The Design and Implementation of laffer1

 I've been working on converting just journal over to Apache Cayenne today.  I figured it would give me an opportunity to get better with it.  Cayenne has a nice modeling program, but their documentation is often lacking.

For example, suppose you're getting an error from your servlet container (Tomcat in my case) about a filter not loading.  With Cayenne, the "correct" way to use it is to load a filter mapped to /* that catches all the sessions and maps a DataContext to them.  You might find after pulling your hair out that there is a syntax error in your DomainNode.driver.xml file.  Even with logging cranked up in log4j, it will not tell you what's wrong with the filter.  However, if you initialize like a traditional app in a jsp page, you can see the error.  Invalid passwords are an example of what can cause this.

I've been using JSON with Dojo Data Grids lately in PHP.  Now, I want to use something similar on this site.  I decided on the json-simple library.  It's actually quite easy to use.

Encoding examples:

http://code.google.com/p/json-simple/wiki/EncodingExamples 

<%@ page contentType="text/json;charset=UTF-8" language="java"%>

<%@ page import="com.justjournal.model.User" %>

<%@ page import="org.apache.cayenne.DataObjectUtils" %>

<%@ page import="org.apache.cayenne.access.DataContext" %>

<%@ page import="org.apache.cayenne.query.SelectQuery" %>

<%@ page import="org.json.simple.JSONObject" %>

<%@ page import="java.util.*" %>

<%@ page import="org.json.simple.JSONArray" %>

<%

    out.clearBuffer();

    SelectQuery query = new SelectQuery(User.class);

    query.addOrdering(User.USERNAME_PROPERTY, true);

    DataContext context = DataContext.getThreadDataContext();

    List users = context.performQuery(query);

    JSONObject obj = new JSONObject();

    JSONArray list1 = new JSONArray();

    Iterator it = users.iterator();

    while (it.hasNext()) {

        Map map = new HashMap();

        User u = (User) it.next();

        if (u.getUsername().length() > 2 && !new com.justjournal.User(u.getUsername()).isPrivateJournal())

        {

            map.put("id", DataObjectUtils.intPKForObject(u));

            map.put("username", u.getUsername());

            map.put("name", u.getName());

            map.put("since", u.getSince());

            list1.add(map);

        }

    }

    obj.put("items", list1);

    obj.put("label","id");

    obj.put("identifier", "id");

    obj.writeJSONString(out);

I've been experimenting with Apache Cayenne for work.  Earlier today, I got stumped by a complex database relationship in terms of Cayenne.  I can easily write a six table join to accomplish the query, but i wanted to find an object mapping approach instead.

I've learned several things about Cayenne.  First, it does not support outer joins from the built in object mapping.  It's possible to create outer and left joins with either SQLTemplate or performing a JDBC query directly.  

A SQLTemplate is a method to write a "real" SQL statement and either execute it as a non returning query or a select mapped to an existing Cayenne object.   You can even have it return raw rows back.  My first thought was that it would suck to do that because you'd lose the auto generating code for multiple database feature.  Actually, you can write alternate manual queries and register them for each major JDBC provider.  You write a generic default one that can work in most cases and then a tuned version (or whatever) for a specific database.   

The other option is to write a direct JDBC query and circumvent most of Cayenne.  You can still use it's database connection pool though.

DataSource ds = context.getParentDataDomain().getNode ("mydatanode").getDataSource(); Connection c = ds.getConnection();

 Today was an interesting day at work.  I managed to get Cayenne working properly for a Project -> User mapping through some intermediate tables.  There were few tricks involved.  First, I had to modify the User class within Cayenne to explicitly have a mapping for the Primary key.  This allows you to do an Expression.fromString and refer to the primary key.  Second, I had to do a double mapping using the relationship name between the project and the mapping table then the mapping table to user table relationship.  The context had to be the name from the project end and from the mapping table end to the user table.  (left to right... )  

One problem I had early on was that we also have a user_id stored as the "owner" with a relationship in the project table.  That means there's a direct relationship and an indirect relationship between project and user.  

The idea here was to map all the projects that the User can access.  The mapping table contains a PK, user_id and project_id.  A user can potentially access many projects.  We knew the user id of the user logged in from the session (username would work too and require less work, but could not benefit from the database indexes).  

This sounds trivial, but there is almost no documentation on something like this with Cayenne.  Those apache developers need to get up some real examples of Cayenne apps.  

I just read that Cayenne 3.0 RC2 is out.  Cayenne is an ORM.   I've been using it for almost a year now.  It's a very easy way to do data access in Java.  There are a few quirks.  Most people love Hibernate and don't consider alternatives.  Cayenne is very easy to work with.  It has a client gui to configure and setup your mappings or you can write an XML file by hand.