fix(rust,python): make python schema_overrides information available to the rust-side inference code when initialising from records/dicts

crates/polars-core/src/schema.rs

@@ -382,6 +382,13 @@ impl Schema {
         self.inner.iter().map(|(_name, dtype)| dtype)
     }
 
+    /// Iterates over mut references to the dtypes in this schema
+    pub fn iter_dtypes_mut(
+        &mut self,
+    ) -> impl Iterator<Item = &mut DataType> + '_ + ExactSizeIterator {
+        self.inner.iter_mut().map(|(_name, dtype)| dtype)
+    }
+
     /// Iterates over references to the names in this schema
     pub fn iter_names(&self) -> impl Iterator<Item = &SmartString> + '_ + ExactSizeIterator {
         self.inner.iter().map(|(name, _dtype)| name)

py-polars/polars/dataframe/frame.py

@@ -425,7 +425,9 @@ def _from_dicts(
         schema_overrides: SchemaDict | None = None,
         infer_schema_length: int | None = N_INFER_DEFAULT,
     ) -> Self:
-        pydf = PyDataFrame.read_dicts(data, infer_schema_length, schema)
+        pydf = PyDataFrame.read_dicts(
+            data, infer_schema_length, schema, schema_overrides
+        )
         if schema or schema_overrides:
             pydf = _post_apply_columns(
                 pydf, list(schema or pydf.columns()), schema_overrides=schema_overrides

py-polars/polars/utils/_construction.py

@@ -1148,7 +1148,9 @@ def _sequence_of_dict_to_pydf(
         if column_names
         else None
     )
-    pydf = PyDataFrame.read_dicts(data, infer_schema_length, dicts_schema)
+    pydf = PyDataFrame.read_dicts(
+        data, infer_schema_length, dicts_schema, schema_overrides
+    )
 
     # TODO: we can remove this `schema_overrides` block completely
     #  once https://github.com/pola-rs/polars/issues/11044 is fixed

py-polars/src/dataframe.rs

@@ -50,42 +50,51 @@ impl PyDataFrame {
     fn finish_from_rows(
         rows: Vec<Row>,
         infer_schema_length: Option<usize>,
-        schema_overwrite: Option<Schema>,
+        schema: Option<Schema>,
+        schema_overrides_by_idx: Option<Vec<(usize, DataType)>>,
     ) -> PyResult<Self> {
         // Object builder must be registered, this is done on import.
-        let schema =
+        let mut final_schema =
             rows_to_schema_supertypes(&rows, infer_schema_length.map(|n| std::cmp::max(1, n)))
                 .map_err(PyPolarsErr::from)?;
+
         // Replace inferred nulls with boolean and erase scale from inferred decimals.
-        let fields = schema.iter_fields().map(|mut fld| match fld.data_type() {
-            DataType::Null => {
-                fld.coerce(DataType::Boolean);
-                fld
-            },
-            DataType::Decimal(_, _) => {
-                fld.coerce(DataType::Decimal(None, None));
-                fld
-            },
-            _ => fld,
-        });
-        let mut schema = Schema::from_iter(fields);
+        for dtype in final_schema.iter_dtypes_mut() {
+            match dtype {
+                DataType::Null => *dtype = DataType::Boolean,
+                DataType::Decimal(_, _) => *dtype = DataType::Decimal(None, None),
+                _ => (),
+            }
+        }
 
-        if let Some(schema_overwrite) = schema_overwrite {
-            for (i, (name, dtype)) in schema_overwrite.into_iter().enumerate() {
-                if let Some((name_, dtype_)) = schema.get_at_index_mut(i) {
+        // Integrate explicit/inferred schema.
+        if let Some(schema) = schema {
+            for (i, (name, dtype)) in schema.into_iter().enumerate() {
+                if let Some((name_, dtype_)) = final_schema.get_at_index_mut(i) {
                     *name_ = name;
 
-                    // If user sets dtype unknown, we use the inferred datatype.
+                    // If schema dtype is Unknown, overwrite with inferred datatype.
                     if !matches!(dtype, DataType::Unknown) {
                         *dtype_ = dtype;
                     }
                 } else {
-                    schema.with_column(name, dtype);
+                    final_schema.with_column(name, dtype);
                 }
             }
         }
 
-        let df = DataFrame::from_rows_and_schema(&rows, &schema).map_err(PyPolarsErr::from)?;
+        // Optional per-field overrides; these supersede default/inferred dtypes.
+        if let Some(overrides) = schema_overrides_by_idx {
+            for (i, dtype) in overrides {
+                if let Some((_, dtype_)) = final_schema.get_at_index_mut(i) {
+                    if !matches!(dtype, DataType::Unknown) {
+                        *dtype_ = dtype;
+                    }
+                }
+            }
+        }
+        let df =
+            DataFrame::from_rows_and_schema(&rows, &final_schema).map_err(PyPolarsErr::from)?;
         Ok(df.into())
     }
 
