Kevin Conaway <kevin.conaway <at> gmail.com> writes: > > I don't know of a reason why it doesn't do buffered reads. I do know that the Reader interface is designed for working with characters whereas the InputStream interface is designed for working with bytes.Can you recompile with an implementation that does buffered reads and see if that solves your issues?KevinOn Tue, Oct 7, 2008 at 6:36 PM, Aaron <dirtyvagabond <at> yahoo.com> wrote:Hi Restlet Community, > We've been doing some performance profiling of our system, which uses > Restlet 1.1-M4 to handle resquests. We noticed that > org.restlet.util.ByteUtils.toString(Reader) is costing us about 40% of > our runtime. > We took a peek at the source code and realized this method is reading in > the payload character-by-character. So this method also stands out in our > profile in terms of invocation count on java.lang.StringBuilder.append()... > we see literally millions of calls to append() since it's making a call per > character of payload. > So we're wondering if it might be better for this to create a very large > buffer and read everything into that at once? Also, is there a reason that > it's dealing with characters rather than bytes? > Many thanks for any insights into this...! > Best wishes,Aaron
Hi Kevin, We've tried a basic char buffer approach which seems to speed it up dramatically. So we'll probably run with that. I was still curious to hear any community feedback, and perhaps see if this might be something the core team would find it worthwhile to investigate.
