|
| 1 | +// SPDX-License-Identifier: Apache-2.0 |
| 2 | + |
| 3 | +//! Checkpoint serialization and restoration for [`SparseVectorManager`]. |
| 4 | +//! |
| 5 | +//! Persists every sparse inverted index so a cold open loads them directly — |
| 6 | +//! no rebuild pass over source documents is required. |
| 7 | +//! |
| 8 | +//! ## Key layout under `Namespace::Vector` |
| 9 | +//! |
| 10 | +//! `nodedb_types::Namespace` has no sparse-specific variant, so the sparse |
| 11 | +//! index shares the vector engine's namespace behind a distinct `sparse:` |
| 12 | +//! key prefix that cannot collide with the dense HNSW keys. |
| 13 | +//! |
| 14 | +//! | Key | Value | |
| 15 | +//! |----------------------------|-------------------------------------------------------| |
| 16 | +//! | `sparse:_indices` | MessagePack `Vec<String>` — index key list | |
| 17 | +//! | `sparse:{index_key}:docs` | MessagePack `Vec<(String, Vec<(u32, f32)>)>` | |
| 18 | +//! |
| 19 | +//! Documents are stored as their own `(dimension, weight)` entries rather than |
| 20 | +//! as posting lists: the inverted index is a pure function of the document |
| 21 | +//! vectors, so storing the source form keeps the blob minimal and makes the |
| 22 | +//! restored index structurally identical to a freshly built one. |
| 23 | +//! |
| 24 | +//! [`SparseVectorManager`]: super::manager::SparseVectorManager |
| 25 | +
|
| 26 | +use std::collections::HashMap; |
| 27 | + |
| 28 | +use nodedb_types::Namespace; |
| 29 | +use nodedb_types::error::{NodeDbError, NodeDbResult}; |
| 30 | + |
| 31 | +use crate::storage::engine::{StorageEngine, WriteOp}; |
| 32 | + |
| 33 | +use super::index::SparseInvertedIndex; |
| 34 | + |
| 35 | +/// Key holding the MessagePack list of live sparse index keys. |
| 36 | +const INDEX_LIST_KEY: &[u8] = b"sparse:_indices"; |
| 37 | + |
| 38 | +/// A single document's sparse vector as persisted. |
| 39 | +type SerDocument = (String, Vec<(u32, f32)>); |
| 40 | + |
| 41 | +/// Per-index documents key: `sparse:{index_key}:docs`. |
| 42 | +fn documents_key(index_key: &str) -> String { |
| 43 | + format!("sparse:{index_key}:docs") |
| 44 | +} |
| 45 | + |
| 46 | +/// Serialize sparse index state into write ops. |
| 47 | +/// |
| 48 | +/// Pure and synchronous, so it is safe to call while holding the manager's |
| 49 | +/// mutex guard; the caller performs the I/O after releasing the lock. |
| 50 | +/// |
| 51 | +/// Empty indexes are still listed so that a checkpoint taken after every |
| 52 | +/// document was deleted restores as empty rather than resurrecting the |
| 53 | +/// previous checkpoint's contents. |
| 54 | +pub(crate) fn serialize_sparse( |
| 55 | + indices: &HashMap<String, SparseInvertedIndex>, |
| 56 | +) -> NodeDbResult<Vec<WriteOp>> { |
| 57 | + let mut ops: Vec<WriteOp> = Vec::with_capacity(indices.len() + 1); |
| 58 | + |
| 59 | + let mut index_keys: Vec<String> = indices.keys().cloned().collect(); |
| 60 | + index_keys.sort(); |
| 61 | + let keys_bytes = zerompk::to_msgpack_vec(&index_keys) |
| 62 | + .map_err(|e| NodeDbError::serialization("msgpack", e))?; |
| 63 | + ops.push(WriteOp::Put { |
| 64 | + ns: Namespace::Vector, |
| 65 | + key: INDEX_LIST_KEY.to_vec(), |
| 66 | + value: keys_bytes, |
| 67 | + }); |
| 68 | + |
| 69 | + for (index_key, index) in indices { |
| 70 | + let documents: Vec<SerDocument> = index.documents(); |
| 71 | + let bytes = zerompk::to_msgpack_vec(&documents) |
| 72 | + .map_err(|e| NodeDbError::serialization("msgpack", e))?; |
| 73 | + ops.push(WriteOp::Put { |
| 74 | + ns: Namespace::Vector, |
| 75 | + key: documents_key(index_key).into_bytes(), |
| 76 | + value: bytes, |
| 77 | + }); |
| 78 | + } |
| 79 | + |
| 80 | + Ok(ops) |
| 81 | +} |
| 82 | + |
| 83 | +/// Write pre-serialized sparse index state to storage. |
| 84 | +pub(crate) async fn write_serialized_sparse<S>(storage: &S, ops: Vec<WriteOp>) -> NodeDbResult<()> |
| 85 | +where |
| 86 | + S: StorageEngine, |
| 87 | +{ |
| 88 | + if ops.is_empty() { |
| 89 | + return Ok(()); |
| 90 | + } |
| 91 | + storage |
| 92 | + .batch_write(&ops) |
| 93 | + .await |
| 94 | + .map_err(|e| NodeDbError::storage(format!("sparse checkpoint batch_write: {e}")))?; |
