feat: make LiteralPredicate serializable via internal IcebergBaseModel

pyiceberg/expressions/__init__.py

@@ -30,10 +30,11 @@
     Type,
     TypeVar,
     Union,
+    cast,
 )
 from typing import Literal as TypingLiteral
 
-from pydantic import Field
+from pydantic import ConfigDict, Field, field_validator
 
 from pyiceberg.expressions.literals import (
     AboveMax,
@@ -52,8 +53,14 @@
     ConfigDict = dict
 
 
-def _to_unbound_term(term: Union[str, UnboundTerm[Any]]) -> UnboundTerm[Any]:
-    return Reference(term) if isinstance(term, str) else term
+def _to_unbound_term(term: Union[str, UnboundTerm[Any], BoundReference[Any]]) -> UnboundTerm[Any]:
+    if isinstance(term, str):
+        return Reference(term)
+    if isinstance(term, UnboundTerm):
+        return term
+    if isinstance(term, BoundReference):
+        return Reference(term.field.name)
+    raise ValueError(f"Expected UnboundTerm | BoundReference | str, got {type(term).__name__}")
 
 
 def _to_literal_set(values: Union[Iterable[L], Iterable[Literal[L]]]) -> Set[Literal[L]]:
@@ -743,12 +750,52 @@ def as_bound(self) -> Type[BoundNotIn[L]]:
         return BoundNotIn[L]
 
 
-class LiteralPredicate(UnboundPredicate[L], ABC):
-    literal: Literal[L]
+class LiteralPredicate(IcebergBaseModel, UnboundPredicate[L], ABC):
+    type: TypingLiteral["lt", "lt-eq", "gt", "gt-eq", "eq", "not-eq", "starts-with", "not-starts-with"] = Field(alias="type")
+    term: UnboundTerm[Any]
+    value: Literal[L] = Field(alias="literal", serialization_alias="value")
+
+    model_config = ConfigDict(populate_by_name=True, frozen=True, arbitrary_types_allowed=True)
+
+    def __init__(
+        self,
+        term: Union[str, UnboundTerm[Any], BoundReference[Any]],
+        literal: Union[L, Literal[L], None] = None,
+        **data: Any,
+    ) -> None:  # pylint: disable=W0621
+        extra = dict(data)
+
+        literal_candidates = []
+        if literal is not None:
+            literal_candidates.append(literal)
+        if "literal" in extra:
+            literal_candidates.append(extra.pop("literal"))
+        if "value" in extra:
+            literal_candidates.append(extra.pop("value"))
+
+        literal_candidates = [candidate for candidate in literal_candidates if candidate is not None]
+
+        if not literal_candidates:
+            raise TypeError("LiteralPredicate requires a literal or value argument")
+        if len(literal_candidates) > 1:
+            raise TypeError("literal/value provided multiple times")
+
+        init = cast("Callable[..., None]", IcebergBaseModel.__init__)
+        init(self, term=_to_unbound_term(term), literal=_to_literal(literal_candidates[0]), **extra)
+
+    @field_validator("term", mode="before")
+    @classmethod
+    def _convert_term(cls, value: Any) -> UnboundTerm[Any]:
+        return _to_unbound_term(value)
+
+    @field_validator("value", mode="before")
+    @classmethod
+    def _convert_value(cls, value: Any) -> Literal[Any]:
+        return _to_literal(value)
 
-    def __init__(self, term: Union[str, UnboundTerm[Any]], literal: Union[L, Literal[L]]):  # pylint: disable=W0621
-        super().__init__(term)
-        self.literal = _to_literal(literal)  # pylint: disable=W0621
+    @property
+    def literal(self) -> Literal[L]:
+        return self.value
 
     def bind(self, schema: Schema, case_sensitive: bool = True) -> BoundLiteralPredicate[L]:
         bound_term = self.term.bind(schema, case_sensitive)
@@ -773,6 +820,10 @@ def __eq__(self, other: Any) -> bool:
             return self.term == other.term and self.literal == other.literal
         return False
 
