any]]:
        raise ValueError(f"Not allowed: {child_result}")

    def visit_and(
        self, left_result: List[Tuple[str, str, Any]], right_result: List[Tuple[str, str, Any]]
    ) -> List[Tuple[str, str, Any]]:
        return left_result + right_result

    def visit_or(
        self, left_result: List[Tuple[str, str, Any]], right_result: List[Tuple[str, str, Any]]
    ) -> List[Tuple[str, str, Any]]:
        raise ValueError(f"Not allowed: {left_result} || {right_result}")


def expression_to_plain_format(
    expressions: Tuple[BooleanExpression, ...], cast_int_to_datetime: bool = False
) -> List[List[Tuple[str, str, Any]]]:
    """Format a Disjunctive Normal Form expression.

    These are the formats that the expression can be fed into:

    - https://arrow.apache.org/docs/python/generated/pyarrow.parquet.read_table.html
    - https://docs.dask.org/en/stable/generated/dask.dataframe.read_parquet.html

    Contrary to normal DNF that may contain Not expressions, but here they should have
    been rewritten. This can be done using ``rewrite_not(...)``.

    Keep in mind that this is only used for page skipping, and still needs to filter
    on a row level.

    Args:
        expressions: Expression in Disjunctive Normal Form.

    Returns:
        Formatter filter compatible with Dask and PyArrow.
    """
    # In the form of expr1 ∨ expr2 ∨ ... ∨ exprN
    visitor = ExpressionToPlainFormat(cast_int_to_datetime)
    return [visit(expression, visitor) for expression in expressions]


class _MetricsEvaluator(BoundBooleanExpressionVisitor[bool], ABC):
    value_counts: Dict[int, int]
    null_counts: Dict[int, int]
    nan_counts: Dict[int, int]
    lower_bounds: Dict[int, bytes]
    upper_bounds: Dict[int, bytes]

    def visit_true(self) -> bool:
        # all rows match
        return ROWS_MIGHT_MATCH

    def visit_false(self) -> bool:
        # all rows fail
        return ROWS_CANNOT_MATCH

    def visit_not(self, child_result: bool) -> bool:
        raise ValueError(f"NOT should be rewritten: {child_result}")

    def visit_and(self, left_result: bool, right_result: bool) -> bool:
        return left_result and right_result

    def visit_or(self, left_result: bool, right_result: bool) -> bool:
        return left_result or right_result

    def _contains_nulls_only(self, field_id: int) -> bool:
        if (value_count := self.value_counts.get(field_id)) and (null_count := self.null_counts.get(field_id)):
            return value_count == null_count
        return False

    def _contains_nans_only(self, field_id: int) -> bool:
        if (nan_count := self.nan_counts.get(field_id)) and (value_count := self.value_counts.get(field_id)):
            return nan_count == value_count
        return False

    def _is_nan(self, val: Any) -> bool:
        try:
            return math.isnan(val)
        except TypeError:
            # In the case of None or other non-numeric types
            return False


class _InclusiveMetricsEvaluator(_MetricsEvaluator):
    struct: StructType
    expr: BooleanExpression

    def __init__(
        self, schema: Schema, expr: BooleanExpression, case_sensitive: bool = True, include_empty_files: bool = False
    ) -> None:
        self.struct = schema.as_struct()
        self.include_empty_files = include_empty_files
        self.expr = bind(schema, rewrite_not(expr), case_sensitive)

    def eval(self, file: DataFile) -> bool:
        """Test whether the file may contain records that match the expression."""
        if not self.include_empty_files and file.record_count == 0:
            return ROWS_CANNOT_MATCH

        if file.record_count < 0:
            # Older version don't correctly implement record count from avro file and thus
            # set record count -1 when importing avro tables to iceberg tables. This should
            # be updated once we implemented and set correct record count.
            return ROWS_MIGHT_MATCH

