Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

feat: Prefix Bloom Filter #103

Open
wants to merge 2 commits into
base: main
Choose a base branch
from
Open

feat: Prefix Bloom Filter #103

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
4d4941d
feat: Prefix Bloom Filter
khlopkov Jan 18, 2025
8ce4dd7
Update mod.rs
marvin-j97 Jan 22, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
11 changes: 6 additions & 5 deletions src/blob_tree/mod.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -125,9 +125,9 @@ impl BlobTree {
// to the tree
}

let iter = self
.index
.create_internal_range::<&[u8], RangeFull>(&.., Some(seqno), None);
let iter =
self.index
.create_internal_range::<&[u8], RangeFull>(&.., Some(seqno), None, None);

// Stores the max seqno of every blob file
let mut seqno_map = crate::HashMap::<SegmentId, SeqNo>::default();
Expand Down Expand Up @@ -330,6 +330,7 @@ impl AbstractTree for BlobTree {
data_block_size: self.index.config.data_block_size,
index_block_size: self.index.config.index_block_size,
folder: lsm_segment_folder,
prefix_extractor: self.index.prefix_extractor.clone(),
})?
.use_compression(self.index.config.compression);

Expand Down Expand Up @@ -549,7 +550,7 @@ impl AbstractTree for BlobTree {
Box::new(
self.index
.0
.create_range(&range, Some(seqno), index)
.create_range(&range, Some(seqno), index, None)
.map(move |item| resolve_value_handle(&vlog, item)),
)
}
Expand Down Expand Up @@ -577,7 +578,7 @@ impl AbstractTree for BlobTree {
Box::new(
self.index
.0
.create_range(&range, None, None)
.create_range(&range, None, None, None)
.map(move |item| resolve_value_handle(&vlog, item)),
)
}
Expand Down
6 changes: 6 additions & 0 deletions src/compaction/worker.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,6 +9,7 @@ use crate::{
level_manifest::LevelManifest,
level_reader::LevelReader,
merge::{BoxedIterator, Merger},
prefix_extractor::PrefixExtractor,
segment::{
block_index::{
full_index::FullBlockIndex, two_level_index::TwoLevelBlockIndex, BlockIndexImpl,
Expand Down Expand Up @@ -50,6 +51,8 @@ pub struct Options {

/// Evicts items that are older than this seqno (MVCC GC).
pub eviction_seqno: u64,

pub prefix_extractor: Option<Arc<dyn PrefixExtractor>>,
}

impl Options {
Expand All @@ -63,6 +66,7 @@ impl Options {
stop_signal: tree.stop_signal.clone(),
strategy,
eviction_seqno: 0,
prefix_extractor: tree.prefix_extractor.clone(),
}
}
}
Expand Down Expand Up @@ -143,6 +147,7 @@ fn create_compaction_stream<'a>(
&(Unbounded, Unbounded),
(Some(lo), Some(hi)),
crate::segment::value_block::CachePolicy::Read,
None,
)));

found += hi - lo + 1;
Expand Down Expand Up @@ -251,6 +256,7 @@ fn merge_segments(
segment_id: 0, // TODO: this is never used in MultiWriter
data_block_size: opts.config.data_block_size,
index_block_size: opts.config.index_block_size,
prefix_extractor: opts.prefix_extractor.clone(),
},
) else {
log::error!("Compaction failed");
Expand Down
13 changes: 13 additions & 0 deletions src/config.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,6 +5,7 @@
use crate::{
descriptor_table::FileDescriptorTable,
path::absolute_path,
prefix_extractor::{self, PrefixExtractor},
segment::meta::{CompressionType, TableType},
BlobTree, BlockCache, Tree,
};
Expand Down Expand Up @@ -102,6 +103,10 @@ pub struct Config {
/// Descriptor table to use
#[doc(hidden)]
pub descriptor_table: Arc<FileDescriptorTable>,

