fix: round integer values during Excel and CSV import (#1365)

Author: xuwei-fit2cloud

backend/apps/datasource/api/datasource.py

@@ -31,7 +31,7 @@
 from ..crud.table import get_tables_by_ds_id
 from ..models.datasource import CoreDatasource, CreateDatasource, TableObj, CoreTable, CoreField, FieldObj, \
     TableSchemaResponse, ColumnSchemaResponse, PreviewResponse, ImportRequest
-from ..utils.excel import parse_excel_preview, USER_TYPE_TO_PANDAS
+from ..utils.excel import parse_excel_preview, read_import_dataframe
 
 router = APIRouter(tags=["Datasource"], prefix="/datasource")
 path = settings.EXCEL_PATH
@@ -576,17 +576,10 @@ def inner():
             fields = sheet_info.fields
 
             field_mapping = {f.fieldName: f.fieldType for f in fields}
-            dtype_dict = {
-                col: USER_TYPE_TO_PANDAS.get(field_mapping.get(col, 'string'), 'string')
-                for col in field_mapping.keys()
-            }
-
             try:
+                df = read_import_dataframe(save_path, sheet_name, field_mapping)
                 if save_path.endswith(".csv"):
-                    df = pd.read_csv(save_path, engine='c', dtype=dtype_dict)
                     sheet_name = "Sheet1"
-                else:
-                    df = pd.read_excel(save_path, sheet_name=sheet_name, engine='calamine', dtype=dtype_dict)
             except Exception as e:
                 raise HTTPException(500, f"{trans('i18n_ds_upload_error')}: {str(e)}")
 

backend/apps/datasource/utils/excel.py

@@ -1,3 +1,5 @@
+from decimal import Decimal, InvalidOperation, ROUND_HALF_UP
+
 import pandas as pd
 
 FIELD_TYPE_MAP = {
@@ -25,6 +27,45 @@ def infer_field_type(dtype) -> str:
     return FIELD_TYPE_MAP.get(dtype_str, 'string')
 
 
+def _round_import_integer(value):
+    """Round explicitly selected integer values without a float intermediate."""
+    if pd.isna(value) or (isinstance(value, str) and not value.strip()):
+        return pd.NA
+    if isinstance(value, bool):
+        return int(value)
+    try:
+        number = Decimal(str(value))
+    except InvalidOperation as exc:
+        raise ValueError('Invalid numeric value for integer field') from exc
+    if not number.is_finite():
+        raise ValueError('Integer fields require finite numeric values')
+    rounded = number.to_integral_value(rounding=ROUND_HALF_UP)
+    if rounded < -(2 ** 63) or rounded > 2 ** 63 - 1:
+        raise ValueError('Rounded value is outside the signed 64-bit integer range')
+    return int(rounded)
+
+
+def read_import_dataframe(save_path: str, sheet_name: str, field_mapping: dict):
+    """Read stored cell values; round only columns explicitly selected as integers."""
+    dtype_dict = {
+        column: 'object' if field_type == 'int' else USER_TYPE_TO_PANDAS.get(field_type, 'string')
+        for column, field_type in field_mapping.items()
+    }
+    if save_path.endswith('.csv'):
+        df = pd.read_csv(save_path, engine='c', dtype=dtype_dict)
+    else:
+        df = pd.read_excel(save_path, sheet_name=sheet_name, engine='calamine', dtype=dtype_dict)
+    for column, field_type in field_mapping.items():
+        if field_type == 'int' and column in df.columns:
+            try:
+                # Build a nullable integer array directly: Series.map can coerce
+                # large integers plus missing values to lossy float64 values.
+                df[column] = pd.array([_round_import_integer(value) for value in df[column]], dtype='Int64')
+            except ValueError as exc:
+                raise ValueError(f"Column '{column}': {exc}") from exc
+    return df
+
+
 def parse_excel_preview(save_path: str, max_rows: int = 10):
     sheets_data = []
     if save_path.endswith(".csv"):