The data pipelining rewriter has a couple of new improvements: - support for multiple batches, e.g. one data element can introduce variables used for the EL in a second element. This is arbitrarily limited to hardcoded 3 batches, but could trivially be made configurable. - elements that cannot be successfully processed aren't deleted
I'd like to get this checked in the next day or so, so please review if you can. http://codereview.appspot.com/14063
