Closed tbg closed 8 years ago
I will back out the second commit (not sure what I was doing) but will leave it in for now since I commented on it, which GitHub sometimes hides after removing. original second commit is gone.
after a chat with @danburkert I've backed out the change which returned owned entries from entry
(the standard user will use his WAL log implementation, and that can use the existing interface just fine thanks to mmap). What's left of the PR is misc refactoring, in particular providing entries(lo,hi)
via the PersistentLog
trait and simplifying some code.
ok, I think I've addressed the TODOs. The changes here are mostly cosmetic, but (in my opinion) net positive. It would be more interesting to get rid of the ad-hoc vector construction (now in log.entries()
, previously in I think two places) since really all we need is a borrow to construct capnp messages from - doesn't seem hard but not worth it until the WAL log is done.
Looks like aster
broke... I'll refire.
Yay, it works! @danburkert does this look good to you? You had some concerns earlier and I'm not sure they were addressed, but it looks like it.
Looks good to me. Thanks @tschottdorf. @homu r+
:pushpin: Commit b1dd39e has been approved by danburkert
:hourglass: Testing commit b1dd39e with merge 456e1e6...
for #101. @danburkert suggested that while the log interface is best driven by the first non-trivial implementations in the works, it still makes sense now to have entry() with an owned argument.