dataSet.GetXml() doesn't return xml for null or blank columns

When I call dataSet.GetXml() I don't get any xml returned for columns with null or blank values. Is there a simple, efficient way to get around this? An example of the problem below. Notice how a2 is missing from the second results section.

<results>
<a1>test1</a1>
<a2>test2</a2>
<a3>te开发者_如何学Cst3</a3>
</results>
<results>
<a1>Atest1</a1>
<a3>Atest3</a3>
</results>

The problem is detailed in this Microsoft KB article: http://support.microsoft.com/default.aspx?scid=kb;EN-US;Q317961. See this previous SO question for more detail: DataSet.GetXml not returning null results.

I don't think there is a good solution to your direct question. Given context, there may be another way to approach the problem though.

One solution that worked for me.

First clone the DataTable, make all columns of type string, replace all null values with string.empty, then call GetXml on a new DataSet.

        DataTable dtCloned = dt.Clone();
        foreach (DataColumn dc in dtCloned.Columns)
            dc.DataType = typeof(string);
        foreach (DataRow row in dt.Rows)
        {
            dtCloned.ImportRow(row);
        }

        foreach (DataRow row in dtCloned.Rows)
        {
            for (int i = 0; i < dtCloned.Columns.Count; i++)
            {
                dtCloned.Columns[i].ReadOnly = false;

                if (string.IsNullOrEmpty(row[i].ToString()))
                    row[i] = string.Empty;
            }
        }

        DataSet ds = new DataSet();
        ds.Tables.Add(dtCloned);
        string xml = ds.GetXml();

I have been searching the whole world for a solution of writing null fields to XML using DataSet.WriteXML(). I found that following works in a performance optimized way. I have created a function for your convenience. Change your dataset tables one after the other by calling the following function and replacing the tables.

    private DataTable GetNullFilledDataTableForXML(DataTable dtSource)
    {
        // Create a target table with same structure as source and fields as strings
        // We can change the column datatype as long as there is no data loaded
        DataTable dtTarget = dtSource.Clone();
        foreach (DataColumn col in dtTarget.Columns)
            col.DataType = typeof(string);

        // Start importing the source into target by ItemArray copying which 
        // is found to be reasonably fast for nulk operations. VS 2015 is reporting
        // 500-525 milliseconds for loading 100,000 records x 10 columns 
        // after null conversion in every cell which may be usable in many
        // circumstances.
        // Machine config: i5 2nd Gen, 8 GB RAM, Windows 7 64bit, VS 2015 Update 1
        int colCountInTarget = dtTarget.Columns.Count;
        foreach (DataRow sourceRow in dtSource.Rows)
        {
            // Get a new row loaded with data from source row
            DataRow targetRow = dtTarget.NewRow();
            targetRow.ItemArray = sourceRow.ItemArray;

            // Update DBNull.Values to empty string in the new (target) row
            // We can safely assign empty string since the target table columns
            // are all of string type
            for (int ctr = 0; ctr < colCountInTarget; ctr++)
                if (targetRow[ctr] == DBNull.Value)
                    targetRow[ctr] = String.Empty;

            // Now add the null filled row to target datatable
            dtTarget.Rows.Add(targetRow);
        }

        // Return the target datatable
        return dtTarget;
    }

            if (ds != null && ds.Tables.Count > 0 && ds.Tables[0].Rows.Count == 0)
            {
                foreach (DataTable dt in ds.Tables)
                {
                    foreach (DataColumn dc in dt.Columns)
                    {
                        dc.DataType = typeof(String);
                    }
                }

                DataRow dr = ds.Tables[0].NewRow();
                for (int i = 0; i < dr.ItemArray.Count(); i++)
                {
                    dr[i] = string.Empty;
                }
                ds.Tables[0].Rows.Add(dr);
            }