respect dtype overwrite when schema is overwritten in lazy csv scanner (

#2915)

polars/polars-lazy/src/frame/csv.rs

@@ -192,13 +192,22 @@ impl<'a> LazyCsvReader<'a> {
             self.delimiter,
             self.infer_schema_length,
             self.has_header,
-            self.schema_overwrite,
+            // we set it to None and modify them after the schema is updated
+            None,
             &mut skip_rows,
             self.comment_char,
             self.quote_char,
             None,
         )?;
-        let schema = f(schema)?;
+        let mut schema = f(schema)?;
+
+        // the dtypes set may be for the new names, so update again
+        if let Some(overwrite_schema) = self.schema_overwrite {
+            for (name, dtype) in overwrite_schema.iter() {
+                schema.with_column(name.clone(), dtype.clone())
+            }
+        }
+
         Ok(self.with_schema(Arc::new(schema)))
     }
 

py-polars/tests/io/test_lazy_csv.py

@@ -45,3 +45,16 @@ def test_row_count(foods_csv: str) -> None:
     )
 
     assert df["foo"].to_list() == [10, 16, 21, 23, 24, 30, 35]
+
+
+def test_scan_csv_schema_overwrite_and_dtypes_overwrite(foods_csv: str) -> None:
+    assert (
+        pl.scan_csv(
+            foods_csv,
+            dtypes={"calories_foo": pl.Utf8, "fats_g_foo": pl.Float32},
+            with_column_names=lambda names: [f"{a}_foo" for a in names],
+        )
+        .collect()
+        .dtypes
+        == [pl.Utf8, pl.Utf8, pl.Float32, pl.Int64]
+    )