armdb 0.4.1

sharded bitcask key-value storage optimized for NVMe
Documentation 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30

/// Common trait for armdb collections managed by [`Db`](super::tree_db::Db).
///
/// Implemented for `TypedTree`, `TypedMap`, `ZeroTree`, `ZeroMap`
/// where the value type implements [`CollectionMeta`](crate::CollectionMeta).
pub trait Collection: Send + Sync {
    /// Collection name (from `CollectionMeta::NAME`).
    fn name(&self) -> &str;
    /// Number of entries in the collection.
    fn len(&self) -> usize;
    fn is_empty(&self) -> bool {
        self.len() == 0
    }
    /// Run a compaction pass across all shards.
    fn compact(&self) -> crate::DbResult<usize>;
    /// Flush in-memory write buffers and pending hint data so an immediate
    /// process exit will not lose committed writes. No default — every impl
    /// must provide this to avoid silent durability gaps.
    fn flush(&self) -> crate::DbResult<()>;
    /// Lightest operation that bounds the periodic durability window for
    /// background use. Backend-specific:
    /// - Bitcask: flush userspace write buffers to the OS page cache (no fsync,
    ///   no hint regeneration) — bounds process-crash (panic/kill/OOM) loss.
    /// - FixedStore: fdatasync dirty pages (data is already in the page cache
    ///   via pwrite) — bounds power-loss loss for idle tail writes.
    ///
    /// Unlike [`flush`](Self::flush), this must NOT rebuild hint files or fsync
    /// the active data file on Bitcask, so it is cheap enough to run on a short
    /// interval. No default — every impl must provide it.
    fn periodic_flush(&self) -> crate::DbResult<()>;
}