Subj : Re: [Rhino] what is better for includes?
To   : Mike C.
From : Igor Bukanov
Date : Sun May 16 2004 11:27 am

Mike C. wrote:
> For using JS in a servlet environment, I translate the page into a
> JavaScript only file, then I compile it with Rhino into a bytecode .class
> file.
> For include files, I can either
> - compile each include into a separate class file, then execute the script
> in the same scope as the main page
> - make a big .js file for each main page, with all include files
> concatenated, then compile the main page only into bytecode.
> Question is, do I have any chance to get better compile-time optimizations
> if I concatenate all include files to the main script? Otherwise I would use
> the first solution, since the translation is simpler for it.

With one big file you may hit a limitation of .class file format where 
the method size is restricted to 64K of bytecode. Since in Rhino all 
code to initialize functions declared in the script is generated as one 
JVM method, the 64K limit is hit when there is arround 2000 top level 
functions. You may also hit this limit if your script contains a lot of 
object and array literals since the generated code to initialize them is 
pretty fat. CVS tip of Rhino improves that somewhat but still the limit 
can be reached quickly.

But if 64K is not a problem the .class file generated for the monolithic 
script is faster to load since JVM does not need to lookup duplicated 
strings and various constants are shared. In addition if your scripts 
contain cross calls then with the optimization level set to 1 or higher 
Rhino will generate faster code for these calls.

Regards, Igor

.