This patch demonstrates the idea and some of the dispatch semantics that need to be accounted for. It stringifies individual functions, which is much too fine-grained for anything that could be deployed.
It seems to me that optimally stringifying a GWT module and performing runAsync splitting is the same problem. It's always bugged me that the current runAsync implementation does a lot of work on the Java side of things to predict the structure of the emitted JS code. Solving this problem well should investigate how the overall impact on the compiler of runAsync could be reduced if it didn't have to juggle both the Java and JS AST structures. -- Bob Vawter Google Web Toolkit Team --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/Google-Web-Toolkit-Contributors -~----------~----~----~----~------~----~------~--~---
lazyEval.patch
Description: Binary data