+    def __str__(self) -> str:
+        """Return the string representation of the LiteralPredicate class."""
+        return f"{str(self.__class__.__name__)}(term={repr(self.term)}, literal={repr(self.literal)})"
+
     def __repr__(self) -> str:
         """Return the string representation of the LiteralPredicate class."""
         return f"{str(self.__class__.__name__)}(term={repr(self.term)}, literal={repr(self.literal)})"
@@ -886,6 +937,8 @@ def as_unbound(self) -> Type[NotStartsWith[L]]:
 
 
 class EqualTo(LiteralPredicate[L]):
+    type: TypingLiteral["eq"] = Field(default="eq", alias="type")
+
     def __invert__(self) -> NotEqualTo[L]:
         """Transform the Expression into its negated version."""
         return NotEqualTo[L](self.term, self.literal)
@@ -896,6 +949,8 @@ def as_bound(self) -> Type[BoundEqualTo[L]]:
 
 
 class NotEqualTo(LiteralPredicate[L]):
+    type: TypingLiteral["not-eq"] = Field(default="not-eq", alias="type")
+
     def __invert__(self) -> EqualTo[L]:
         """Transform the Expression into its negated version."""
         return EqualTo[L](self.term, self.literal)
@@ -906,6 +961,8 @@ def as_bound(self) -> Type[BoundNotEqualTo[L]]:
 
 
 class LessThan(LiteralPredicate[L]):
+    type: TypingLiteral["lt"] = Field(default="lt", alias="type")
+
     def __invert__(self) -> GreaterThanOrEqual[L]:
         """Transform the Expression into its negated version."""
         return GreaterThanOrEqual[L](self.term, self.literal)
@@ -916,6 +973,8 @@ def as_bound(self) -> Type[BoundLessThan[L]]:
 
 
 class GreaterThanOrEqual(LiteralPredicate[L]):
+    type: TypingLiteral["gt-eq"] = Field(default="gt-eq", alias="type")
+
     def __invert__(self) -> LessThan[L]:
         """Transform the Expression into its negated version."""
         return LessThan[L](self.term, self.literal)
@@ -926,6 +985,8 @@ def as_bound(self) -> Type[BoundGreaterThanOrEqual[L]]:
 
 
 class GreaterThan(LiteralPredicate[L]):
+    type: TypingLiteral["gt"] = Field(default="gt", alias="type")
+
     def __invert__(self) -> LessThanOrEqual[L]:
         """Transform the Expression into its negated version."""
         return LessThanOrEqual[L](self.term, self.literal)
@@ -936,6 +997,8 @@ def as_bound(self) -> Type[BoundGreaterThan[L]]:
 
 
 class LessThanOrEqual(LiteralPredicate[L]):
+    type: TypingLiteral["lt-eq"] = Field(default="lt-eq", alias="type")
+
     def __invert__(self) -> GreaterThan[L]:
         """Transform the Expression into its negated version."""
         return GreaterThan[L](self.term, self.literal)
@@ -946,6 +1009,8 @@ def as_bound(self) -> Type[BoundLessThanOrEqual[L]]:
 
 
 class StartsWith(LiteralPredicate[L]):
+    type: TypingLiteral["starts-with"] = Field(default="starts-with", alias="type")
+
     def __invert__(self) -> NotStartsWith[L]:
         """Transform the Expression into its negated version."""
         return NotStartsWith[L](self.term, self.literal)
@@ -956,6 +1021,8 @@ def as_bound(self) -> Type[BoundStartsWith[L]]:
 
 
 class NotStartsWith(LiteralPredicate[L]):
+    type: TypingLiteral["not-starts-with"] = Field(default="not-starts-with", alias="type")
+
     def __invert__(self) -> StartsWith[L]:
         """Transform the Expression into its negated version."""
         return StartsWith[L](self.term, self.literal)

pyiceberg/transforms.py

@@ -120,7 +120,7 @@ def _try_import(module_name: str, extras_name: Optional[str] = None) -> types.Mo
         raise NotInstalledError(msg) from None
 
 
-def _transform_literal(func: Callable[[L], L], lit: Literal[L]) -> Literal[L]:
+def _transform_literal(func: Callable[[Any], Any], lit: Literal[L]) -> Literal[L]:
     """Small helper to upwrap the value from the literal, and wrap it again."""
     return literal(func(lit.value))
 

tests/expressions/test_evaluator.py

@@ -22,6 +22,7 @@
 from pyiceberg.conversions import to_bytes
 from pyiceberg.expressions import (
     And,
+    BooleanExpression,
     EqualTo,
     GreaterThan,
     GreaterThanOrEqual,
@@ -30,6 +31,7 @@
     IsNull,
     LessThan,
     LessThanOrEqual,
+    LiteralPredicate,
     Not,
     NotEqualTo,
     NotIn,
@@ -301,7 +303,7 @@ def test_missing_stats() -> None:
         upper_bounds=None,
     )
 
