> On Mon, 25 Mar 2002, Kin-Man Chung wrote: > > > I have noticed that the current Jasper does a lot of string manipulations > > very slowly, e.g. there are unnecessary copying or concatenations, and > > the uses of String where StringBuffer should be used. Hopefully I can > > avoid all of these. > > > > But I don't have high hopes that Jasper performance can be improved to > > the point that an user will notice, because the time spent in Jasper > > is probably small compared with that spent in javac. Now if we could > > have jasper generate byte code directly... > > I wouldn't worry too much about this kind of optimizations - the page > compilation happens only once and it's quite easy to pre-compile the pages > with jspc. > > The place where optimizations do matter the most is the runtime - I would > be happy with a generator/compiler that is slower but generates faster > code. Even if it is possible to use BCEL to generate bytecodes directly, > I think this would be a huge waste of time if done before the runtime > is optimized. > > A clean generator that makes it easier to improve the generated code > is better than the fastest generator. > > Costin
I happen to agree with you. My initial effort in performance improvement would certainly be on the generated codes. I wasn't really serious when I suggested that Jasper generates byte codes. -- To unsubscribe, e-mail: <mailto:[EMAIL PROTECTED]> For additional commands, e-mail: <mailto:[EMAIL PROTECTED]>