margaret
margaret copied to clipboard
ssbc
a flume-like persisted append-only log implementation
Why does this need to be here, and what would it take to eliminate it? https://github.com/ssbc/margaret/blob/22ed48add4909da6153dfc3c1ed82642efd9c225/indexes/badger/index.go#L32 It pops up in every (properly built) testing run of `go-ssb`.
So, overall, the Badger indexes seem to be the best guarded against race conditions of any I've seen. However, there's one spot where I can see a possibility for data...
Related to #30 These both make modifications to the database, each in its own separate transaction, but they can be called independently and the sequence number is often used by...
Probably ought to wrap this in a mutex. Index can be set, potentially multiple times if done concurrently, before the observable finished updating: https://github.com/ssbc/margaret/blob/22ed48add4909da6153dfc3c1ed82642efd9c225/indexes/mkv/index.go#L114-L128
(Trying to keep track of some of the things I'm finding so I can remember to look at them later.) This section should probably be wrapped in a mutex: https://github.com/ssbc/margaret/blob/22ed48add4909da6153dfc3c1ed82642efd9c225/indexes/sinkindex.go#L42-L49
I have yet to see a Margaret query in go-ssb that doesn't use SeqWrap. What is the purpose of it, and why isn't it the default (or only option) if...
Related to #19 For someone not already familiar with this project, I'm finding that the only way to actually figure out how anything does is to read through the code....
Currently, the size of an entry is abused to indicate that it was deleted (the size is negated) It would be better to have a user metadata thing for this...
right opening a query on offset2 with reverse and greater or less then constructions doesn't work correctly. Re-enable the tests and add the needed handling to qry.Next.