/// Custom implementation to extract prefix from a key
/// when using read or updates in bloom filter to optimize DB scans by prefix
pub prefix_extractor: Option<Arc<dyn PrefixExtractor>>,
}

impl Default for Config {
Expand All @@ -123,6 +128,7 @@ impl Default for Config {
blob_cache: Arc::new(BlobCache::with_capacity_bytes(/* 16 MiB */ 16 * 1_024 * 1_024)),
blob_file_target_size: /* 64 MiB */ 64 * 1_024 * 1_024,
blob_file_separation_threshold: /* 4 KiB */ 4 * 1_024,
prefix_extractor: None,
}
}
}
Expand Down Expand Up @@ -304,6 +310,13 @@ impl Config {
self
}

#[must_use]
#[doc(hidden)]
pub fn prefix_extractor(mut self, prefix_extractor: Arc<dyn PrefixExtractor>) -> Self {
self.prefix_extractor = Some(prefix_extractor);
self
}

/// Opens a tree using the config.
///
/// # Errors
Expand Down
89 changes: 53 additions & 36 deletions src/level_reader.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -4,7 +4,7 @@

use crate::{
level_manifest::level::Level,
segment::{range::Range, value_block::CachePolicy},
segment::{range::Range, value_block::CachePolicy, Segment},
InternalValue, UserKey,
};
use std::{ops::Bound, sync::Arc};
Expand All @@ -17,6 +17,7 @@ pub struct LevelReader {
lo_reader: Option<Range>,
hi_reader: Option<Range>,
cache_policy: CachePolicy,
prefix_hash: Option<(u64, u64)>,
}

impl LevelReader {
Expand All @@ -25,6 +26,7 @@ impl LevelReader {
level: Arc<Level>,
range: &(Bound<UserKey>, Bound<UserKey>),
cache_policy: CachePolicy,
prefix_hash: Option<(u64, u64)>,
) -> Self {
assert!(!level.is_empty(), "level reader cannot read empty level");

Expand All @@ -39,10 +41,17 @@ impl LevelReader {
lo_reader: None,
hi_reader: None,
cache_policy,
prefix_hash,
};
};

Self::from_indexes(level, range, (Some(lo), Some(hi)), cache_policy)
Self::from_indexes(
level,
range,
(Some(lo), Some(hi)),
cache_policy,
prefix_hash,
)
}

#[must_use]
Expand All @@ -51,6 +60,7 @@ impl LevelReader {
range: &(Bound<UserKey>, Bound<UserKey>),
(lo, hi): (Option<usize>, Option<usize>),
cache_policy: CachePolicy,
prefix_hash: Option<(u64, u64)>,
) -> Self {
let lo = lo.unwrap_or_default();
let hi = hi.unwrap_or(level.len() - 1);
Expand All @@ -74,8 +84,14 @@ impl LevelReader {
lo_reader: Some(lo_reader),
hi_reader,
cache_policy,
prefix_hash,
}
}

fn may_segment_contain_hash(&self, segment: &Segment) -> bool {
self.prefix_hash
.map_or(true, |hash| segment.may_contain_hash(hash))
}
}

impl Iterator for LevelReader {
Expand All @@ -90,24 +106,20 @@ impl Iterator for LevelReader {

// NOTE: Lo reader is empty, get next one
self.lo_reader = None;
self.lo += 1;

if self.lo < self.hi {
self.lo_reader = Some(
self.segments
.get(self.lo)
.expect("should exist")
.iter()
.cache_policy(self.cache_policy),
);
}
} else if let Some(hi_reader) = &mut self.hi_reader {
} else if self.lo == self.hi {
// NOTE: We reached the hi marker, so consume from it instead
//
// If it returns nothing, it is empty, so we are done
return hi_reader.next();
return self.hi_reader.as_mut().and_then(|r| r.next());
} else {
return None;
self.lo += 1;

if self.lo < self.hi {
let segment = self.segments.get(self.lo).expect("should exist");
if self.may_segment_contain_hash(segment) {
self.lo_reader = Some(segment.iter().cache_policy(self.cache_policy));
}
}
}
}
}
Expand All @@ -121,26 +133,22 @@ impl DoubleEndedIterator for LevelReader {
return Some(item);
}

