I also think that avoiding the use of final at all is not something
practical. In some cases marking a method as final improves the performance
by a sigficiant amount. Many of the Java books and reference materials
actually encourage the usage of final for performance reasons. Of course I
would agree that we should use it carefully, in order not to limit the
extensibility. There are already many final methods in Xalan source code. I
do not think there is a reason not to introduce any new final method at
all.
Morris Kwan
XSLT Development
IBM Toronto Lab
Tel: (905)413-3729
Email: [EMAIL PROTECTED]
"Santiago
Pericas-Geertsen" To:
<[email protected]>
<Santiago.PericasGeerts cc: <[EMAIL
PROTECTED]>
[EMAIL PROTECTED]> Subject: Re: cvs
commit: xml-xalan/java/src/org/apache/xalan/xsltc/dom
DOMAdapter.java
SAXImpl.java
03/07/2003 03:58 PM
Please respond to
xalan-dev
----- Original Message -----
From: "Joseph Kesselman" <[EMAIL PROTECTED]>
> Note that final methods also cause problems if/when you need to introduce
> subclassing/polymorphism, and that SUN's own performance team generally
> recommends strongly *AGAINST* trying to use the final keyword as an
> optimization trick.
Sun's Hotspots has a mechanism by which it essentially infers the final
keyword (often, e.g. if a certain class loaded, it can also "uninfer" it).
The main advantage of a final method is that it can be inlined, thus saving
the overhead of method invocation and late binding. I can't think of any
reason for which not to use final: for variables it promotes constant
propagation (typically at compile time and when the init expression has not
side effects), for methods it promotes inlining.
This is just my opinion though, I can always contact the performance team
if we want to make sure. Do you have a reference Joe?
-- Santiago
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]
