-
cargo stuff
- spell checker rust documentation and other md files.
- md-book
- release management
- configuration via features
- platform binary generation
- publishing onto crates.io
- package dependencies
- declared but unused dependencies
- outdated dependencies, upgrades
- licensing analysis
- source code analysis
- modulewise, imports.
- list of types (type, struct, enum, const, static)
- list of functions, traits, trait implementation
- type methods, public and private.
- featured gated source code items.
-
README: create document links in a separate section, linking to docs/ directory.
-
implement logging using log facade.
-
benches: Fix benches/ and add it as part of perf.sh
-
rdms-perf: add performance suite for dgm, robt, wral, shllrb, shrobt, croaring
-
rdms-perf: add latency measurements.
-
rdms-perf: plot graphs.
-
rdms-test: migrate test-suites from ixtest
-
rdms-perf for robt
- try initial build with 1M, 10, 100M entries; with value as Binary(1K)
- try with nobitmap, xor8 bitmaps.
- try incremental build with 1M, 10, 100M entries; with value as Binary(1K)
- try the incremental builds with and without compaction
- measure concurrent read performance for 1, 2, 4, 8, 16 threads
- try with and without lru cache.
- try initial build with 1M, 10, 100M entries; with value as Binary(1K)
-
wral: journal-limit, adjust the algorithm to not to exceed the journal limit.
(a) review 5c71164f6d9e57ce60ed0030f1fa7dba7d5056b5 fix errors before refactoring llrb out into ppom
rdms-clru:
Concurrent access to least-recently-used-cache need its backing datastructure like a disk-btree to be immutable. Otherwise, we may have to deal with synchronization problem in building the cache and evicting the entries.
Access-1: Get(cache) fail Get(disk-btree) Set(cache)
Access-2: Set(disk-btree) Set(cache)
Access-3: Remove(disk-btree) Remove(cache)
Access-4: Evict(cache)
There will synchronization issues when above listed access scenarios happen concurrently on the disk-btree and the cache.