Hi Pinaki,

On Feb 26, 2007, at 10:01 AM, Pinaki Poddar wrote:

Respected Sir,
Sorry to disagree. If you fetch Department instances, I think you always want the _employees collection to contain all the related Employees. You never want a filtered collection unless you provide a filter via some Java behavior.
I agree that department.getEmployees() must return all the employees.

However, i was considering the state of a Department instance in memory that has just been selected by a query. Which of the native (String/int
etc.) field values are populated and which of its related objects are
instantiated is controlled by the currently active fetch plan (if not
explicitly then at least implicitly by 'default' fetch group).

Right.

If the application accesses any unrealized field  of this 'partially
filled' department, then that field is populated irrespective of whether
it is in active fetch group or not via a new, separate database access
call.

Right.

In a sense, the fetch plan determines  a 'partial' view of the
'complete' department -- both in terms of its native state and its
relations -- as and when it appears for the first time in memory. In
fact, if the department instance leaves the managed environment via
detachment at that point -- the detached department in a remote process
will only have this partially filled 'view'. And, of course, the
application can explictly load any unrealized state of the partial view
via any getter method.

Right.

I was trying to draw an equivalence/parallel of this normal fetch
behavior to the use case where a owner object has a high,
multi-cardinalty relationship but only wants a subset of the elements of
that relationship. In such a scheme, the traversal on a fetch path
relation (which is now a binary decision) can be further adorned to
support a predicate i.e.
Query query = // a JPQL query that selects a Department based on some
critria
query.getFetchPlan().addField("employees", "yearsOfSevice>10");
Department dept = query.getSingleResult();
// dept will be constructed in memory with a collection of Employees
with yearsOfService>10

The standard way to get the employees with the filter you show is to explicitly return the employee instances. The query language does not accommodate returning the collection as a result value. One of the reasons to exclude this is to disambiguate the result collection as being filtered or not. Disallowing return values of collection types neatly avoids the issue.

dept.getEmployees();
// now the dept will load a collection of all Employees

Right.

And you can also envision filtering the collection directly via non- standard techniques. Given the collection, that might not be fully instantiated, you can pass that collection to a proprietary method that does the filtering in the database.

Of course, the suggested alternative approach where the application
explictly constructs this partial/filtered multi-cardinality relatinship
by choosing the elements via query is feasible and available right now
with current features.

Right.

Craig


Pinaki Poddar
BEA Systems
415.402.7317


-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Friday, February 23, 2007 8:32 PM
To: open-jpa-dev@incubator.apache.org
Subject: Re: possible to write a JPA Query to that filters both an
Entity and its relationship entities?


On Feb 23, 2007, at 7:49 PM, Pinaki Poddar wrote:

One way to realize a owner object with a partially filled
multi-cardinality relationship is to expand fetchplan concept. For
example, if we consider a Department of 100 Employees of which only 20

are oldtimers, then a query can select the particular Department from
the datastore but to realize that Department as a Java object in
memory with only 20 senior Employees in _employees collection will
need to tune the fetch plan acconrdingly.
In JDO, fetch plan is specified to certain details to extract a subset of native states and subset of relationships. JPA does not yet specify
fetch plan, but OpenJPA does via @FetchGroup extension. However
currently, no filtering criterion is applied to filter the content
while a particular relationship path is traversed. It is a
all-or-nothing affair. Either Department is fetched with all its 100
Employees or the Employees relation is not traversed at all. The only
control available now is to specify which native states and
relationship will be considered for fetching and in case of recursive
relationship how many times a particular relation path will be
traversed.
Essentially query and fecth paths are working in conjunction -- query
selects the root object and fetch plan decides which subset of the
closure of this root object is realized in memory.  It will surely be
a useful feature to consider expanding fetch plan with filtering
criteria.

Sorry to disagree. If you fetch Department instances, I think you always want the _employees collection to contain all the related Employees. You
never want a filtered collection unless you provide a filter via some
Java behavior.

In the query case, you want to get references to the filtered instances
but there's no connection to the _employees field of the department.

This is a good thing.

Craig



Pinaki Poddar
BEA Systems
415.402.7317


-----Original Message-----
From: Patrick Linskey
Sent: Friday, February 23, 2007 4:37 PM
To: open-jpa-dev@incubator.apache.org
Subject: RE: possible to write a JPA Query to that filters both an
Entity and its relationship entities?

I would expect that your query would return the unfiltered employees
collection. But I have not run any tests.

In other words, I would expect that the oldtimers collection would be
the full set of employees for each dept that had any oldtimers >= 15.

That said, I don't think that your interpretation is necessarily
wrong.
Just not my expectation.

-Patrick

--
Patrick Linskey
BEA Systems, Inc.

_____________________________________________________________________ _
_
Notice:  This email message, together with any attachments, may
contain
information  of  BEA Systems,  Inc.,  its subsidiaries  and
affiliated
entities,  that may be confidential,  proprietary,  copyrighted
and/or
legally privileged, and is intended solely for the use of the
individual or entity named in this message. If you are not the
intended recipient, and have received this message in error, please
immediately return this by email and then delete it.

