+rockot, FYI grok is part of code search

Now that we're using mojom for all new IPC & converting existing IPC, having this would be super useful in the near term.

Emma or Dave, is this something you could take on soon?

Some more details on what's involved (details from the grok team):

1) Find a way to capture "compilation records" for the files, including the compiler settings and dependencies. 

2) Write an indexer that can use a captured compilation record to "replay" the compiler front-end, and emit graph data according to the Kythe schema.

How much work these are depends on how reproducible the build is, and the availability of existing analysis tools for the language in question. For a typical compiled, statically typed programming language, this is morally no more than hooking up the front end of the compiler to generate a typed, resolved AST, and then walking around in that AST to emit some nodes and edges. For languages without types, or which do not have existing libraries for analysis, it may take more work. As long as the indexer conforms to the schema, existing tools and APIs should "just work", modulo bugs. 

For code-generation schema languages, like protobuf or mojom in this case, there is some additional work to instrument the code generator so that the target language indexer can find the connections back to the schema.