Open johnynek opened 1 year ago
I think moving to use more or have the parser be a primary source or some hybrid will be an improvement due to its speed and ability to respond to refactoring. But things like protobufs, or external jars, or any genrule that happens to be about i guess bazelfe's indexer will respond to since it works on the jars and parses the target information out that way. Maybe some layered approach of trying to use the parsed info and falling back only then to what the sniffing of target building does?
this code creates a jar index here:
https://github.com/bazeltools/bazelfe/tree/main/bazelfe-core/src/jvm_indexer
but also this code creates indexes to build the dependency graph:
https://github.com/bazeltools/bzl-gen-build
bazelfe indexes predate the gen build tool. It seems like maybe we should remove indexing from this code and instead rely on bzl-gen-build indexes.