hawdb-storage 0.12.0

An embedded Rust graph database for local data-plane workloads
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
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234

// Copyright 2026 Nowledge
// SPDX-License-Identifier: Apache-2.0

use super::*;
use crate::cow::CowPageWeight;
use crate::test_allocator::{AllocationObservation, AllocationStats};
use crate::version::{VersionIndex, VersionKey, VersionStamp};

fn update(id: u64, value: i64) -> GraphMutation {
    GraphMutation::SetNodeProperty {
        label: "Memory".into(),
        filter: Some(PropertyFilter::IdEq {
            value: Value::Int(i64::try_from(id).unwrap()),
        }),
        property: "value".into(),
        value: Value::Int(value),
    }
}

fn stats(stats: AllocationStats) -> serde_json::Value {
    serde_json::json!({
        "allocation_calls": stats.allocations,
        "allocated_bytes": stats.allocated_bytes,
        "deallocation_calls": stats.deallocations,
        "deallocated_bytes": stats.deallocated_bytes,
    })
}

// Run each profile in its own process when recording RSS. All observed work
// runs on this thread; fixture construction and JSON output are not observed.
fn allocation_profile(rows: usize, age: usize) {
    const BODY_BYTES: usize = 16 * 1024;
    let mut catalog = Catalog::default();
    let mut live = GraphStore::default();
    for id in 0..rows {
        live.create_node(
            &mut catalog,
            "Memory",
            BTreeMap::from([
                ("id".into(), Value::Int(i64::try_from(id).unwrap())),
                ("value".into(), Value::Int(0)),
                ("body".into(), Value::String("x".repeat(BODY_BYTES))),
            ]),
        )
        .unwrap();
    }
    live.checkpoint(&catalog).unwrap();
    let pins = live.version_snapshot_pins.clone();
    let before = AllocationObservation::start_all();
    let reader = live.snapshot_for_read();
    let capture = before.finish_stats();
    assert!(reader.nodes.shares_storage_with(&live.nodes));
    assert!(capture.allocated_bytes < BODY_BYTES);
    let base_epoch = reader.commit_epoch();
    let mut roots = Vec::with_capacity(age);
    let mut shared_segments = Vec::with_capacity(age);
    let before = AllocationObservation::start_all();
    for value in 1..=age {
        live.commit_mutations(&mut catalog, vec![update(0, i64::try_from(value).unwrap())])
            .unwrap();
        // A changed directory/page still shares every untouched row owner.
        assert!(std::ptr::eq(
            live.nodes.get(&NodeId(1)).unwrap(),
            reader.nodes.get(&NodeId(1)).unwrap(),
        ));
        shared_segments.push(reader.nodes.shared_segment_count_with(&live.nodes));
        roots.push(live.snapshot());
    }
    let writes = before.finish_stats();
    let current_payload =
        live.version_index.estimated_bytes() - VersionIndex::default().estimated_bytes();
    let retained_payload = live.version_index.retained_history_bytes();
    assert_eq!(retained_payload, age * current_payload);
    assert!(retained_payload > 0);
    let counter = live.version_index.read_baseline(live.commit_epoch());
    assert_eq!(
        reader.node(NodeId(0)).unwrap().properties["value"],
        Value::Int(0)
    );
    assert_eq!(pins.oldest_epoch(), Some(base_epoch));
    // Closing the producer cannot retire snapshots or their history charges.
    drop(live);
    assert_eq!(counter.retained_history_bytes(), retained_payload);
    assert_eq!(reader.node_count_for_label(None), rows);
    let before = AllocationObservation::start_all();
    drop(roots);
    drop(reader);
    let retired = before.finish_stats();
    assert_eq!(counter.retained_history_bytes(), 0);
    assert_eq!(pins.oldest_epoch(), None);
    assert!(retired.deallocated_bytes >= rows * BODY_BYTES);
    println!(
        "cow_snapshot_retention {}",
        serde_json::json!({
            "rows": rows,
            "body_bytes": BODY_BYTES,
            "reader_age_commits": age,
            "capture": stats(capture),
            "updates_and_precise_snapshots": stats(writes),
            "retirement": stats(retired),
            "base_epoch": base_epoch,
            "current_version_payload_estimate": current_payload,
            "retained_version_payload_estimate": retained_payload,
            "retained_version_payload_after_retirement": counter.retained_history_bytes(),
            "shared_segments_after_each_update": shared_segments,
            "untouched_row_owner_shared": true,
            "scope": "in-memory GraphStore; calling-thread allocator traffic; conflict payload estimates exclude allocator overhead and canonical values; RSS recorded separately",
        }),
    );
}

