Tpt
|
e59c4612b2
|
RocksDB: Removes merge and compact operators
|
3 years ago |
|
Tpt
|
b7ee3a6767
|
Removes string usage counting
|
3 years ago |
|
Tpt
|
2c304aa29d
|
Share environment between RocksDB instances
Allow multiple RocksDB instance in the same process to share resources instead of fighting to get them.
|
3 years ago |
|
Tpt
|
6f44a5956b
|
Simplifies bulk load API
|
3 years ago |
|
Tpt
|
7b1c4e0ad5
|
Returns a clean error on transaction read after commit
|
3 years ago |
|
Tpt
|
607aa0b0dd
|
WIP: Transactions
|
3 years ago |
|
Tpt
|
9f414c13fd
|
Simplifies RocksDB SST API
|
3 years ago |
|
Tpt
|
367a1b4585
|
Reuse RocksDB transactional DB
First step for snapshots and transactions support
|
3 years ago |
|
Tpt
|
196d6d6576
|
Adds bulk load by directly write RocksDB SSTs
|
3 years ago |
|
Tpt
|
1d3108d27f
|
RocksDB: Uses the most recent storage format
|
3 years ago |
|
Tpt
|
569000b5ea
|
Storage: adds write batch
Atomic insertions and huge performance improvements for RocksDB
|
3 years ago |
|
Tpt
|
986d3e60bb
|
RocksDB: bulk load
TODO: we do a lot of lookups during load for GC, so the option is not very useful
|
3 years ago |
|
Tpt
|
98f9a307b8
|
Exposes a function to compact the database
Use it after load
|
3 years ago |
|
Tpt
|
8c0b4f5322
|
RocksDB: do not use transactiondb for now.
Not used yet
|
3 years ago |
|
Tpt
|
00ad6e634f
|
Cleaner storage backend architecture
|
3 years ago |
|
Tpt
|
73cd761229
|
Removes unused TermEncoder
|
3 years ago |
|
Tpt
|
0ee25d7579
|
RocksDB: Adds prefix bloom filters
|
3 years ago |
|
Tpt
|
821cac9c4c
|
Sets some RocksDB options doing perf improvements
|
3 years ago |
|
Tpt
|
25c808c9eb
|
Reduces successive merge
Fixes awful reads
|
3 years ago |
|
Tpt
|
20376ba0dd
|
Makes Clippy happy
|
3 years ago |
|
Tpt
|
5da9fd4f14
|
Fixes Storage::contains_str
Takes care of the counter if not a key might be removed after the next compaction even if assumed to be contained
|
3 years ago |
|
Tpt
|
a7a0b7bbf3
|
String garbage collector: do not decrease counters set to MAX.
We do not keep track of counting beyond i32::MAX
|
3 years ago |
|
Tpt
|
0c9a46fcd1
|
Adds runtime checks to insert_term_values
Might help to spot logic bugs in the future
|
3 years ago |
|
Tpt
|
4c71327c95
|
RocksDB: Simplifies SlicesIterator
|
3 years ago |
|
Tpt
|
cddb5900ef
|
Storage: Manages strings GC using merge operation
Removes old strings during compaction
|
3 years ago |
|
Tpt
|
045f40ccad
|
Configures RocksDB logs to avoid logs overloading
|
3 years ago |
|
Tpt
|
bf0f178c41
|
Adds a ColumnFamilyDefinition struct to store column family options
|
3 years ago |
|
Tpt
|
def7a3ce72
|
RocksDB: Adds low priority writes
Allows RocksDB to defer garbage collection if useful
|
3 years ago |
|
Tpt
|
f68d747308
|
RocksDB: Uses lz4 compression on disk
|
3 years ago |
|
Tpt
|
3119e7748e
|
Small RocksDB optimizations
Use native iterator bound
|
3 years ago |
|
Tpt
|
dfefe6cd1c
|
Uses transactional RocksDB
First stop to real transaction support
|
3 years ago |
|
Tpt
|
3fd0332e32
|
Use our own Custom rocksdb bindings
Will allow exposing more C++ functions
|
3 years ago |
|
Tpt
|
e297de73d1
|
Refactors backend code
Makes it easier to implement transactions
|
3 years ago |
|
Tpt
|
dfe50cff11
|
Fixes WASM builds
|
3 years ago |
|
Tpt
|
0bd512a14c
|
Avoids using API without column families
|
3 years ago |
|
Tpt
|
cb146efd7c
|
Use directly RocksDB C API
Will allow more customizations in the future
|
3 years ago |
|
Tpt
|
063683088d
|
Moves back from Sled to RocksDB
Simple implementation without transactions
|
3 years ago |
|
Tpt
|
2a82fe4f75
|
Adds on-disk benchmarks
|
3 years ago |
|
Tpt
|
728a5dadf4
|
Adds a native SPARQL query and update benchmark based on BSBM
|
3 years ago |
|
Tpt
|
e1c8da49a2
|
Sorts blank nodes with ORDER BY
|
3 years ago |
|
Tpt
|
45f612b6e3
|
Allows comparing rdf:langString literals with the same language tag
|
3 years ago |
|
Tpt
|
97bb227042
|
Migrates to Rust 2021
|
3 years ago |
|
Tpt
|
f6dbb96327
|
Makes Clippy happy
|
3 years ago |
|
Tpt
|
d27d62f973
|
Evaluation: Adds size_hint to iterators
Removes some iterators that could be written as filter
|
3 years ago |
|
Tpt
|
6d4bb36425
|
Uses Reduced everywhere cardinality is not important
Allows avoiding dumb downstream duplicated work and should hopefully unlock optimizations in the future.
|
3 years ago |
|
Tpt
|
505980f026
|
Implements reduced operation during evaluation
Cheap streaming implementation (no malloc needed)
|
3 years ago |
|
Tpt
|
47c4734d51
|
Uses cheaper key in hash joins
|
3 years ago |
|
Tpt
|
4673cd7979
|
Takes care of VALUES when choosing joins
|
3 years ago |
|
Tpt
|
32660ecfba
|
Use HashJoin in BGP if no shared variables
Allows avoiding reading multiple times the same table
|
3 years ago |
|
Tpt
|
836a7ad4a1
|
Push filters as deep as possible
Allows to remove some possible bindings as soon as possible
|
3 years ago |
