LGTM. Makes sense. Good news on the savings!!
kathrin On Wed, Nov 12, 2008 at 6:24 PM, Lex Spoon <[EMAIL PROTECTED]> wrote: > Hi, Kathrin, > > Can you review this patch for me? > > This implements an idea the group has been talking about a long time: > when generating serializers and deserializers for RPC, don't support > sending types *to* the browser that are only sent *from* the browser, > and vice versa. > > The main changes are: > > 1. Instead of a single set of types, several classes now have two sets > named "typesSentToBrowser" and "typesSentFromBrowser". > > 2. ProxyCreator.writeTypeMethods leaves out entries for methods that > are not used. > > 3. The format of the RPC serialization policy file needed updating. > Previously it only had to indicate which types could be sent across > the wire. Now it needs to include info about which direction the > types can be sent in. > > > There are also some rippling changes: > > - Many of the instance methods of SerializableTypeOracle are > made static methods in SerializableUtils. It got weird > to have to choose which of to STO's to query, even though > they would both have the same answer. > > - JsParser can now handle "undefined". > > > > -Lex > --~--~---------~--~----~------------~-------~--~----~ http://groups.google.com/group/Google-Web-Toolkit-Contributors -~----------~----~----~----~------~----~------~--~---
