FIX: Cursor.describe invalid data

mssql_python/constants.py

@@ -115,6 +115,7 @@ class ConstantsDDBC(Enum):
     SQL_FETCH_RELATIVE = 6
     SQL_FETCH_BOOKMARK = 8
     SQL_DATETIMEOFFSET = -155
+    SQL_SS_UDT = -151  # SQL Server User-Defined Types (geometry, geography, hierarchyid)
     SQL_C_SS_TIMESTAMPOFFSET = 0x4001
     SQL_SCOPE_CURROW = 0
     SQL_BEST_ROWID = 1
@@ -502,3 +503,5 @@ def get_attribute_set_timing(attribute):
     # internally.
     "packetsize": "PacketSize",
 }
+
+# (Function removed; no replacement needed)

mssql_python/cursor.py

@@ -16,6 +16,7 @@
 import datetime
 import warnings
 from typing import List, Union, Any, Optional, Tuple, Sequence, TYPE_CHECKING
+import xml
 from mssql_python.constants import ConstantsDDBC as ddbc_sql_const, SQLTypes
 from mssql_python.helpers import check_error
 from mssql_python.logging import logger
@@ -142,6 +143,9 @@ def __init__(self, connection: "Connection", timeout: int = 0) -> None:
         )
         self.messages = []  # Store diagnostic messages
 
+        # Store raw column metadata for converter lookups
+        self._column_metadata = None
+
     def _is_unicode_string(self, param: str) -> bool:
         """
         Check if a string contains non-ASCII characters.
@@ -942,8 +946,12 @@ def _initialize_description(self, column_metadata: Optional[Any] = None) -> None
         """Initialize the description attribute from column metadata."""
         if not column_metadata:
             self.description = None
+            self._column_metadata = None  # Clear metadata too
             return
 
+        # Store raw metadata for converter map building
+        self._column_metadata = column_metadata
+
         description = []
         for _, col in enumerate(column_metadata):
             # Get column name - lowercase it if the lowercase flag is set
@@ -957,7 +965,7 @@ def _initialize_description(self, column_metadata: Optional[Any] = None) -> None
             description.append(
                 (
                     column_name,  # name
-                    self._map_data_type(col["DataType"]),  # type_code
+                    col["DataType"],  # type_code (SQL type integer) - CHANGED THIS LINE
                     None,  # display_size
                     col["ColumnSize"],  # internal_size
                     col["ColumnSize"],  # precision - should match ColumnSize
@@ -975,18 +983,17 @@ def _build_converter_map(self):
         """
         if (
             not self.description
+            or not self._column_metadata
             or not hasattr(self.connection, "_output_converters")
             or not self.connection._output_converters
         ):
             return None
 
         converter_map = []
 
-        for desc in self.description:
-            if desc is None:
-                converter_map.append(None)
-                continue
-            sql_type = desc[1]
+        for col_meta in self._column_metadata:
+            # Use the raw SQL type code from metadata, not the mapped Python type
+            sql_type = col_meta["DataType"]
             converter = self.connection.get_output_converter(sql_type)
             # If no converter found for the SQL type, try the WVARCHAR converter as a fallback
             if converter is None:
@@ -1053,6 +1060,11 @@ def _map_data_type(self, sql_type):
             ddbc_sql_const.SQL_VARBINARY.value: bytes,
             ddbc_sql_const.SQL_LONGVARBINARY.value: bytes,
             ddbc_sql_const.SQL_GUID.value: uuid.UUID,
+            ddbc_sql_const.SQL_SS_UDT.value: bytes,  # UDTs mapped to bytes
+            ddbc_sql_const.SQL_XML.value: str,  # XML mapped to str
+            ddbc_sql_const.SQL_DATETIME2.value: datetime.datetime,
+            ddbc_sql_const.SQL_SMALLDATETIME.value: datetime.datetime,
+            ddbc_sql_const.SQL_DATETIMEOFFSET.value: datetime.datetime,
             # Add more mappings as needed
         }
         return sql_to_python_type.get(sql_type, str)
@@ -2572,7 +2584,6 @@ def __del__(self):
         Destructor to ensure the cursor is closed when it is no longer needed.
         This is a safety net to ensure resources are cleaned up
         even if close() was not called explicitly.
-        If the cursor is already closed, it will not raise an exception during cleanup.
         """
         if "closed" not in self.__dict__ or not self.closed:
             try:

mssql_python/pybind/ddbc_bindings.cpp

@@ -27,6 +27,9 @@
 #define MAX_DIGITS_IN_NUMERIC 64
 #define SQL_MAX_NUMERIC_LEN 16
 #define SQL_SS_XML (-152)
+#define SQL_SS_UDT (-151)  // SQL Server User-Defined Types (geometry, geography, hierarchyid)
+#define SQL_DATETIME2 (42)
+#define SQL_SMALLDATETIME (58)
 
