I'm using a ResultSet in Java, and am not sure how to properly close it. I'm considering using the ResultSet to construct a HashMap and then closing the ResultSet after that. Is this HashMap technique efficient, or are there more efficient ways of h开发者_C百科andling this situation? I need both keys and values, so using a HashMap seemed like a logical choice.
If using a HashMap is the most efficient method, how do I construct and use the HashMap in my code?
Here's what I've tried:
public HashMap resultSetToHashMap(ResultSet rs) throws SQLException {
ResultSetMetaData md = rs.getMetaData();
int columns = md.getColumnCount();
HashMap row = new HashMap();
while (rs.next()) {
for (int i = 1; i <= columns; i++) {
row.put(md.getColumnName(i), rs.getObject(i));
}
}
return row;
}
- Iterate over the ResultSet
- Create a new Object for each row, to store the fields you need
- Add this new object to ArrayList or Hashmap or whatever you fancy
- Close the ResultSet, Statement and the DB connection
Done
EDIT: now that you have posted code, I have made a few changes to it.
public List resultSetToArrayList(ResultSet rs) throws SQLException{
ResultSetMetaData md = rs.getMetaData();
int columns = md.getColumnCount();
ArrayList list = new ArrayList(50);
while (rs.next()){
HashMap row = new HashMap(columns);
for(int i=1; i<=columns; ++i){
row.put(md.getColumnName(i),rs.getObject(i));
}
list.add(row);
}
return list;
}
I just cleaned up RHT's answer to eliminate some warnings and thought I would share. Eclipse did most of the work:
public List<HashMap<String,Object>> convertResultSetToList(ResultSet rs) throws SQLException {
ResultSetMetaData md = rs.getMetaData();
int columns = md.getColumnCount();
List<HashMap<String,Object>> list = new ArrayList<HashMap<String,Object>>();
while (rs.next()) {
HashMap<String,Object> row = new HashMap<String, Object>(columns);
for(int i=1; i<=columns; ++i) {
row.put(md.getColumnName(i),rs.getObject(i));
}
list.add(row);
}
return list;
}
RHT pretty much has it. Or you could use a RowSetDynaClass and let someone else do all the work :)
this is my alternative solution, instead of a List of Map, i'm using a Map of List. Tested on tables of 5000 elements, on a remote db, times are around 350ms for eiter method.
private Map<String, List<Object>> resultSetToArrayList(ResultSet rs) throws SQLException {
ResultSetMetaData md = rs.getMetaData();
int columns = md.getColumnCount();
Map<String, List<Object>> map = new HashMap<>(columns);
for (int i = 1; i <= columns; ++i) {
map.put(md.getColumnName(i), new ArrayList<>());
}
while (rs.next()) {
for (int i = 1; i <= columns; ++i) {
map.get(md.getColumnName(i)).add(rs.getObject(i));
}
}
return map;
}
A couple of things to enhance the other answers. First, you should never return a HashMap
, which is a specific implementation. Return instead a plain old java.util.Map
. But that's actually not right for this example, anyway. Your code only returns the last row of the ResultSet as a (Hash)Map. You instead want to return a List<Map<String,Object>>
. Think about how you should modify your code to do that. (Or you could take Dave Newton's suggestion).
i improved the solutions of RHTs/Brad Ms and of Lestos answer.
i extended both solutions in leaving the state there, where it was found. So i save the current ResultSet position and restore it after i created the maps.
The rs is the ResultSet, its a field variable and so in my solutions-snippets not visible.
I replaced the specific Map in Brad Ms solution to the gerneric Map.
public List<Map<String, Object>> resultAsListMap() throws SQLException
{
var md = rs.getMetaData();
var columns = md.getColumnCount();
var list = new ArrayList<Map<String, Object>>();
var currRowIndex = rs.getRow();
rs.beforeFirst();
while (rs.next())
{
HashMap<String, Object> row = new HashMap<String, Object>(columns);
for (int i = 1; i <= columns; ++i)
{
row.put(md.getColumnName(i), rs.getObject(i));
}
list.add(row);
}
rs.absolute(currRowIndex);
return list;
}
In Lestos solution, i optimized the code. In his code he have to lookup the Maps each iteration of that for-loop. I reduced that to only one array-acces each for-loop iteration. So the program must not seach each iteration step for that string-key.
public Map<String, List<Object>> resultAsMapList() throws SQLException
{
var md = rs.getMetaData();
var columns = md.getColumnCount();
var tmp = new ArrayList[columns];
var map = new HashMap<String, List<Object>>(columns);
var currRowIndex = rs.getRow();
rs.beforeFirst();
for (int i = 1; i <= columns; ++i)
{
tmp[i - 1] = new ArrayList<>();
map.put(md.getColumnName(i), tmp[i - 1]);
}
while (rs.next())
{
for (int i = 1; i <= columns; ++i)
{
tmp[i - 1].add(rs.getObject(i));
}
}
rs.absolute(currRowIndex);
return map;
}
Here is the code little modified that i got it from google -
List data_table = new ArrayList<>();
Class.forName("oracle.jdbc.driver.OracleDriver");
con = DriverManager.getConnection(conn_url, user_id, password);
Statement stmt = con.createStatement();
System.out.println("query_string: "+query_string);
ResultSet rs = stmt.executeQuery(query_string);
ResultSetMetaData rsmd = rs.getMetaData();
int row_count = 0;
while (rs.next()) {
HashMap<String, String> data_map = new HashMap<>();
if (row_count == 240001) {
break;
}
for (int i = 1; i <= rsmd.getColumnCount(); i++) {
data_map.put(rsmd.getColumnName(i), rs.getString(i));
}
data_table.add(data_map);
row_count = row_count + 1;
}
rs.close();
stmt.close();
con.close();
public static List<HashMap<Object, Object>> GetListOfDataFromResultSet(ResultSet rs) throws SQLException {
ResultSetMetaData metaData = rs.getMetaData();
int count = metaData.getColumnCount();
String[] columnName = new String[count];
List<HashMap<Object,Object>> lst=new ArrayList<>();
while(rs.next()) {
HashMap<Object,Object> map=new HashMap<>();
for (int i = 1; i <= count; i++){
columnName[i-1] = metaData.getColumnLabel(i);
map.put(columnName[i-1], rs.getObject(i));
}
lst.add(map);
}
return lst;
}
精彩评论