#[test]
fn cow_snapshot_profile_small_young() {
    allocation_profile(64, 1);
}

#[test]
fn cow_snapshot_profile_small_old() {
    allocation_profile(64, 4);
}

#[test]
fn cow_snapshot_profile_large_young() {
    allocation_profile(512, 1);
}

#[test]
fn cow_snapshot_profile_large_old() {
    allocation_profile(512, 4);
}

#[test]
fn restored_savepoint_preserves_conflicts_after_delete_reinsert_and_pruning() {
    let mut catalog = Catalog::default();
    let mut live = GraphStore::default();
    for id in 0..2 {
        live.create_node(
            &mut catalog,
            "Memory",
            BTreeMap::from([
                ("id".into(), Value::Int(id)),
                ("value".into(), Value::Int(0)),
            ]),
        )
        .unwrap();
    }
    live.checkpoint(&catalog).unwrap();
    let reserve = VersionIndex::default().estimated_bytes();
    let weight =
        VersionKey::GraphNode(NodeId(0)).cow_page_bytes() + std::mem::size_of::<VersionStamp>();
    live.version_index = VersionIndex::with_history_limit(reserve + 3 * weight, 6 * weight);
    let epoch = live.commit_epoch();
    let mut stale = live.begin_mutation_transaction(&catalog);
    for value in 1..=3 {
        stale
            .stage_mutation_with_limits(update(0, value), MutationLimits::default())
            .unwrap();
    }
    let saved = stale.savepoint();
    stale
        .stage_mutation_with_limits(update(1, 99), MutationLimits::default())
        .unwrap();
    stale.restore(saved);
    assert_eq!(stale.store.commit_epoch(), epoch + 3);
    // The restored private epoch is newer than a real conflicting commit.
    // Reclaiming at that private epoch would erase the evidence it still needs.
    live.commit_mutations(
        &mut catalog,
        vec![
            GraphMutation::DeleteNode {
                label: "Memory".into(),
                filter: Some(PropertyFilter::IdEq {
                    value: Value::Int(0),
                }),
                detach: true,
            },
            GraphMutation::CreateNode {
                label: "Memory".into(),
                properties: BTreeMap::from([
                    ("id".into(), Value::Int(0)),
                    ("value".into(), Value::Int(8)),
                ]),
            },
        ],
    )
    .unwrap();
    live.checkpoint(&catalog).unwrap();
    assert_eq!(live.version_snapshot_pins.oldest_epoch(), Some(epoch));
    assert!(live
        .version_index
        .stamp(&VersionKey::GraphNode(NodeId(0)))
        .is_some());
    let before = live.commit_epoch();
    let error = live
        .commit_mutation_transaction_and_relational(
            &mut catalog,
            stale,
            RelationalTransaction::default(),
            MutationLimits::default(),
        )
        .unwrap_err();
    assert!(error.is_retryable_transaction_conflict());
    assert_eq!(live.commit_epoch(), before);
    live.reclaim_version_history();
    assert_eq!(live.version_index.retained_history_bytes(), 0);
    assert_eq!(
        live.node(NodeId(1)).unwrap().properties["value"],
        Value::Int(0)
    );
    assert_eq!(
        live.node(NodeId(2)).unwrap().properties["value"],
        Value::Int(8)
    );
    // After history removal, new disjoint workspaces still commit.
    let mut left = live.begin_mutation_transaction(&catalog);
    let mut right = live.begin_mutation_transaction(&catalog);
    left.stage_mutation_with_limits(update(1, 1), MutationLimits::default())
        .unwrap();
    right
        .stage_mutation_with_limits(update(2, 2), MutationLimits::default())
        .unwrap();
    for tx in [left, right] {
        live.commit_mutation_transaction_and_relational(
            &mut catalog,
            tx,
            RelationalTransaction::default(),
            MutationLimits::default(),
        )
        .unwrap();
    }
    live.checkpoint(&catalog).unwrap();
    assert_eq!(live.version_snapshot_pins.oldest_epoch(), None);
    assert_eq!(live.version_index.retained_history_bytes(), 0);
}