 #define STRINGIFY_FOR_CASE(x)                                                                      \
     case x:                                                                                        \
@@ -2987,6 +2990,11 @@ SQLRETURN SQLGetData_wrap(SqlHandlePtr StatementHandle, SQLUSMALLINT colCount, p
                 }
                 break;
             }
+            case SQL_SS_UDT: {
+                LOG("SQLGetData: Streaming UDT (geometry/geography) for column %d", i);
+                row.append(FetchLobColumnData(hStmt, i, SQL_C_BINARY, false, true));
+                break;
+            }
             case SQL_SS_XML: {
                 LOG("SQLGetData: Streaming XML for column %d", i);
                 row.append(FetchLobColumnData(hStmt, i, SQL_C_WCHAR, true, false, "utf-16le"));
@@ -3211,6 +3219,8 @@ SQLRETURN SQLGetData_wrap(SqlHandlePtr StatementHandle, SQLUSMALLINT colCount, p
             }
             case SQL_TIMESTAMP:
             case SQL_TYPE_TIMESTAMP:
+            case SQL_DATETIME2:
+            case SQL_SMALLDATETIME:
             case SQL_DATETIME: {
                 SQL_TIMESTAMP_STRUCT timestampValue;
                 ret = SQLGetData_ptr(hStmt, i, SQL_C_TYPE_TIMESTAMP, &timestampValue,
@@ -3795,6 +3805,8 @@ SQLRETURN FetchBatchData(SQLHSTMT hStmt, ColumnBuffers& buffers, py::list& colum
                 }
                 case SQL_TIMESTAMP:
                 case SQL_TYPE_TIMESTAMP:
+                case SQL_DATETIME2:
+                case SQL_SMALLDATETIME:
                 case SQL_DATETIME: {
                     const SQL_TIMESTAMP_STRUCT& ts = buffers.timestampBuffers[col - 1][i];
                     PyObject* datetimeObj = PythonObjectCache::get_datetime_class()(
@@ -3974,6 +3986,9 @@ size_t calculateRowSize(py::list& columnNames, SQLUSMALLINT numCols) {
             case SQL_SS_TIMESTAMPOFFSET:
                 rowSize += sizeof(DateTimeOffset);
                 break;
+            case SQL_SS_UDT:
+                rowSize += columnSize;  // UDT types use column size as-is
+                break;
             default:
                 std::wstring columnName = columnMeta["ColumnName"].cast<std::wstring>();
                 std::ostringstream errorString;
@@ -4028,7 +4043,7 @@ SQLRETURN FetchMany_wrap(SqlHandlePtr StatementHandle, py::list& rows, int fetch
 
         if ((dataType == SQL_WVARCHAR || dataType == SQL_WLONGVARCHAR || dataType == SQL_VARCHAR ||
              dataType == SQL_LONGVARCHAR || dataType == SQL_VARBINARY ||
-             dataType == SQL_LONGVARBINARY || dataType == SQL_SS_XML) &&
+             dataType == SQL_LONGVARBINARY || dataType == SQL_SS_XML || dataType == SQL_SS_UDT) &&
             (columnSize == 0 || columnSize == SQL_NO_TOTAL || columnSize > SQL_MAX_LOB_SIZE)) {
             lobColumns.push_back(i + 1);  // 1-based
         }
@@ -4162,7 +4177,7 @@ SQLRETURN FetchAll_wrap(SqlHandlePtr StatementHandle, py::list& rows,
 
         if ((dataType == SQL_WVARCHAR || dataType == SQL_WLONGVARCHAR || dataType == SQL_VARCHAR ||
              dataType == SQL_LONGVARCHAR || dataType == SQL_VARBINARY ||
-             dataType == SQL_LONGVARBINARY || dataType == SQL_SS_XML) &&
+             dataType == SQL_LONGVARBINARY || dataType == SQL_SS_XML || dataType == SQL_SS_UDT) &&
             (columnSize == 0 || columnSize == SQL_NO_TOTAL || columnSize > SQL_MAX_LOB_SIZE)) {
             lobColumns.push_back(i + 1);  // 1-based
         }