HOWTO: Retrieving Multiple Resultsets with Varying Columns

• Microsoft Visual C++, 32-bit Enterprise Edition, version 5.0

SUMMARY

Using the Microsoft Foundation Class (MFC) Open Database Connectivity (ODBC) database classes, you can call a stored procedure that can return multiple resultsets with a varying number of columns. However, some additional work is required.

To retrieve multiple resultsets with a varying number of columns and types, you should use the CRecordset class directly with the GetFieldValue() method. Derived CRecordset classes using RFX functions are not designed to bind to multiple resultsets with a differing number of columns and types, although it may be possible to place conditionals in the DoFieldExchange() function to call different RFX functions for different results.

Why is the additional work required? Consider a stored procedure returning two resultsets; one with one column and the other with two columns. If a Visual C++ MFC ODBC application executes this stored procedure and captures the results, the second resultset returns only one column; in other words, it returns only as many columns as the first resultset returned.

This occurs because the field cache information created by the MFC database classes is not refreshed inside FlushResultSet(). Therefore, the resultset is retrieved with the existing cache information, which was built in AllocAndCacheInfo() in CRecordset::Open(). To properly retrieve the second resultset, call CRecordset::AllocAndCacheFieldInfo() after FlushResultSet() and before MoveNext().

Because CRecordset::AllocAndCacheFieldInfo() is a protected member function, create a new function that does the same thing as AllocAndCacheFieldInfo() from the Dbcore.cpp MFC sources.

MORE INFORMATION

Copy CRecordset::AllocAndCacheFieldInfo() from Dbcore.cpp and include that function in the application. Modify the function as shown below to destroy the field information and do other cleanup. The following is example code in which MyFunction() is calling AllocAndCacheFieldInfo(). The modified AllocAndCacheFieldInfo() function below is based on the Visual C++ 5.0 Enterprise Edition:

   void MyFunction()
   {
      // First resultset has only one column;
      // the second resultset has two columns.

      CString strTemp;
      CDatabase dbase;
      BOOL bRet = dbase.Open("SQLServer", FALSE,FALSE,"ODBC;",FALSE);
      CRecordset recSet(&dbase);
      bRet = recSet.Open(CRecordset::forwardOnly,
                         "{call sp_myProc}",CRecordset::readOnly);

      //Check the number of columns in the first resultset.
      short nCols = recSet.GetODBCFieldCount();

      recSet.GetFieldValue((short)0,strTemp);
      AfxMessageBox(strTemp);

      recSet.FlushResultSet();

      //Rebuild the cache info to retrieve new column info.
      AllocAndCacheFieldInfo(recSet);

      recSet.MoveNext();

      //Check the number of columns in the second resultset.
      nCols = recSet.GetODBCFieldCount();

      recSet.GetFieldValue((short)0,strTemp);
      AfxMessageBox(strTemp);

      recSet.MoveNext();
      recSet.GetFieldValue(1,strTemp);
      AfxMessageBox(strTemp);
   }


   void CCurLib2View::AllocAndCacheFieldInfo(CRecordset &pSet)
   {
      pSet.m_nResultCols = -1;
      ASSERT(pSet.GetODBCFieldCount() < 0);

      if(pSet.m_rgODBCFieldInfos != NULL)
         delete [] pSet.m_rgODBCFieldInfos;


      RETCODE nRetCode;
      SWORD nActualLen;

      // Cache the number of result columns.
      AFX_ODBC_CALL(::SQLNumResultCols(pSet.m_hstmt, &pSet.m_nResultCols));
      if (!pSet.Check(nRetCode))
      {
         TRACE0("Error: Can't get field info.\n");
         AfxThrowDBException(nRetCode, pSet.m_pDatabase,pSet.m_hstmt);
      }

      // If there are no fields, quit now.
      if (pSet.m_nResultCols == 0)
         return;

      // Allocate buffer and get the ODBC meta data.
      pSet.m_rgODBCFieldInfos = new CODBCFieldInfo[pSet.m_nResultCols];
      LPSTR lpszFieldName;

   #ifdef _UNICODE
      // Need proxy to temporarily store non-UNICODE name.
      lpszFieldName = new char[MAX_FNAME_LEN + 1];
   #endif

      // Get the field info for each field.
      for (WORD n = 1; n <= pSet.GetODBCFieldCount(); n++)
      {
   #ifndef _UNICODE
     // Reset the buffer to point to next element.
     lpszFieldName =
     pSet.m_rgODBCFieldInfos[n - 1].m_strName.GetBuffer(MAX_FNAME_LEN + 1);
   #endif

         AFX_ODBC_CALL(::SQLDescribeCol(pSet.m_hstmt, n,
            (UCHAR*)lpszFieldName, MAX_FNAME_LEN, &nActualLen,
            &pSet.m_rgODBCFieldInfos[n - 1].m_nSQLType,
            &pSet.m_rgODBCFieldInfos[n - 1].m_nPrecision,
            &pSet.m_rgODBCFieldInfos[n - 1].m_nScale,
            &pSet.m_rgODBCFieldInfos[n - 1].m_nNullability));

   #ifndef _UNICODE
        pSet.m_rgODBCFieldInfos[n - 1].m_strName.ReleaseBuffer(nActualLen);
   #else
         // Copy the proxy data to correct element.
         pSet.m_rgODBCFieldInfos[n - 1].m_strName = lpszFieldName;
   #endif

      }

   #ifdef _UNICODE
      delete[] lpszFieldName;
   #endif
   } 

Additional query words: recordset

Keywords : kbcode kbDatabase kbMFC kbODBC kbVC
Version : WINDOWS:5.0,5.0sp1,5.0sp2,5.0sp3
Platform : WINDOWS
Issue type : kbhowto