steve Barf

ok 2 things :

(1) I realise I didn't got the manager properly in the BeanManagersStaff constructor, fixed below.
The question is still the same - why is the request scope BeanManagersStaff created in the Render_Response(6) phase ? Is there a better way ?

public BeanManagersStaff(){ System.out.println("BeanManagersStaff.constructor"); FacesContext context = FacesContext.getCurrentInstance(); ExternalContext extContext = context.getExternalContext(); Map sessMap = extContext.getSessionMap(); manager=(UserBean)sessMap.get("user"); manager=new UserBean(manager.getContactDetails().getManagerId()); populateStaff(); }

(2) The user id is also used as the password hence the database call :

"SELECT * FROM hr.employees WHERE manager_id="+ manager.getPassword() //ie=manager's user id

I have a userDetails screen based on a managed session scope UserBean. I have created a simple outputLink on the Manager Id label to go to a managerPlusStaff screen :

<h:outputLink value="/Nod/pages/managerPlusStaff.jsf"> <h:outputText value="Manager Id:"/> </h:outputLink>

The managerPlusStaff.jsp is based on a managed request scope BeanManagersStaff bean. The constructor for BeanManagersStaff populates the bean properties (manager+staff) by accessing the session scope UserBean to get the Manager Id.

This all works fine and I can see from a trace that the BeanManagersStaff is created in the Render_Response(6) step. BUT WHY AND IS THIS THE BEST WAY.

BeanManagersStaff below:

package beans; import java.sql.ResultSet; import java.util.List; import java.util.ArrayList; import java.util.Map; import javax.faces.context.ExternalContext; import javax.faces.context.FacesContext; import util.DBHandler; public class BeanManagersStaff { private List staff=new ArrayList(); private UserBean manager; public BeanManagersStaff(){ System.out.println("BeanManagersStaff.constructor"); FacesContext context = FacesContext.getCurrentInstance(); ExternalContext extContext = context.getExternalContext(); Map sessMap = extContext.getSessionMap(); manager=(UserBean)sessMap.get("user"); populateStaff(); } public List getStaff() { return staff; } public void setStaff(List staff) { this.staff = staff; } public UserBean getManager() { return manager; } public void setManager(UserBean manager) { this.manager = manager; } private void populateStaff(){ System.out.println("managersStaff populateStaff start"); FacesContext context = FacesContext.getCurrentInstance(); ExternalContext extContext = context.getExternalContext(); Map sessMap = extContext.getSessionMap(); DBHandler dBHandler = (DBHandler) sessMap.get("DBHandler"); System.out.println("managersStaff populateStaff 2"); try { ResultSet rs = dBHandler .runQuery("SELECT * FROM hr.employees WHERE manager_id=" + manager.getPassword()); System.out.println("managersStaff populateStaff 3"); while (rs.next()){ UserBean u= new UserBean(); u.setEmail(rs.getString("email")); u.setName(rs.getString("last_name")); staff.add(u); } dBHandler.closeQuery(rs); System.out.println("managersStaff populateStaff 4"); } catch (Exception e) { System.out.println(e.toString()); } } }

Or 25 <navigation-case>s ...

Don't know much about JSF yet - being 'urged' to use it by customer. But if one of the benefits is that the 'controller' part of mvc is exposed in the faces-config.xml file - how would one handle navigation from a menu eg. a jscookmenu where you can navigate to say 25 different destinations ? Surely you wouldn't have 25 navigation-rules for each screen ?

<navigation-rule> <from-view-id>/userLogin.jsp</from-view-id <navigation-case> <from-outcome>success</from-outcome> <to-view-id>/userLoginSuccess.jsp</to-view-id> </navigation-case> <navigation-case> <from-outcome>failure</from-outcome> <to-view-id>/userLogin.jsp</to-view-id> </navigation-case> </navigation-rule>

We specifically aimed for JSP's with an absolute minimum of code. This is because I inherited an applicaion with >1000 line JSP's which I swore I would never duplicate.

Each request from a given JSP first went to a corresponding servlet. Processing was performed by the servlet and a javabean with request scope. Shared data was held in session and application beans. The final response was generated by the JSP retrieving stored data in the session(s)/application(a) bean plus derived data in the request(r) scope javabean. ie. the jsp handeled the displaying and not the processing.

Along Model 2 lines.

This'll probably make you all laugh but who cares, in a recent jsp/servlet/bean program I wrote with someone new to java who couldn't keep track of the scope of different variables I ended up prefixing the variables as follows :
application - a
session - s
request - r
instance - i
local - l
and arguments/parameters into methods with 'parm'.

I had to check a piece of the code today not having looked at it since before Christmas and found it very useful.

