malaperle added a comment. In https://reviews.llvm.org/D35406#809609, @ilya-biryukov wrote:
> The idea is to allows us changing the way we manage/rebuild PCHs and ASTs. > ASTUnit is used for many things and has a fairly complicated and verbose > interface and does a lot of mutations all other the place inside it, so > making changes to it is not particularly easy. > On the other hand, it doesn't add a lot of value for clangd specifically, > since we use a very simple subset of it. > > The next step I planned was to allow code completion to be run in parallel > with recomputing preamble/computing diagnostics. (And in general, allow > multiple TUs to be processed in parallel). > That would probably involve changes to the interface, so we should > definitely sync with you on the work you've been doing. > And we were also planning to look into indexing at a later point, so > discussing the design there might be interesting as well. Thanks a lot for the additional information. For synching with what I am doing. I am "only" looking right now at the modeling of the index and its on-disk storage, not really on the speeding up of the parsing of the TUs (i.e. the input of the index). I use index::IndexDataConsumer to feed the index and I see that in your change it is still pretty much used the same way as before so there is no problem. I will post on the indexing thread from before with a more detailed proposal from before once it is a bit further along. https://reviews.llvm.org/D35406 _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
