----- Original Message -----
From: "Antonio Goncalves"
<antonio.mailing(a)gmail.com>
To: "forge-dev List" <forge-dev(a)lists.jboss.org>
Sent: Sunday, October 20, 2013 10:20:49 PM
Subject: [forge-dev] Wondering about coding convention #philosophy
Hi all,
I love Forge because it generates code. And that's why my customers start to
love it too. Basically, they look at Forge as the "way to write Java EE
code" or if you like "if those guys write code like this, then we
should".
I am digging into some details of the generated code (I am writing a blog
about several architectural styles starting with Forge) and I feel coding
convention should be homogenized . I know extensions are written by
different individuals, but some basic coding conventions should be applied.
+1.
I would consider these architectural styles more than coding conventions,
and I believe it would be wise to 'standardize' some of these for Java EE 6+.
For example, when you generate a web app with REST and Faces
scaffolding,
you get some difference :
* Faces Backing Bean use query builder (e.g getAll method is
entityManager.createQuery(criteria.select(criteria.from(Book.class))).getResultList();
and
* REST Endpoint use dynamic queries (the list all method is "SELECT
DISTINCT b FROM Book b ORDER BY b.id "))
Method names are different and do the same :
* JSF : getAll
* REST : listAll
They do look similar, but they're not the same in all respects.
The one in REST will contain JOIN FETCH expressions when relationships (@OneToMany,
@ManyToMany etc.) are found in the class.
This was done to ensure that serialization would not emit incomplete object graphs.
There are various factors that would be at work here in determining what JPQL queries
should be generated.
Obviously in this context, exposing JPA entities directly is bad idea and
projections/views of the entities should be used instead.
IMHO we should not be putting persistence concerns in either the JSF beans or the REST
resources.
They should go into a service or a repository or whatever data access pattern is suitable
for the context.
This is where we lack any standardization at the moment, and it would be better to not
limit this exercise to improving the conventions alone, but also the architecture.
I don't believe in packing in persistence concerns however small, into these beans for
* we run the risk of generating God classes, and
* we'd leave users with the task of creating/extracting the persistence layer (which
should have been done by Forge in the first place).
Attributes
* private EntityManager em;
* private EntityManager entityManager; // em would be better
Or the use of this keyword (JSF beans use this.entityManager instead of
directly em in REST)
And there are several examples like this. If Forge is seen as "the way of
writing code" maybe something should be created to get homogenized code.
PMD, Checkstyle, human review and so one.....
Just wondering....
--
Antonio Goncalves
Software architect and Java Champion
Web site | Twitter | LinkedIn | Paris JUG | Devoxx France
_______________________________________________
forge-dev mailing list
forge-dev(a)lists.jboss.org
https://lists.jboss.org/mailman/listinfo/forge-dev