People who know Xtext and ECMAScript a little bit have no problem describing such a project with a single word: it's crazy. And in deed, it is challenging for a number of reasons. One of them is the ECMAScript grammar, which requires certain parser features which cannot be implemented with a pure Xtext grammar. Another is performance. While all the nice default Xtext solutions work great for smaller to midsize modeling projects, it is a different story to process, i.e. validate and compile, not only hundreds, but thousands of files. And when you think you're done, more surprises are awaiting you, such as Maven build system difficulties or rough edges coming with new technologies as Xcore.
In this talk I will sketch out our solutions and thereby demonstrate how to customize Xtext to achieve things which will not work out of the box. I will also compare our Xtext based solution, and Xtext as a framework, with the DLTK, as a predecessor solution of our IDE was based on the DLTK and serves now as an indicator.