        self.value_counts = file.value_counts or EMPTY_DICT
        self.null_counts = file.null_value_counts or EMPTY_DICT
        self.nan_counts = file.nan_value_counts or EMPTY_DICT
        self.lower_bounds = file.lower_bounds or EMPTY_DICT
        self.upper_bounds = file.upper_bounds or EMPTY_DICT

        return visit(self.expr, self)

    def _may_contain_null(self, field_id: int) -> bool:
        return self.null_counts is None or (field_id in self.null_counts and self.null_counts.get(field_id) is not None)

    def _contains_nans_only(self, field_id: int) -> bool:
        if (nan_count := self.nan_counts.get(field_id)) and (value_count := self.value_counts.get(field_id)):
            return nan_count == value_count
        return False

    def visit_is_null(self, term: BoundTerm[L]) -> bool:
        field_id = term.ref().field.field_id

        if self.null_counts.get(field_id) == 0:
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_null(self, term: BoundTerm[L]) -> bool:
        # no need to check whether the field is required because binding evaluates that case
        # if the column has no non-null values, the expression cannot match
        field_id = term.ref().field.field_id

        if self._contains_nulls_only(field_id):
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_is_nan(self, term: BoundTerm[L]) -> bool:
        field_id = term.ref().field.field_id

        if self.nan_counts.get(field_id) == 0:
            return ROWS_CANNOT_MATCH

        # when there's no nanCounts information, but we already know the column only contains null,
        # it's guaranteed that there's no NaN value
        if self._contains_nulls_only(field_id):
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_nan(self, term: BoundTerm[L]) -> bool:
        field_id = term.ref().field.field_id

        if self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if lower_bound_bytes := self.lower_bounds.get(field_id):
            lower_bound = from_bytes(field.field_type, lower_bound_bytes)

            if self._is_nan(lower_bound):
                # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                return ROWS_MIGHT_MATCH

            if lower_bound >= literal.value:
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if lower_bound_bytes := self.lower_bounds.get(field_id):
            lower_bound = from_bytes(field.field_type, lower_bound_bytes)
            if self._is_nan(lower_bound):
                # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                return ROWS_MIGHT_MATCH

            if lower_bound > literal.value:
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if upper_bound_bytes := self.upper_bounds.get(field_id):
            upper_bound = from_bytes(field.field_type, upper_bound_bytes)
            if upper_bound <= literal.value:
                if self._is_nan(upper_bound):
                    # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                    return ROWS_MIGHT_MATCH

                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if upper_bound_bytes := self.upper_bounds.get(field_id):
            upper_bound = from_bytes(field.field_type, upper_bound_bytes)
            if upper_bound < literal.value:
                if self._is_nan(upper_bound):
                    # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                    return ROWS_MIGHT_MATCH

                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if lower_bound_bytes := self.lower_bounds.get(field_id):
            lower_bound = from_bytes(field.field_type, lower_bound_bytes)
            if self._is_nan(lower_bound):
                # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                return ROWS_MIGHT_MATCH

            if lower_bound > literal.value:
                return ROWS_CANNOT_MATCH

        if upper_bound_bytes := self.upper_bounds.get(field_id):
            upper_bound = from_bytes(field.field_type, upper_bound_bytes)
            if self._is_nan(upper_bound):
                # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                return ROWS_MIGHT_MATCH

            if upper_bound < literal.value:
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return ROWS_MIGHT_MATCH

