Improve re-raises of Exceptions (#4142)

* Improve re-raises of Exceptions

When we use try-except in our code base, often we raise a new exception to replace the original exception. It is good practice to either wrap around the original exception (`from exc`), or bin the original exception (`from None`) if the new exception is a self-contained description of what happened. The latter is for instance the case when we wrap around a dict, catch the KeyError, and provide more meaningful Exception with message, as the `KeyError` is not what the user is after, it wants to know that dtype1 cannot be converted to dtype2, etc.

Without re-raise, you will see `During handling of the above exception, another exception occurred`, which means the two exceptions, original and new, are treated unrelated. With `from exc`, you will see `The above exception was the direct cause of the following exception:`. With `from None`, you will get a clean traceback without the original exception.

Found all cases using pylint (`pylint polars --disable=all --enable=raise-missing-from`). Discussion on the rollout of pylint in CI takes place in #4044.

Background reading: https://stefan.sofa-rockers.org/2020/10/28/raise-from/

* Remove test_example

py-polars/polars/datatypes.py

@@ -421,21 +421,27 @@ def dtype_to_ctype(dtype: PolarsDataType) -> type[_SimpleCData]:
     try:
         return _DTYPE_TO_CTYPE[dtype]
     except KeyError:  # pragma: no cover
-        raise NotImplementedError
+        raise NotImplementedError(
+            f"Conversion of polars data type {dtype} to C-type not implemented."
+        ) from None
 
 
 def dtype_to_ffiname(dtype: PolarsDataType) -> str:
     try:
         return _DTYPE_TO_FFINAME[dtype]
     except KeyError:  # pragma: no cover
-        raise NotImplementedError
+        raise NotImplementedError(
+            f"Conversion of polars data type {dtype} to FFI not implemented."
+        ) from None
 
 
 def dtype_to_py_type(dtype: PolarsDataType) -> type:
     try:
         return _DTYPE_TO_PY_TYPE[dtype]
     except KeyError:  # pragma: no cover
-        raise NotImplementedError
+        raise NotImplementedError(
+            f"Conversion of polars data type {dtype} to Python type not implemented."
+        ) from None
 
 
 def is_polars_dtype(data_type: Any) -> bool:
@@ -453,7 +459,9 @@ def py_type_to_dtype(data_type: Any) -> type[DataType]:
     try:
         return _PY_TYPE_TO_DTYPE[data_type]
     except KeyError:  # pragma: no cover
-        raise NotImplementedError
+        raise NotImplementedError(
+            f"Conversion of Python data type {data_type} to Polars data type not implemented."
+        ) from None
 
 
 def py_type_to_arrow_type(dtype: type[Any]) -> pa.lib.DataType:
@@ -463,7 +471,9 @@ def py_type_to_arrow_type(dtype: type[Any]) -> pa.lib.DataType:
     try:
         return _PY_TYPE_TO_ARROW_TYPE[dtype]
     except KeyError:  # pragma: no cover
-        raise ValueError(f"Cannot parse dtype {dtype} into Arrow dtype.")
+        raise ValueError(
+            f"Cannot parse Python data type {dtype} into Arrow data type."
+        ) from None
 
 
 def dtype_to_arrow_type(dtype: PolarsDataType) -> pa.lib.DataType:
@@ -473,7 +483,9 @@ def dtype_to_arrow_type(dtype: PolarsDataType) -> pa.lib.DataType:
     try:
         return _DTYPE_TO_ARROW_TYPE[dtype]
     except KeyError:  # pragma: no cover
-        raise ValueError(f"Cannot parse dtype {dtype} into Arrow dtype.")
+        raise ValueError(
+            f"Cannot parse data type {dtype} into Arrow data type."
+        ) from None
 
 
 def supported_numpy_char_code(dtype: str) -> bool:
@@ -484,7 +496,9 @@ def numpy_char_code_to_dtype(dtype: str) -> type[DataType]:
     try:
         return _NUMPY_CHAR_CODE_TO_DTYPE[dtype]
     except KeyError:  # pragma: no cover
-        raise NotImplementedError
+        raise NotImplementedError(
+            f"Cannot parse numpy data type {dtype} into Polars data type."
+        ) from None
 
 
 def maybe_cast(

py-polars/polars/datatypes_constructor.py

@@ -73,7 +73,7 @@ def polars_type_to_constructor(
     try:
         return _POLARS_TYPE_TO_CONSTRUCTOR[dtype]
     except KeyError:  # pragma: no cover
-        raise ValueError(f"Cannot construct PySeries for type {dtype}.")
+        raise ValueError(f"Cannot construct PySeries for type {dtype}.") from None
 
 
 if _NUMPY_AVAILABLE and not _DOCUMENTING:
@@ -102,7 +102,9 @@ def numpy_type_to_constructor(dtype: type[np.dtype]) -> Callable[..., PySeries]:
     except KeyError:
         return PySeries.new_object
     except NameError:  # pragma: no cover
-        raise ImportError("'numpy' is required for this functionality.")
+        raise ImportError(
+            f"'numpy' is required to convert numpy dtype {dtype}."
+        ) from None
 
 
 if not _DOCUMENTING:

py-polars/polars/internals/frame.py

@@ -1587,8 +1587,8 @@ def __getattr__(self, item: Any) -> PySeries:
                 DeprecationWarning,
             )
             return pli.wrap_s(self._df.column(item))
-        except Exception:
-            raise AttributeError(item)
+        except Exception as exc:
+            raise AttributeError(item) from exc
 
     def __contains__(self, key: str) -> bool:
         return key in self.columns
@@ -6179,7 +6179,7 @@ def get_group(self, group_value: Any | tuple[Any]) -> DF:
         try:
             groups_idx = groups[mask][0]  # type: ignore[index]
         except IndexError:
-            raise ValueError(f"no group: {group_value} found")
+            raise ValueError(f"no group: {group_value} found") from None
 
         df = self._dataframe_class._from_pydf(self._df)
         return df[groups_idx]

py-polars/polars/internals/lazy_frame.py

@@ -489,17 +489,17 @@ def show_graph(
                     ["dot", "-Nshape=box", "-Tsvg"], input=f"{dot}".encode()
                 )
                 return display(SVG(svg))
-            except Exception:
+            except Exception as exc:
                 raise ImportError(
                     "Graphviz dot binary should be on your PATH and matplotlib should be installed to show graph."
-                )
+                ) from exc
         try:
             import matplotlib.image as mpimg
             import matplotlib.pyplot as plt
         except ImportError:
             raise ImportError(
                 "Graphviz dot binary should be on your PATH and matplotlib should be installed to show graph."
-            )
+            ) from None
         dot = self._ldf.to_dot(optimized)
         if raw_output:
             return dot

py-polars/polars/io.py

@@ -1064,7 +1064,7 @@ def read_excel(
     except ImportError:
         raise ImportError(
             "xlsx2csv is not installed. Please run `pip install xlsx2csv`."
-        )
+        ) from None
 
     if isinstance(file, (str, Path)):
         file = format_path(file)