Merge branch 'main' into ignore_corrupt

Author: chenghuichen

.github/ISSUE_TEMPLATE/config.yml

@@ -4,5 +4,5 @@ contact_links:
   url: https://github.com/Eventual-Inc/Daft/discussions/new/choose
   about: Please ask questions here.
 - name: Slack
-  url: https://dist-data.slack.com/archives/C052CA6Q9N1
+  url: https://daft.ai/slack
   about: Or the `#daft-dev` channel in the Daft Slack.

README.rst

@@ -125,7 +125,7 @@ Daft has an Apache 2.0 license - please see the LICENSE file.
    :alt: Coverage
 
 .. |Slack| image:: https://img.shields.io/badge/slack-@distdata-purple.svg?logo=slack
-   :target: https://join.slack.com/t/dist-data/shared_invite/zt-3rh9jr9iv-tmmTNOlQpfvhEy2NTMWS_w
+   :target: https://daft.ai/slack
    :alt: slack community
 
 .. |TrendShift| image:: https://trendshift.io/api/badge/repositories/8239

daft/daft/__init__.pyi

@@ -1536,6 +1536,7 @@ class PySchema:
 class PyExpr:
     def alias(self, name: str) -> PyExpr: ...
     def cast(self, dtype: PyDataType) -> PyExpr: ...
+    def try_cast(self, dtype: PyDataType) -> PyExpr: ...
     def if_else(self, if_true: PyExpr, if_false: PyExpr) -> PyExpr: ...
     def count(self, mode: CountMode) -> PyExpr: ...
     def count_distinct(self) -> PyExpr: ...
@@ -1830,6 +1831,7 @@ class PySeries:
     def agg_list(self) -> PySeries: ...
     def agg_set(self) -> PySeries: ...
     def cast(self, dtype: PyDataType) -> PySeries: ...
+    def try_cast(self, dtype: PyDataType) -> PySeries: ...
     def pow(self, exp: float) -> PySeries: ...
     def log2(self) -> PySeries: ...
     def log10(self) -> PySeries: ...
@@ -2808,8 +2810,8 @@ class PyFileReference:
     def writable(self) -> bool: ...
     def path(self) -> str: ...
     def name(self) -> str: ...
-    def offset(self) -> int | None: ...
-    def length(self) -> int | None: ...
+    def position(self) -> int | None: ...
+    def size(self) -> int | None: ...
 
 class PyDaftFile:
     def __init__(self, path: str | None = None, data: bytes | None = None) -> None: ...

daft/expressions/expressions.py

@@ -497,6 +497,19 @@ def cast(self, dtype: DataTypeLike) -> Expression:
 
         return cast(self, dtype)
 
+    def try_cast(self, dtype: DataTypeLike) -> Expression:
+        """Attempts to cast an expression to the given datatype, returning null on failure.
+
+        Unlike `cast`, this method does not raise an error when the conversion fails.
+        Instead, it returns null for values that cannot be converted.
+
+        Tip: See Also
+            [`daft.functions.try_cast`](https://docs.daft.ai/en/stable/api/functions/try_cast/)
+        """
+        from daft.functions import try_cast
+
+        return try_cast(self, dtype)
+
     if TYPE_CHECKING:
 
         def as_int8(self) -> Expression: ...

daft/expressions/pyarrow_visitor.py

@@ -38,6 +38,12 @@ def visit_cast(self, expr: Expression, dtype: DataType) -> pc.Expression:
         pc_type = dtype.to_arrow_dtype()
         return pc_expr.cast(pc_type)
 
+    def visit_try_cast(self, expr: Expression, dtype: DataType) -> pc.Expression:
+        """Converts the try_cast with safe=True cast options."""
+        pc_expr = self.visit(expr)
+        pc_type = dtype.to_arrow_dtype()
+        return pc_expr.cast(pc_type, safe=True)
+
     def visit_list(self, items: list[Expression]) -> pc.Expression:
         raise ValueError("pyarrow.compute does not have a make_list function.")
 

daft/expressions/visitor.py

@@ -32,6 +32,10 @@ class ExpressionVisitor(ABC, Generic[R]):
         ...         print(f"Cast: {dtype}")
         ...         self.visit(expr)
         ...
+        ...     def visit_try_cast(self, expr: Expression, dtype: DataType) -> None:
+        ...         print(f"TryCast: {dtype}")
+        ...         self.visit(expr)
+        ...
         ...     def visit_function(self, name: str, args: list[Expression]) -> None:
         ...         print(f"Function: {name}")
         ...         for arg in args:
@@ -78,6 +82,13 @@ def visit_cast(self, expr: Expression, dtype: DataType) -> R:
         """Visit a cast expression."""
         ...
 
+    def visit_try_cast(self, expr: Expression, dtype: DataType) -> R:
+        """Visit a try_cast expression.
+
+        Default implementation delegates to visit_cast for backwards compatibility.
+        """
+        return self.visit_cast(expr, dtype)
+
     @abstractmethod
     def visit_function(self, name: str, args: list[Expression]) -> R:
         """Visit a function call expression."""
@@ -173,6 +184,9 @@ def visit_alias(self, expr: Expression, alias: str) -> set[str]:
     def visit_cast(self, expr: Expression, dtype: DataType) -> set[str]:
         return self.visit(expr)
 