    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        field = term.ref().field
        field_id = field.field_id

        if self._contains_nulls_only(field_id) or self._contains_nans_only(field_id):
            return ROWS_CANNOT_MATCH

        if len(literals) > IN_PREDICATE_LIMIT:
            # skip evaluating the predicate if the number of values is too big
            return ROWS_MIGHT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        if lower_bound_bytes := self.lower_bounds.get(field_id):
            lower_bound = from_bytes(field.field_type, lower_bound_bytes)
            if self._is_nan(lower_bound):
                # NaN indicates unreliable bounds. See the InclusiveMetricsEvaluator docs for more.
                return ROWS_MIGHT_MATCH

            literals = {lit for lit in literals if lower_bound <= lit}
            if len(literals) == 0:
                return ROWS_CANNOT_MATCH

        if upper_bound_bytes := self.upper_bounds.get(field_id):
            upper_bound = from_bytes(field.field_type, upper_bound_bytes)
            # this is different from Java, here NaN is always larger
            if self._is_nan(upper_bound):
                return ROWS_MIGHT_MATCH

            literals = {lit for lit in literals if upper_bound >= lit}
            if len(literals) == 0:
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        # because the bounds are not necessarily a min or max value, this cannot be answered using
        # them. notIn(col, {X, ...}) with (X, Y) doesn't guarantee that X is a value in col.
        return ROWS_MIGHT_MATCH

    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id: int = field.field_id

        if self._contains_nulls_only(field_id):
            return ROWS_CANNOT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        prefix = str(literal.value)
        len_prefix = len(prefix)

        if lower_bound_bytes := self.lower_bounds.get(field_id):
            lower_bound = str(from_bytes(field.field_type, lower_bound_bytes))

            # truncate lower bound so that its length is not greater than the length of prefix
            if lower_bound and lower_bound[:len_prefix] > prefix:
                return ROWS_CANNOT_MATCH

        if upper_bound_bytes := self.upper_bounds.get(field_id):
            upper_bound = str(from_bytes(field.field_type, upper_bound_bytes))

            # truncate upper bound so that its length is not greater than the length of prefix
            if upper_bound is not None and upper_bound[:len_prefix] < prefix:
                return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH

    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        field = term.ref().field
        field_id: int = field.field_id

        if self._may_contain_null(field_id):
            return ROWS_MIGHT_MATCH

        if not isinstance(field.field_type, PrimitiveType):
            raise ValueError(f"Expected PrimitiveType: {field.field_type}")

        prefix = str(literal.value)
        len_prefix = len(prefix)

        # not_starts_with will match unless all values must start with the prefix. This happens when
        # the lower and upper bounds both start with the prefix.
        if (lower_bound_bytes := self.lower_bounds.get(field_id)) and (upper_bound_bytes := self.upper_bounds.get(field_id)):
            lower_bound = str(from_bytes(field.field_type, lower_bound_bytes))
            upper_bound = str(from_bytes(field.field_type, upper_bound_bytes))

            # if lower is shorter than the prefix then lower doesn't start with the prefix
            if len(lower_bound) < len_prefix:
                return ROWS_MIGHT_MATCH

            if lower_bound[:len_prefix] == prefix:
                # if upper is shorter than the prefix then upper can't start with the prefix
                if len(upper_bound) < len_prefix:
                    return ROWS_MIGHT_MATCH

                if upper_bound[:len_prefix] == prefix:
                    return ROWS_CANNOT_MATCH

        return ROWS_MIGHT_MATCH


def strict_projection(
    schema: Schema, spec: PartitionSpec, case_sensitive: bool = True
) -> Callable[[BooleanExpression], BooleanExpression]:
    return StrictProjection(schema, spec, case_sensitive).project


class StrictProjection(ProjectionEvaluator):
    def visit_bound_predicate(self, predicate: BoundPredicate[Any]) -> BooleanExpression:
        parts = self.spec.fields_by_source_id(predicate.term.ref().field.field_id)

        result: BooleanExpression = AlwaysFalse()
        for part in parts:
            # consider (ts > 2019-01-01T01:00:00) with day(ts) and hour(ts)
            # projections: d >= 2019-01-02 and h >= 2019-01-01-02 (note the inclusive bounds).
            # any timestamp where either projection predicate is true must match the original
            # predicate. For example, ts = 2019-01-01T03:00:00 matches the hour projection but not
            # the day, but does match the original predicate.
            strict_projection = part.transform.strict_project(name=part.name, pred=predicate)
            if strict_projection is not None:
                result = Or(result, strict_projection)

