Flatten Spark data frame fields structure, via SQL in Java. This fork also supports column names having a hyphe in it (e.g. "col-1"), dots and @-sign.

flatten.java

class Toto
{
    public void Main()
    {
        final DataFrame source = GetDataFrame();
        final String querySelectSQL = flattenSchema(source.schema(), null);

        source.registerTempTable("source"); 
      
        final DataFrame flattenData = sqlContext.sql("SELECT " + querySelectSQL + " FROM source")
    }
    
   /**
    * Generate SQL to select columns as flat.
    */
   public static String flattenSchema(StructType schema, String prefix) {
        final StringBuilder selectSQLQuery = new StringBuilder();

        for (StructField field : schema.fields()) {

            final String fieldName = field.name();

            String colName = prefix == null ? (fieldName.contains("-") || fieldName.contains("@") || fieldName.contains(".") ? "`" + fieldName + "`" : fieldName) : (prefix + "." + (fieldName.contains("-") || fieldName.contains("@") || fieldName.contains(".")? "`" + fieldName + "`" : fieldName));
            String colNameTarget = colName.replace(".", "_").replace("-", "").replace("`", "").replace("@", "");

            if (field.dataType().getClass().equals(StructType.class)) {
                selectSQLQuery.append(flattenSchema((StructType) field.dataType(), colName));
            } else {
                selectSQLQuery.append(colName);
                selectSQLQuery.append(" as ");
                selectSQLQuery.append(colNameTarget);
            }

            selectSQLQuery.append(",");
        }

        if (selectSQLQuery.length() > 0) {
            selectSQLQuery.deleteCharAt(selectSQLQuery.length() - 1);
        }

        return selectSQLQuery.toString();
    }
}