-    expressions = [
+    expressions: list[BooleanExpression] = [
         LessThan("no_stats", 5),
         LessThanOrEqual("no_stats", 30),
         EqualTo("no_stats", 70),
@@ -324,7 +326,7 @@ def test_zero_record_file_stats(schema_data_file: Schema) -> None:
         file_path="file_1.parquet", file_format=FileFormat.PARQUET, partition=Record(), record_count=0
     )
 
-    expressions = [
+    expressions: list[BooleanExpression] = [
         LessThan("no_stats", 5),
         LessThanOrEqual("no_stats", 30),
         EqualTo("no_stats", 70),
@@ -683,26 +685,27 @@ def data_file_nan() -> DataFile:
 
 
 def test_inclusive_metrics_evaluator_less_than_and_less_than_equal(schema_data_file_nan: Schema, data_file_nan: DataFile) -> None:
-    for operator in [LessThan, LessThanOrEqual]:  # type: ignore
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+    operators: tuple[type[LiteralPredicate[Any]], ...] = (LessThan, LessThanOrEqual)
+    for operator in operators:
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: all nan column doesn't contain number"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: 1 is smaller than lower bound"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 10)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 10)).eval(data_file_nan)
         assert should_read, "Should match: 10 is larger than lower bound"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("min_max_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("min_max_nan", 1)).eval(data_file_nan)
         assert should_read, "Should match: no visibility"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan_null_bounds", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan_null_bounds", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: all nan column doesn't contain number"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: 1 is smaller than lower bound"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 10)).eval(  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 10)).eval(
             data_file_nan
         )
         assert should_read, "Should match: 10 larger than lower bound"
@@ -711,31 +714,32 @@ def test_inclusive_metrics_evaluator_less_than_and_less_than_equal(schema_data_f
 def test_inclusive_metrics_evaluator_greater_than_and_greater_than_equal(
     schema_data_file_nan: Schema, data_file_nan: DataFile
 ) -> None:
-    for operator in [GreaterThan, GreaterThanOrEqual]:  # type: ignore
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+    operators: tuple[type[LiteralPredicate[Any]], ...] = (GreaterThan, GreaterThanOrEqual)
+    for operator in operators:
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: all nan column doesn't contain number"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 1)).eval(data_file_nan)
         assert should_read, "Should match: upper bound is larger than 1"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 10)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("max_nan", 10)).eval(data_file_nan)
         assert should_read, "Should match: upper bound is larger than 10"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("min_max_nan", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("min_max_nan", 1)).eval(data_file_nan)
         assert should_read, "Should match: no visibility"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan_null_bounds", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan_null_bounds", 1)).eval(data_file_nan)
         assert not should_read, "Should not match: all nan column doesn't contain number"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 1)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 1)).eval(data_file_nan)
         assert should_read, "Should match: 1 is smaller than upper bound"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 10)).eval(  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("some_nan_correct_bounds", 10)).eval(
             data_file_nan
         )
         assert should_read, "Should match: 10 is smaller than upper bound"
 
-        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 30)).eval(data_file_nan)  # type: ignore[arg-type]
+        should_read = _InclusiveMetricsEvaluator(schema_data_file_nan, operator("all_nan", 30)).eval(data_file_nan)
         assert not should_read, "Should not match: 30 is greater than upper bound"
 
 
@@ -1162,7 +1166,7 @@ def test_strict_missing_stats(strict_data_file_schema: Schema, strict_data_file_
         upper_bounds=None,
     )
 
-    expressions = [
+    expressions: list[BooleanExpression] = [
         LessThan("no_stats", 5),
         LessThanOrEqual("no_stats", 30),
         EqualTo("no_stats", 70),
@@ -1185,7 +1189,7 @@ def test_strict_zero_record_file_stats(strict_data_file_schema: Schema) -> None:
         file_path="file_1.parquet", file_format=FileFormat.PARQUET, partition=Record(), record_count=0
     )
 
-    expressions = [
+    expressions: list[BooleanExpression] = [
         LessThan("no_stats", 5),
         LessThanOrEqual("no_stats", 30),
         EqualTo("no_stats", 70),