// NOTE: Hi reader is empty, get orev one
// NOTE: Hi reader is empty, get the previous one
self.hi_reader = None;
self.hi -= 1;

if self.lo < self.hi {
self.hi_reader = Some(
self.segments
.get(self.hi)
.expect("should exist")
.iter()
.cache_policy(self.cache_policy),
);
}
} else if let Some(lo_reader) = &mut self.lo_reader {
} else if self.lo == self.hi {
// NOTE: We reached the lo marker, so consume from it instead
//
// If it returns nothing, it is empty, so we are done
return lo_reader.next_back();
return self.lo_reader.as_mut().and_then(|r| r.next_back());
} else {
return None;
self.hi -= 1;

if self.lo < self.hi {
let segment = self.segments.get(self.hi).expect("should exist");
if self.may_segment_contain_hash(segment) {
self.hi_reader = Some(segment.iter().cache_policy(self.cache_policy));
}
}
}
}
}
Expand Down Expand Up @@ -190,8 +198,12 @@ mod tests {

#[allow(clippy::unwrap_used)]
{
let multi_reader =
LevelReader::new(level.clone(), &(Unbounded, Unbounded), CachePolicy::Read);
let multi_reader = LevelReader::new(
level.clone(),
&(Unbounded, Unbounded),
CachePolicy::Read,
None,
);

let mut iter = multi_reader.flatten();

Expand All @@ -211,8 +223,12 @@ mod tests {

#[allow(clippy::unwrap_used)]
{
let multi_reader =
LevelReader::new(level.clone(), &(Unbounded, Unbounded), CachePolicy::Read);
let multi_reader = LevelReader::new(
level.clone(),
&(Unbounded, Unbounded),
CachePolicy::Read,
None,
);

let mut iter = multi_reader.rev().flatten();

Expand All @@ -232,7 +248,8 @@ mod tests {

#[allow(clippy::unwrap_used)]
{
let multi_reader = LevelReader::new(level, &(Unbounded, Unbounded), CachePolicy::Read);
let multi_reader =
LevelReader::new(level, &(Unbounded, Unbounded), CachePolicy::Read, None);

let mut iter = multi_reader.flatten();

Expand Down
3 changes: 3 additions & 0 deletions src/lib.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -133,6 +133,9 @@ mod block_cache;
#[cfg(feature = "bloom")]
pub mod bloom;

#[doc(hidden)]
pub mod prefix_extractor;

#[doc(hidden)]
pub mod coding;
pub mod compaction;
Expand Down
11 changes: 11 additions & 0 deletions src/prefix_extractor.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
/// A trait allowing to customize prefix extraction in operations with bloom filter.
/// It defines how prefix should be extracted from the key, when update or read
/// a bloom filter.
pub trait PrefixExtractor: Sync + Send {
/// Extracts prefix from original key
fn transform<'a>(&self, key: &'a [u8]) -> &'a [u8];
/// Checks if a key is in domain and prefix can be extracted from it.
/// For example if `PrefixExtractor` suppose to extract first 4 bytes from key,
/// `in_domain(&[0, 2, 3])` should return false
fn in_domain(&self, key: &[u8]) -> bool;
Copy link
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I don't think this is necessary.
If a key is [1, 2, 3] and we extract the first 4 bytes, we would just take the entire string, so:

let len = Math.min(key.len(), 4);
return key.slice(0, len);

Copy link
Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Let me better explain why is this needed in more practical example:

