Complaint number one by developers using FLOW3 – and that includes the core team – has been the compilation speed in a development context. There are times FLOW3 needs to flush all code caches which can result in several seconds of waiting time on the next hit, when the proxy classes are rebuilt. In order to tackle this, we scheduled a whole week for a code sprint in Lübeck.
The proxy building routines are one of the oldest parts of FLOW3 and although we revised the actual compiler routines some time ago, the logic for matching classes and methods for the Aspect-Oriented Programming support has remained unchanged. Andreas Förthner, Christopher Hlubek and myself discussed ways to optimize the matching and during the past week Andi refined the concept and turned it into code.
We always hoped that our approach would speed up compilation time, but we didn't expect the big performance boost we eventually reached. Now, in turns of numbers that is … well, we keep the actual numbers as a little surprise for the first FLOW3 conference, but you'll certainly like it!
Christian Müller also joined us here in Lübeck and did a great job going through all pending issues, checking if they are still valid, asking questions to the reporters and solved various tickets right away. Christian really is our good soul when it comes to community support, there's hardly a day or night time you don't spot him in our IRC channel.
We seem to have most of the code done for a FLOW3 1.1 release, except the new HTTP support which is still in the making. Nevertheless, you may expect a 1.1 release during the next weeks, including a few neat surprises. We could also fix the few failing tests which gave Smilla reason to cry so desperately (see picture).
What are you looking forward to for future FLOW3 releases, what would delight you most?
Leave a reply