Quinlan Pfiffer
Quinlan Pfiffer
Idle thoughts on ways to improve the transaction implementation: - ~~`dirty` state: if a transaction didn't actually do anything (quite common for `ol_unjar` operations) then we should just discard the...
Tests need to be split up. We need tests for the frontend. This ticket is for: - Where do we put tests? - The testing binary needs to have everything...
Platforms we want to be able to target with at least `liboleg`: - SmartOS/Illumos - OS X - Linux - Win32 (Cygwin or otherwise) Which is going to facilitate some...
This is kind of a vague, general clean up ticket. There are no clear metrics to go after, really, but there are some definite things we could fix: - Repeated...
Could possibly use jump_hash to avoid reorganizing the entire hashing space on a rehash.
Values stored in the values file should be padded with extra space (at least 4k) to reduce page faults. This will also allow some headroom on updates.
I keep breaking the AOL file's format by changing shit, mosty recently the removal of `Content-Type` as a first-class citizen. If we read how we're supposed to read the data...
There might be some weirdness happening during replay in regards to bucket allocation. `db->val_size` might be expanded unnecessarily when: 1. A bucket is reallocated during AOL replay, and 2. That...
The more of these we can store in memory the better. Some ideas: - ~~Remove content type stuff~~ - Make a secondary struct for data that isn't necessary everytime (next_bucket,...
General ticket for transactions. API, implementation, ideas, thoughts, etc. Transactions should have the following goals: - Make all operations inside a transaction atomic - Lock the database to prevent unwanted...