        return result


class _StrictMetricsEvaluator(_MetricsEvaluator):
    struct: StructType
    expr: BooleanExpression

    def __init__(
        self, schema: Schema, expr: BooleanExpression, case_sensitive: bool = True, include_empty_files: bool = False
    ) -> None:
        self.struct = schema.as_struct()
        self.include_empty_files = include_empty_files
        self.expr = bind(schema, rewrite_not(expr), case_sensitive)

    def eval(self, file: DataFile) -> bool:
        """Test whether all records within the file match the expression.

        Args:
            file: A data file

        Returns: false if the file may contain any row that doesn't match
                    the expression, true otherwise.
        """
        if file.record_count <= 0:
            # Older version don't correctly implement record count from avro file and thus
            # set record count -1 when importing avro tables to iceberg tables. This should
            # be updated once we implemented and set correct record count.
            return ROWS_MUST_MATCH

        self.value_counts = file.value_counts or EMPTY_DICT
        self.null_counts = file.null_value_counts or EMPTY_DICT
        self.nan_counts = file.nan_value_counts or EMPTY_DICT
        self.lower_bounds = file.lower_bounds or EMPTY_DICT
        self.upper_bounds = file.upper_bounds or EMPTY_DICT

        return visit(self.expr, self)

    def visit_is_null(self, term: BoundTerm[L]) -> bool:
        # no need to check whether the field is required because binding evaluates that case
        # if the column has any non-null values, the expression does not match
        field_id = term.ref().field.field_id

        if self._contains_nulls_only(field_id):
            return ROWS_MUST_MATCH
        else:
            return ROWS_MIGHT_NOT_MATCH

    def visit_not_null(self, term: BoundTerm[L]) -> bool:
        # no need to check whether the field is required because binding evaluates that case
        # if the column has any non-null values, the expression does not match
        field_id = term.ref().field.field_id

        if (null_count := self.null_counts.get(field_id)) is not None and null_count == 0:
            return ROWS_MUST_MATCH
        else:
            return ROWS_MIGHT_NOT_MATCH

    def visit_is_nan(self, term: BoundTerm[L]) -> bool:
        field_id = term.ref().field.field_id

        if self._contains_nans_only(field_id):
            return ROWS_MUST_MATCH
        else:
            return ROWS_MIGHT_NOT_MATCH

    def visit_not_nan(self, term: BoundTerm[L]) -> bool:
        field_id = term.ref().field.field_id

        if (nan_count := self.nan_counts.get(field_id)) is not None and nan_count == 0:
            return ROWS_MUST_MATCH

        if self._contains_nulls_only(field_id):
            return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_less_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when: <----------Min----Max---X------->

        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        if upper_bytes := self.upper_bounds.get(field_id):
            field = self._get_field(field_id)
            upper = _from_byte_buffer(field.field_type, upper_bytes)

            if upper < literal.value:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_less_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when: <----------Min----Max---X------->

        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        if upper_bytes := self.upper_bounds.get(field_id):
            field = self._get_field(field_id)
            upper = _from_byte_buffer(field.field_type, upper_bytes)

            if upper <= literal.value:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_greater_than(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when: <-------X---Min----Max---------->

        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        if lower_bytes := self.lower_bounds.get(field_id):
            field = self._get_field(field_id)
            lower = _from_byte_buffer(field.field_type, lower_bytes)

            if self._is_nan(lower):
                # NaN indicates unreliable bounds.
                # See the _StrictMetricsEvaluator docs for more.
                return ROWS_MIGHT_NOT_MATCH

            if lower > literal.value:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_greater_than_or_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when: <-------X---Min----Max---------->
        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        if lower_bytes := self.lower_bounds.get(field_id):
            field = self._get_field(field_id)
            lower = _from_byte_buffer(field.field_type, lower_bytes)