| 95 | + Ok(()) |
| 96 | +} |
| 97 | + |
| 98 | +/// Restore sparse index state from storage on cold open. |
| 99 | +/// |
| 100 | +/// Returns `None` when no checkpoint has ever been written, which the caller |
| 101 | +/// distinguishes from a checkpoint that legitimately holds no documents — only |
| 102 | +/// the former warrants a rebuild from source documents. A corrupt blob for one |
| 103 | +/// index is logged and that index restored empty rather than failing the whole |
| 104 | +/// open; the remaining indexes stay usable. |
| 105 | +pub(crate) async fn restore_sparse<S>( |
| 106 | + storage: &S, |
| 107 | +) -> NodeDbResult<Option<HashMap<String, SparseInvertedIndex>>> |
| 108 | +where |
| 109 | + S: StorageEngine, |
| 110 | +{ |
| 111 | + let Some(keys_bytes) = storage.get(Namespace::Vector, INDEX_LIST_KEY).await? else { |
| 112 | + return Ok(None); |
| 113 | + }; |
| 114 | + let Ok(index_keys) = zerompk::from_msgpack::<Vec<String>>(&keys_bytes) else { |
| 115 | + tracing::warn!("sparse checkpoint: index list undecodable — starting fresh"); |
| 116 | + return Ok(None); |
| 117 | + }; |
| 118 | + |
| 119 | + let mut indices: HashMap<String, SparseInvertedIndex> = |
| 120 | + HashMap::with_capacity(index_keys.len()); |
| 121 | + |
| 122 | + for index_key in &index_keys { |
| 123 | + let key = documents_key(index_key); |
| 124 | + let Some(bytes) = storage.get(Namespace::Vector, key.as_bytes()).await? else { |
| 125 | + indices.insert(index_key.clone(), SparseInvertedIndex::new()); |
| 126 | + continue; |
| 127 | + }; |
| 128 | + match zerompk::from_msgpack::<Vec<SerDocument>>(&bytes) { |
| 129 | + Ok(documents) => { |
| 130 | + indices.insert( |
| 131 | + index_key.clone(), |
| 132 | + SparseInvertedIndex::from_documents(documents), |
| 133 | + ); |
| 134 | + } |
| 135 | + Err(e) => { |
| 136 | + tracing::warn!( |
| 137 | + index_key, |
| 138 | + error = %e, |
| 139 | + "sparse checkpoint: document blob undecodable — index restored empty" |
| 140 | + ); |
| 141 | + indices.insert(index_key.clone(), SparseInvertedIndex::new()); |
| 142 | + } |
| 143 | + } |
| 144 | + } |
| 145 | + |
| 146 | + tracing::debug!(index_count = indices.len(), "sparse checkpoint restored"); |
| 147 | + Ok(Some(indices)) |
| 148 | +} |
| 149 | + |
| 150 | +#[cfg(test)] |
| 151 | +mod tests { |
| 152 | + use nodedb_types::SparseVector; |
| 153 | + |
| 154 | + use super::*; |
| 155 | + use crate::PagedbStorageMem; |
| 156 | + |
| 157 | + fn sv(entries: &[(u32, f32)]) -> SparseVector { |
| 158 | + SparseVector::from_entries(entries.to_vec()).expect("valid sparse vector") |
| 159 | + } |
| 160 | + |
| 161 | + #[tokio::test] |
| 162 | + async fn round_trip_through_storage() { |
| 163 | + let storage = PagedbStorageMem::open_in_memory() |
| 164 | + .await |
| 165 | + .expect("in-memory pagedb"); |
| 166 | + |
| 167 | + let mut index = SparseInvertedIndex::new(); |
| 168 | + index.insert("d1", &sv(&[(1, 0.5), (9, 0.25)])); |
| 169 | + index.insert("d2", &sv(&[(9, 1.0)])); |
| 170 | + |
| 171 | + let mut indices = HashMap::new(); |
| 172 | + indices.insert("docs:emb".to_string(), index); |
| 173 | + |
| 174 | + let ops = serialize_sparse(&indices).expect("serialize"); |
| 175 | + write_serialized_sparse(&storage, ops).await.expect("write"); |
| 176 | + |
| 177 | + let restored = restore_sparse(&storage) |
| 178 | + .await |
| 179 | + .expect("restore") |
| 180 | + .expect("checkpoint present"); |
| 181 | + let restored_index = restored.get("docs:emb").expect("index restored"); |
| 182 | + assert_eq!(restored_index.doc_count(), 2); |
| 183 | + |
| 184 | + let hits = restored_index.search(&sv(&[(9, 1.0)]), 10); |
| 185 | + assert_eq!(hits.len(), 2); |
| 186 | + assert_eq!(hits[0].doc_id, "d2"); |
| 187 | + } |
| 188 | + |
| 189 | + #[tokio::test] |
| 190 | + async fn absent_checkpoint_reports_none() { |
| 191 | + let storage = PagedbStorageMem::open_in_memory() |
| 192 | + .await |
| 193 | + .expect("in-memory pagedb"); |
| 194 | + let restored = restore_sparse(&storage).await.expect("restore"); |
| 195 | + assert!(restored.is_none()); |
| 196 | + } |
| 197 | +} |
0 commit comments