+    def visit_try_cast(self, expr: Expression, dtype: DataType) -> set[str]:
+        return self.visit(expr)
+
     def visit_function(self, name: str, args: list[Expression]) -> set[str]:
         return set().union(*(self.visit(arg) for arg in args))
 

daft/file/file.py

@@ -2,6 +2,7 @@
 
 import shutil
 import tempfile
+import warnings
 from typing import TYPE_CHECKING
 
 from daft.daft import PyDaftFile, PyFileReference
@@ -57,10 +58,30 @@ def __init__(
         url: str,
         io_config: IOConfig | None = None,
         media_type: MediaType = MediaType.unknown(),
+        position: int | None = None,
+        size: int | None = None,
         offset: int | None = None,
         length: int | None = None,
     ) -> None:
-        self._inner = PyFileReference._from_tuple((media_type._media_type, url, io_config, offset, length))  # type: ignore
+        if offset is not None:
+            warnings.warn(
+                "`offset` is deprecated; use `position` instead.",
+                DeprecationWarning,
+                stacklevel=2,
+            )
+            if position is None:
+                position = offset
+
+        if length is not None:
+            warnings.warn(
+                "`length` is deprecated; use `size` instead.",
+                DeprecationWarning,
+                stacklevel=2,
+            )
+            if size is None:
+                size = length
+
+        self._inner = PyFileReference._from_tuple((media_type._media_type, url, io_config, position, size))  # type: ignore
 
     def open(self, buffer_size: int | None = None) -> PyDaftFile:
         return PyDaftFile._from_file_reference(self._inner, buffer_size=buffer_size)
@@ -110,17 +131,37 @@ def name(self) -> str:
         """
         return self._inner.name()
 
+    @property
+    def position(self) -> int | None:
+        """The starting byte position for range reads, or None for full-file reads."""
+        return self._inner.position()
+
     @property
     def offset(self) -> int | None:
-        """The byte offset for range reads, or None for full-file reads."""
-        return self._inner.offset()
+        """Deprecated alias for `position`. The byte offset for range reads, or None for full-file reads."""
+        warnings.warn(
+            "`File.offset` is deprecated; use `File.position` instead.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+        return self._inner.position()
 
     @property
     def length(self) -> int | None:
-        """The byte length for range reads, or None for full-file reads."""
-        return self._inner.length()
+        """Deprecated alias for the byte-range read window size, or None for full-file reads.
+
+        Note: this returns the requested range size (caller intent), not the derived file
+        size. Use `File.size()` for the actual file size.
+        """
+        warnings.warn(
+            "`File.length` is deprecated; use `File.size()` instead.",
+            DeprecationWarning,
+            stacklevel=2,
+        )
+        return self._inner.size()
 
     def size(self) -> int:
+        """The size of the file in bytes, derived from the underlying file."""
         return PyDaftFile._from_file_reference(self._inner, buffer_size=BUFFER_SNIFF).size()
 
     def mime_type(self) -> str:

daft/functions/__init__.py

@@ -147,6 +147,7 @@
     random_int,
     eq_null_safe,
     cast,
+    try_cast,
     is_null,
     not_null,
     fill_null,
@@ -570,6 +571,7 @@
     "total_nanoseconds",
     "total_seconds",
     "trunc",
+    "try_cast",
     "try_compress",
     "try_decode",
     "try_decompress",

daft/functions/misc.py

@@ -189,6 +189,45 @@ def cast(expr: Expression, dtype: DataTypeLike) -> Expression:
     return Expression._from_pyexpr(expr._expr.cast(dtype._dtype))
 
 
+def try_cast(expr: Expression, dtype: DataTypeLike) -> Expression:
+    """Attempts to cast an expression to the given datatype, returning null on failure.
+
+    Unlike `cast`, this function does not raise an error when the conversion fails.
+    Instead, it returns null for values that cannot be converted.
+
+    Returns:
+        Expression: Expression with the specified new datatype, with null for failed conversions
+
+    Note:
+        - If a string is provided, it will use the sql engine to parse the string into a data type.
+        - A python `type` can also be provided, in which case the corresponding Daft data type will be used.
+
+    Examples:
+        >>> import daft
+        >>> df = daft.from_pydict({"str_val": ["1", "2", "abc", None]})
+        >>> df = df.select(df["str_val"].try_cast(daft.DataType.int64()))
+        >>> df.show()
+        ╭─────────╮
+        │ str_val │
+        │ ---     │
+        │ Int64   │
+        ╞═════════╡
+        │ 1       │
+        ├╌╌╌╌╌╌╌╌╌┤
+        │ 2       │
+        ├╌╌╌╌╌╌╌╌╌┤
+        │ None    │
+        ├╌╌╌╌╌╌╌╌╌┤
+        │ None    │
+        ╰─────────╯
+        <BLANKLINE>
+        (Showing first 4 of 4 rows)
+    """
+    dtype = DataType._infer(dtype)
+    expr = Expression._to_expression(expr)
+    return Expression._from_pyexpr(expr._expr.try_cast(dtype._dtype))
+
+
 def is_null(expr: Expression) -> Expression:
     """Checks if values in the Expression are Null (a special value indicating missing data).
 

daft/io/iceberg/_visitors.py

@@ -69,6 +69,9 @@ def visit_alias(self, expr: Expression, alias: str) -> BooleanExpression:
     def visit_cast(self, expr: Expression, dtype: DataType) -> BooleanExpression:
         return self.visit(expr)
 
+    def visit_try_cast(self, expr: Expression, dtype: DataType) -> BooleanExpression:
+        return self.visit(expr)
+
     def visit_function(self, name: str, args: list[Expression]) -> BooleanExpression:
         raise ValueError(f"Iceberg does not support function '{name}' in filter expressions")