openjpa-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Patrick Linskey" <plins...@bea.com>
Subject RE: Forced getter/setter access
Date Wed, 04 Apr 2007 06:11:29 GMT
Ah... I was referring to the fact that if an implementation isn't
tracking field accesses, then it must sit around and do in-mem
comparisons (and hold onto hard refs to copies of all read data) in
order to figure out what to write back to the DB. So, I was referring
solely to the in-mem computation time, not to SQL generation.

-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: Evan Ireland [mailto:eireland@sybase.com] 
> Sent: Tuesday, April 03, 2007 11:04 PM
> To: open-jpa-dev@incubator.apache.org
> Subject: RE: Forced getter/setter access
> 
> Patrick,
> 
> I meant "tuned" as in not including columns in the SQL update "set"
> clause if they weren't really changed by the app (even if a setter
> was called, the old and new values may be the same).
> 
> The update "set" clause is 100% portable SQL.
> 
> So I'm not entirely clear what you mean by:
> 
>     "tuned" for your implementation
> 
> > -----Original Message-----
> > From: Patrick Linskey [mailto:plinskey@bea.com] 
> > Sent: Wednesday, 4 April 2007 5:44 p.m.
> > To: open-jpa-dev@incubator.apache.org
> > Subject: RE: Forced getter/setter access
> > 
> > > If you are going to issue "tuned" updates to the DB, 
> > determining what 
> > > "really" changed (as opposed to what setter methods were
> > > called) can avoid unnecessary DB overheads.
> > 
> > ... "tuned" for your implementation, that is. Often, business 
> > needs require that a transaction involves a large number of 
> > objects, and it would be unfortunate to have to compromise on 
> > transaction integrity just for the sake of a particular 
> > implementation. It's not just a matter of needing to 
> > periodically call flush() during a tx, but rather of having 
> > to design transactions that don't read too many objects.
> > 
> > Happily, with OpenJPA, this is a non-issue, regardless of 
> > property or field access.
> > 
> > -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: Evan Ireland [mailto:eireland@sybase.com]
> > > Sent: Tuesday, April 03, 2007 10:39 PM
> > > To: open-jpa-dev@incubator.apache.org
> > > Subject: RE: Forced getter/setter access
> > > 
> > > Granted, but with a reasonable implementation the cost 
> > should be low 
> > > for:
> > > 
> > > "at commit / flush time compare the current values with the 
> > original 
> > > values to figure out what to write back to the database."
> > > 
> > > If you are going to issue "tuned" updates to the DB, 
> > determining what 
> > > "really" changed (as opposed to what setter methods were
> > > called) can avoid unnecessary DB overheads.
> > > 
> > > Anyway at the end of the day, you must use the 
> > getters/setters because 
> > > the spec says so, and you should write portable apps 
> where possible 
> > > :-)
> > > 
> > > > -----Original Message-----
> > > > From: Patrick Linskey [mailto:plinskey@bea.com]
> > > > Sent: Wednesday, 4 April 2007 5:21 p.m.
> > > > To: open-jpa-dev@incubator.apache.org
> > > > Subject: RE: Forced getter/setter access
> > > > 
> > > > The main reason to support getter / setter access is for 
> > > > implementations that cannot intercept field accesses. So, the 
> > > > getters and setters are there so that the JPA 
> implementation can 
> > > > create a subclass of your entity type (hence the 
> no-final-classes 
> > > > rule) and track what happens as you invoke the setters 
> > and getters. 
> > > > In other words, your business methods become part of the JPA 
> > > > implementation's domain.
> > > > 
> > > > So, when using property access, your contract with the 
> > JPA provider 
> > > > is that you'll access persistent attributes only through 
> > the setters 
> > > > and getters, which allows the implementation to track 
> what you do 
> > > > and when you do it. If you could directly access the underlying 
> > > > state, the implementation would have no way to know 
> what happened 
> > > > during the course of a transaction. This, in turn, would 
> > mean that 
> > > > the implementation would have to keep a copy of every 
> bit of data 
> > > > that you read during a transaction, and then at commit / 
> > flush time 
> > > > compare the current values with the original values to 
> figure out 
> > > > what to write back to the database.
> > > > 
> > > > As it turns out, when you use OpenJPA, all your direct field 
> > > > accesses are replaced with synthetic static methods 
> > anyways, so from 
> > > > a performance standpoint, you'll see equivalent behavior 
> > either way. 
> > > > In my experience, persistent domain model field access 
> > performance 
> > > > in tight loops is rarely actually a performance 
> bottleneck; it's 
> > > > almost always going back and forth to the database that ends up 
> > > > being the bottleneck, and thus the most important place 
> > to optimize.
> > > > 
> > > > -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: Phill Moran [mailto:pjmoran@rogers.com]
> > > > > Sent: Tuesday, April 03, 2007 10:02 PM
> > > > > To: open-jpa-dev@incubator.apache.org
> > > > > Subject: Forced getter/setter access
> > > > > 
> > > > > Can anyone explain why this rule is in effect:
> > > > > 
> > > > > When using property access, only the getter and setter
> > > method for a
> > > > > property should ever access the underlying persistent field
> > > > directly.
> > > > > Other methods,
> > > > > including internal business methods in the persistent
> > > > class, should go
> > > > > through the getter and setter methods when manipulating
> > > persistent
> > > > > state. (section 2.1.4 OpenJPA manual)
> > > > > 
> > > > > This seems rather execution costly. If ,for instance, I
> > > have a Size
> > > > > class with hieght, width and length then to calculate 
> > and return 
> > > > > volume I suffer a three method call overhead:
> > > > > return getWidth() * getLength() * getHieght();
> > > > > 
> > > > > This is opposed to a more efficient
> > > > > 
> > > > > Return height * width * length
> > > > > 
> > > > > Phill
> > > > > 
> > > > > 
> > > > 
> > > > 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.
> > > > 
> > > 
> > > 
> > 
> > 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.
> > 
> 
> 

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.

Mime
View raw message