Tyler Neely
Tyler Neely
 https://arxiv.org/pdf/1509.05053.pdf We spend a lot of time doing binary search in index nodes to find an interesting leaf. This is partially because we don't do interpolation search, but mostly...
we use `BTreeSet` in a lot of places, resulting in a huge memory overhead compared to better compact structures like the ART. generally, these structures don't need to support concurrent...
A nice feature for more structured queries, especially once transactions are in place on the Tree, is to be able to define a relatively thin Record or Table-like overlay where...
dictionaries allow us to get far higher compression ratios when compressing small data items. they can not be incrementally generated, as far as I know. the same dictionary that was...
gradually apply https://github.com/dtolnay/no-panic to a new test feature that just compiles optimized code and asserts that panic does not happen in a wider and wider subtree. biggest offenders for panics...
we perform the same calculations on various types all over the place in the sled codebase, in ad-hoc error-prone ways that cost time when debugging, refactoring, or implementing any change...
since all IO and cross-thread communication points are manually instrumented with failpoints and debug_delay, we can use these calls to record traces of IO and linearization points. * run a...
We need to support parallel writes for building on windows. I believe the way to go about this is through https://retep998.github.io/doc/kernel32/fn.WriteFile.html and passing in an [OVERLAPPED](https://retep998.github.io/doc/winapi/minwinbase/struct.OVERLAPPED.html) struct, as per [SO](https://stackoverflow.com/questions/3852758/equivalent-to-pread-pwrite-in-msvc)....
A key piece of complexity in Caffeine is the need to have a concurrent hashmap for maintaining the logical linked lists. sled currently uses an allocation-heavy doubly-linked-list. sled already has...
I want to build distributed databases on top of sled over time. Distributed databases often shard their data lexicographically. sled can make this easier by supporting keyspace fusing / splitting....