On Wed, Aug 21, 2013 at 10:47:20PM -0400, Corey Richardson wrote: > - Metadata is scary. It's a hairy part of the codebase that I sure > don't understand. I know its purpose, more or less, but not the > specifics of how or why things are encoded. Michael Sullivan could > speak more to this, he is the last one to have touched it. The > compiler can't help you when you make a mistake, either.
FWIW, I had a branch that replaced the type encoding part of metadata with the standard encoder. It did not use an EBML output but rather a very specialized, minimal encoder. This did not improve compile times, though. I think the reason was that I never got around to implementing the "type abbreviations" code that optimizes the case of repeated types. Anyway, I still think the best *long-term* fix for making metadata more approachable and maintainable is to transition to using the standard encoding stuff. Clearly, we'd need to structure the encoder so that we first encode an index and then encode the various bits individually, presumably using gzip. It is unclear to me though whether this would increase perf -- perhaps as a side-effect of general rearchitecting. It might also help more in the general metadata code, where EBML costs something, than in the type encoding code, which is already pretty specialized. Niko _______________________________________________ Rust-dev mailing list Rust-dev@mozilla.org https://mail.mozilla.org/listinfo/rust-dev