Imagine we have a database where keys are constructed using the following template: {datatype}#{id} and a prefix extractor that would extract {datatype}# prefix from a key to optimize scans of all rows of datatype#. Then, in table were populated following rows table1#a, table1#b, table2#a. The keys for bloom filter would be table1#, table2# in this case. Then, if we use prefix scan by prefixes table1# or table2# everything will work correctly. But if scan is performed with table prefix without in_domain the full prefix will be extracted and will try to check bloom filter by table key, which may return false negative. in_domain in this case will help to prevent such cases and will use Bloom Filter only in case when prefix is in domain. In this example everything starting with {datatype}# will be considered as in_domain and use Bloom Filter, other prefixes will be not in_domain and will just use normal prefix search ignoring Bloom Filter

Copy link
Contributor

@marvin-j97 marvin-j97 Jan 20, 2025 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Makes sense.

However, for the prefix extractor to support structured prefixes, the trait needs to return an Iterator over extracted prefixed (see #97 and fjall-rs/fjall#116). At that point, the trait could just return std::iter::empty, which would signal that the key is "not in domain".

For example we may extract prefixes, like:

eu#germany#berlin

-> extract "eu#" as prefix
-> also extract "eu#germany#"

Now prefix searches over both continent and [continent + country] can be filtered by the prefix filter.

}
19 changes: 14 additions & 5 deletions src/range.rs
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -83,6 +83,7 @@ impl DoubleEndedIterator for TreeIter {
fn collect_disjoint_tree_with_range(
level_manifest: &LevelManifest,
bounds: &(Bound<UserKey>, Bound<UserKey>),
prefix_hash: Option<(u64, u64)>,
) -> MultiReader<LevelReader> {
debug_assert!(level_manifest.is_disjoint());

Expand Down Expand Up @@ -120,7 +121,7 @@ fn collect_disjoint_tree_with_range(

let readers = levels
.into_iter()
.map(|lvl| LevelReader::new(lvl, bounds, CachePolicy::Write))
.map(|lvl| LevelReader::new(lvl, bounds, CachePolicy::Write, prefix_hash))
.collect();

MultiReader::new(readers)
Expand All @@ -134,6 +135,7 @@ impl TreeIter {
bounds: (Bound<UserKey>, Bound<UserKey>),
seqno: Option<SeqNo>,
level_manifest: ArcRwLockReadGuardian<LevelManifest>,
prefix_hash: Option<(u64, u64)>,
) -> Self {
Self::new(guard, |lock| {
let lo = match &bounds.0 {
Expand Down Expand Up @@ -185,7 +187,8 @@ impl TreeIter {

// NOTE: Optimize disjoint trees (e.g. timeseries) to only use a single MultiReader.
if level_manifest.is_disjoint() {
let reader = collect_disjoint_tree_with_range(&level_manifest, &bounds);
let reader =
collect_disjoint_tree_with_range(&level_manifest, &bounds, prefix_hash);

if let Some(seqno) = seqno {
iters.push(Box::new(reader.filter(move |item| match item {
Expand All @@ -199,8 +202,12 @@ impl TreeIter {
for level in &level_manifest.levels {
if level.is_disjoint {
if !level.is_empty() {
let reader =
LevelReader::new(level.clone(), &bounds, CachePolicy::Write);
let reader = LevelReader::new(
level.clone(),
&bounds,
CachePolicy::Write,
prefix_hash,
);

if let Some(seqno) = seqno {
iters.push(Box::new(reader.filter(move |item| match item {
Expand All @@ -213,7 +220,9 @@ impl TreeIter {
}
} else {
for segment in &level.segments {
if segment.check_key_range_overlap(&bounds) {
let may_contain_hash =
prefix_hash.map_or(true, |hash| segment.may_contain_hash(hash));
if may_contain_hash && segment.check_key_range_overlap(&bounds) {
let reader = segment.range(bounds.clone());

if let Some(seqno) = seqno {
Expand Down
Loading
Loading