@@ -512,35 +521,40 @@ impl PyDataFrame {
     pub fn read_rows(
         rows: Vec<Wrap<Row>>,
         infer_schema_length: Option<usize>,
-        schema_overwrite: Option<Wrap<Schema>>,
+        schema: Option<Wrap<Schema>>,
     ) -> PyResult<Self> {
         // SAFETY: Wrap<T> is transparent.
         let rows = unsafe { std::mem::transmute::<Vec<Wrap<Row>>, Vec<Row>>(rows) };
-        Self::finish_from_rows(
-            rows,
-            infer_schema_length,
-            schema_overwrite.map(|wrap| wrap.0),
-        )
+        Self::finish_from_rows(rows, infer_schema_length, schema.map(|wrap| wrap.0), None)
     }
 
     #[staticmethod]
     pub fn read_dicts(
         dicts: &PyAny,
         infer_schema_length: Option<usize>,
-        schema_overwrite: Option<Wrap<Schema>>,
+        schema: Option<Wrap<Schema>>,
+        schema_overrides: Option<Wrap<Schema>>,
     ) -> PyResult<Self> {
         // If given, read dict fields in schema order.
         let mut schema_columns = PlIndexSet::new();
-        if let Some(schema) = &schema_overwrite {
-            schema_columns.extend(schema.0.iter_names().map(|n| n.to_string()))
+        if let Some(s) = &schema {
+            schema_columns.extend(s.0.iter_names().map(|n| n.to_string()))
         }
         let (rows, names) = dicts_to_rows(dicts, infer_schema_length, schema_columns)?;
+        let mut schema_overrides_by_idx: Vec<(usize, DataType)> = Vec::new();
+        if let Some(overrides) = schema_overrides {
+            for (idx, name) in names.iter().enumerate() {
+                if let Some(dtype) = overrides.0.get(name) {
+                    schema_overrides_by_idx.push((idx, dtype.clone()));
+                }
+            }
+        }
         let mut pydf = Self::finish_from_rows(
             rows,
             infer_schema_length,
-            schema_overwrite.map(|wrap| wrap.0),
+            schema.map(|wrap| wrap.0),
+            Some(schema_overrides_by_idx),
         )?;
-
         unsafe {
             for (s, name) in pydf.df.get_columns_mut().iter_mut().zip(&names) {
                 s.rename(name);

py-polars/tests/unit/dataframe/test_df.py

@@ -4,6 +4,7 @@
 import sys
 import textwrap
 import typing
+from collections import OrderedDict
 from datetime import date, datetime, time, timedelta, timezone
 from decimal import Decimal
 from io import BytesIO
@@ -1424,6 +1425,49 @@ def test_from_rows_of_dicts() -> None:
         assert df3.schema == {"id": pl.Int16, "value": pl.Int32}
 
 
+def test_from_records_with_schema_overrides_12032() -> None:
+    # the 'id' fields contains an int value that exceeds Int64 and doesn't have an exact
+    # Float64 representation; confirm that the override is applied *during* inference,
+    # not as a post-inference cast, so we maintain the accuracy of the original value.
+    rec = [
+        {"id": 9187643043065364490, "x": 333, "y": None},
+        {"id": 9223671840084328467, "x": 666.5, "y": 1698177261953686},
+        {"id": 9187643043065364505, "x": 999, "y": 9223372036854775807},
+    ]
+    df = pl.from_records(rec, schema_overrides={"x": pl.Float32, "id": pl.UInt64})
+    assert df.schema == OrderedDict(
+        [
+            ("id", pl.UInt64),
+            ("x", pl.Float32),
+            ("y", pl.Int64),
+        ]
+    )
+    assert rec == df.rows(named=True)
+
+
+def test_from_large_uint64_misc() -> None:
+    uint_data = [[9187643043065364490, 9223671840084328467, 9187643043065364505]]
+
+    df = pl.DataFrame(uint_data, orient="col", schema_overrides={"column_0": pl.UInt64})
+    assert df["column_0"].dtype == pl.UInt64
+    assert df["column_0"].to_list() == uint_data[0]
+
+    for overrides in ({}, {"column_1": pl.UInt64}):
+        df = pl.DataFrame(
+            uint_data,
+            orient="row",
+            schema_overrides=overrides,  # type: ignore[arg-type]
+        )
+        assert df.schema == OrderedDict(
+            [
+                ("column_0", pl.Int64),
+                ("column_1", pl.UInt64),
+                ("column_2", pl.Int64),
+            ]
+        )
+        assert df.row(0) == tuple(uint_data[0])
+
+
 def test_repeat_by_unequal_lengths_panic() -> None:
     df = pl.DataFrame(
         {