            if self._is_nan(lower):
                # NaN indicates unreliable bounds.
                # See the _StrictMetricsEvaluator docs for more.
                return ROWS_MIGHT_NOT_MATCH

            if lower >= literal.value:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when Min == X == Max
        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        if (lower_bytes := self.lower_bounds.get(field_id)) and (upper_bytes := self.upper_bounds.get(field_id)):
            field = self._get_field(field_id)
            lower = _from_byte_buffer(field.field_type, lower_bytes)
            upper = _from_byte_buffer(field.field_type, upper_bytes)

            if lower != literal.value or upper != literal.value:
                return ROWS_MIGHT_NOT_MATCH
            else:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_not_equal(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        # Rows must match when X < Min or Max < X because it is not in the range
        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MUST_MATCH

        field = self._get_field(field_id)

        if lower_bytes := self.lower_bounds.get(field_id):
            lower = _from_byte_buffer(field.field_type, lower_bytes)

            if self._is_nan(lower):
                # NaN indicates unreliable bounds.
                # See the _StrictMetricsEvaluator docs for more.
                return ROWS_MIGHT_NOT_MATCH

            if lower > literal.value:
                return ROWS_MUST_MATCH

        if upper_bytes := self.upper_bounds.get(field_id):
            upper = _from_byte_buffer(field.field_type, upper_bytes)

            if upper < literal.value:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MIGHT_NOT_MATCH

        field = self._get_field(field_id)

        if (lower_bytes := self.lower_bounds.get(field_id)) and (upper_bytes := self.upper_bounds.get(field_id)):
            # similar to the implementation in eq, first check if the lower bound is in the set
            lower = _from_byte_buffer(field.field_type, lower_bytes)
            if lower not in literals:
                return ROWS_MIGHT_NOT_MATCH

            # check if the upper bound is in the set
            upper = _from_byte_buffer(field.field_type, upper_bytes)
            if upper not in literals:
                return ROWS_MIGHT_NOT_MATCH

            # finally check if the lower bound and the upper bound are equal
            if lower != upper:
                return ROWS_MIGHT_NOT_MATCH

            # All values must be in the set if the lower bound and the upper bound are
            # in the set and are equal.
            return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_not_in(self, term: BoundTerm[L], literals: Set[L]) -> bool:
        field_id = term.ref().field.field_id

        if self._can_contain_nulls(field_id) or self._can_contain_nans(field_id):
            return ROWS_MUST_MATCH

        field = self._get_field(field_id)

        if lower_bytes := self.lower_bounds.get(field_id):
            lower = _from_byte_buffer(field.field_type, lower_bytes)

            if self._is_nan(lower):
                # NaN indicates unreliable bounds.
                # See the StrictMetricsEvaluator docs for more.
                return ROWS_MIGHT_NOT_MATCH

            literals = {val for val in literals if lower <= val}
            if len(literals) == 0:
                return ROWS_MUST_MATCH

        if upper_bytes := self.upper_bounds.get(field_id):
            upper = _from_byte_buffer(field.field_type, upper_bytes)

            literals = {val for val in literals if upper >= val}

            if len(literals) == 0:
                return ROWS_MUST_MATCH

        return ROWS_MIGHT_NOT_MATCH

    def visit_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return ROWS_MIGHT_NOT_MATCH

    def visit_not_starts_with(self, term: BoundTerm[L], literal: Literal[L]) -> bool:
        return ROWS_MIGHT_NOT_MATCH

    def _get_field(self, field_id: int) -> NestedField:
        field = self.struct.field(field_id=field_id)
        if field is None:
            raise ValueError(f"Cannot find field, might be nested or missing: {field_id}")

        return field

    def _can_contain_nulls(self, field_id: int) -> bool:
        return (null_count := self.null_counts.get(field_id)) is not None and null_count > 0

    def _can_contain_nans(self, field_id: int) -> bool:
        return (nan_count := self.nan_counts.get(field_id)) is not None and nan_count > 0