Go on then laugh.

Ben,
Thanks, that book seems better than others I have read on servlets/jsps.
But I didn't see anything on 'context listeners' - if this is not the same as the ServletContext then this is new to me. Searches on google have resulted in articles relating to EJB's or products that use it, rather than using it in the raw. Could you supply some more guidance please ?

My other option is to use the ServletContext as first mentioned :

appBean myAppBean = new appBean(); myAppBean.getData(); getServletContext().setAttribute("myAppBean");

but to do it in a servlet that runs on server startup.

Steve

Ben,
Thanks for your quick reply. I'd be grateful for your comments on my understanding of your points.

1) a) Could there be a problem that the bean is GCed between creation by the servlet and given application scope in the jsp ?

b) Does the application scope implicitly generate a reference which will therefore prevent GC ?

2) Could you expand on this please or point me to a simple reference.

3) Maybe I could do both. I would like to have a go at a singleton, does the following look suitable :

public class AppBean { public static final AppBean appBean = new AppBean(); private AppBean(){} public void getData() { //get DB Data } public Vector returnData() { // return DB data } }

There is another singleton method I've seen but I can't see how the getInstance() method would fit in with serrvlets/jsps
public class AppBean { private static AppBean appBean = new AppBean(); private AppBean(){} public static AppBean getInstance() { return appBean; } public void getData() { //get Data } }


Steve

I am thinking of writing a bean to hold a quantity of database information to prevent every user of a web application having to retrieve all the data.



But I have a few concerns :

1) what would stop this bean from being garbage collected ?

2) what would be the best way to create the bean ?

something like this ? :
appBean myAppBean = (appBean) getServletContext().getAttribute("myAppBean"); if (myAppBean==null) { appBean myAppBean = new appBean(); myAppBean.getData(); getServletContext().setAttribute("myAppBean"); }
and then reference it in jsp's by :
<jsp:useBean id="myAppBean" scope=""application" class="appBean">

3) should the bean be created as a singleton ?

Steve

I am thinking of writing a bean to hold a quantity of database information to prevent every user of a web application having to retrieve all the data.



But I have a few concerns :

1) what would stop this bean from being garbage collected ?

2) what would be the best way to create the bean ?

something like this ? :
[CODE}
appBean myAppBean = (appBean) getServletContext().getAttribute("myAppBean");
if (myAppBean==null)
{
appBean myAppBean = new appBean();
myAppBean.getData();
getServletContext().setAttribute("myAppBean");
}
[/CODE]
and then reference it in jsp's by :
<jsp:useBean id="myAppBean" scope=""application" class="appBean">

3) should the bean be created as a singleton ?

Steve

which book is this by K&B ?
this one

http://www.pcworld-books.co.uk/catalog/browse.asp?advert=Kelkoo&ref=693405

is not published til sep 2005 and this one

http://www.amazon.co.uk/exec/obidos/ASIN/0596009208/qid=1120576727/sr=2-3/ref=sr_2_3_3/026-8814143-7100451

seems to be treated as a joke

Kevin,
Thanks for your replies - I was wondering if there was a completely different technique. I've seen code using Principals but I'm not clear how it works or if the client needs something on their workstation first. I feel sure other Java users must have been faced with this problem and resisted using IIS.
<BR/><BR/>
My fear with NTLM is :

what happens if the offsets were to change

do all browsers use it

there can be problems if the user sets the security level to high

<BR/><BR/>
Are you familiar with any non NTLM techniques ?
Steve

http://www.jguru.com/faq/view.jsp?EID=1045412

This link shows a technique for getting the users Windows login + domain. But it uses hardcoded offsets. Is there a cleaner way of doing it ?

CL,
I think the problem is a network problem (the SMTP server chap started talking about reverse pings ..),
on odd occasions the smtpclient isn't created properly, and this is what I was trying to handle.

Looking back at your first reply you seem to have separate threads for IO probably for this reason.

I would appreciate being able to see a code example from your website, if it's not too complicated
and if you could direct me to a suitable piece of code I'd be grateful.

I was wondering if the application could have 3 goes at creating the smtpclient, and putting a time limit
of say 2 seconds on each attempt to create the client.

In pseudo-code I was thinking of something like :


clientcreated=false;

loop (3 times while clientcreated=false)

create thread

loop (for 2 seconds while clientcreated=false)
end loop

if (clientcreated=false)
cancel thread

end loop

if (clientcreated=false)
error

thread
{
create smtpclient
clientcreated=true
}


Remembering that the rest of the application is synchronised, this shouldn't hold up other users
except when there is this intermittent problem, and then only for 6 seconds.

But again it may not make sense to have a thread in a synchronised application.

Anybody ?