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Friday, February 23, 2007 4:14 PM
To: open-jpa-dev@incubator.apache.org
Subject: Re: possible to write a JPA Query to that filters both an
Entity and its relationship entities?

Hi Patrick,

 From the JPQL
select dept, oldtimers from Department dept LEFT JOIN
dept.employeeCollection oldtimers WHERE dept.deptno >= 100 AND
oldtimers.yearsOfService >= 15

I expect to get SQL that looks something like select dept.id,
dept.name, oldtimers.id, oldtimers.firstname, oldtimers.lastname,
oldtimers.salary, oldtimers.ssn from Department dept, Employee
oldtimers WHERE dept.id = oldtimers.deptid AND dept.deptno >= 100 AND

oldtimers.yearsOfService >= 15

I then construct instances from the returned dept and oldtimers
columns and give back to the user the results, with one result row
for

each SQL ResultSet row. So the filtering from the JPQL translated
into

the WHERE clause in the SQL should exclude oldtimers that don't
qualify.

If you don't use the oldtimers.yearsOfService >= 15 in the SQL, then
you will get all of the employees in the department returned in the
result. But why do you not use the oldtimers.yearsOfService >= 15 in
the SQL?

Craig

On Feb 23, 2007, at 2:12 PM, Patrick Linskey wrote:

In fact, you have to do a bit of work to get SQL to return you non-

filtered instances. So I don't get it. Does OpenJPA not construct
the obvious SQL that filters oldtimers?

I don't think that you and I are really in sync here. Maybe
some more
concrete examples are in order?

-Patrick

--
Patrick Linskey
BEA Systems, Inc.


______________________________________________________________
________
_
Notice:  This email message, together with any attachments, may
contain information  of  BEA Systems,  Inc.,  its subsidiaries  and
affiliated entities,  that may be confidential,  proprietary,
copyrighted and/or legally privileged, and is intended solely for
the use of the individual or entity named in this message. If you
are not the intended recipient, and have received this message in
error, please immediately return this by email and then delete it.

-----Original Message-----
From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED]
Sent: Friday, February 23, 2007 2:09 PM
To: open-jpa-dev@incubator.apache.org
Subject: Re: possible to write a JPA Query to that filters both an
Entity and its relationship entities?

So,

I can tell you that having a result column that is filtered in the
query is not what a user would expect. In JDO, the oldtimers column

would be filtered.

In fact, you have to do a bit of work to get SQL to return you non-

filtered instances. So I don't get it. Does OpenJPA not construct
the obvious SQL that filters oldtimers?

I've re-read the JPA specification, and it appears to be silent on
the issue of filtering. Is this a portability issue?

Craig

On Feb 23, 2007, at 1:57 PM, Tom Mutdosch wrote:

Hi Patrick,
Thanks for the query suggestion.  I guess what I was initially
thinking of doing was incorrect in that JPA doesn't really give me

a "view" of what I want.  That is, I can never get a Department
object containing a list of filtered Employees.  A JPA object
returned from a query is always going to be an exact
representation
of the database.  So your Department object is always going to
contain all of the Employees in its relationship.

So like you mentioned, I can still get all the information using
one query, and then just process those results as I want them.  I
imagine that this would entail some sort of wrapper bean
that would
house the Department and the filtered list of Employees.  Or what
if I added a regular method to my Department entity called
getFilterEmployees() which would return a List that I populated
with the filtered results from my query?  Does that seem like a
reasonable thing to do -- if I didn't want to deal with a wrapper
object but still have all of my desired data captured by a single
Entity?
Thanks
Tom


Patrick Linskey wrote:
It is, but it doesn't buy you much in this situation --
the oldTimers
collection in your example won't be filtered to just the
ones that
are
old. It'll be all the employees in the dept.

-Patrick



Craig Russell
Architect, Sun Java Enterprise System
http://java.sun.com/products/
jdo
408 276-5638 mailto:[EMAIL PROTECTED] P.S. A good JDO? O, Gasp!



Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/
jdo
408 276-5638 mailto:[EMAIL PROTECTED] P.S. A good JDO? O, Gasp!



Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!

______________________________________________________________________ _ Notice: This email message, together with any attachments, may contain information of BEA Systems, Inc., its subsidiaries and affiliated entities, that may be confidential, proprietary, copyrighted and/or legally privileged, and is intended solely for the use of the individual or entity named in this message. If you are not the intended recipient, and have received this message in error, please immediately return this
by email and then delete it.

Craig Russell
Architect, Sun Java Enterprise System http://java.sun.com/products/jdo
408 276-5638 mailto:[EMAIL PROTECTED]
P.S. A good JDO? O, Gasp!

Attachment: smime.p7s
Description: S/MIME cryptographic signature

Reply via email to