nodedb 0.2.0

Local-first, real-time, edge-to-cloud hybrid database for multi-modal 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

// SPDX-License-Identifier: BUSL-1.1

// Integration tests for the structural sparse-tile codec introduced in segment
// format version 4. Tests cover: large roundtrip, legacy payload compatibility,
// compression ratio, and mixed sparse+dense segment decoding.

use nodedb_array::codec::tile_encode::encode_sparse_tile;
use nodedb_array::schema::ArraySchemaBuilder;
use nodedb_array::schema::attr_spec::{AttrSpec, AttrType};
use nodedb_array::schema::dim_spec::{DimSpec, DimType};
use nodedb_array::tile::dense_tile::DenseTile;
use nodedb_array::tile::sparse_tile::{RowKind, SparseRow, SparseTile, SparseTileBuilder};
use nodedb_array::types::TileId;
use nodedb_array::types::cell_value::value::CellValue;
use nodedb_array::types::coord::value::CoordValue;
use nodedb_array::types::domain::{Domain, DomainBound};
use nodedb_array::{SegmentReader, SegmentWriter, TilePayload};
use nodedb_types::{OPEN_UPPER, Surrogate};

fn genomic_schema() -> nodedb_array::ArraySchema {
    ArraySchemaBuilder::new("genome")
        .dim(DimSpec::new(
            "chrom",
            DimType::Int64,
            Domain::new(DomainBound::Int64(0), DomainBound::Int64(24)),
        ))
        .dim(DimSpec::new(
            "pos",
            DimType::Int64,
            Domain::new(DomainBound::Int64(0), DomainBound::Int64(300_000_000)),
        ))
        .attr(AttrSpec::new("qual", AttrType::Float64, true))
        .tile_extents(vec![1, 1_000_000])
        .build()
        .unwrap()
}

fn simple_schema() -> nodedb_array::ArraySchema {
    ArraySchemaBuilder::new("s")
        .dim(DimSpec::new(
            "x",
            DimType::Int64,
            Domain::new(DomainBound::Int64(0), DomainBound::Int64(1_000_000)),
        ))
        .attr(AttrSpec::new("v", AttrType::Int64, true))
        .tile_extents(vec![1000])
        .build()
        .unwrap()
}

/// Test 1: 10k-cell tile write + read exact equality.
#[test]
fn ten_k_tile_roundtrip_exact_equality() {
    let s = genomic_schema();
    let mut b = SparseTileBuilder::new(&s);
    for i in 0..10_000usize {
        b.push_row(SparseRow {
            coord: &[
                CoordValue::Int64((i % 24) as i64),
                CoordValue::Int64(i as i64 * 100),
            ],
            attrs: &[CellValue::Float64(i as f64 * 0.1)],
            surrogate: Surrogate::new(i as u32),
            valid_from_ms: i as i64 * 10,
            valid_until_ms: OPEN_UPPER,
            kind: RowKind::Live,
        })
        .unwrap();
    }
    let tile = b.build();

    let mut seg_writer = SegmentWriter::new(0xDEAD_BEEF);
    seg_writer
        .append_sparse(TileId::snapshot(1), &tile)
        .unwrap();
    let bytes = seg_writer.finish(None).unwrap();

    let reader = SegmentReader::open(&bytes).unwrap();
    assert_eq!(reader.tile_count(), 1);
    match reader.read_tile(0).unwrap() {
        TilePayload::Sparse(decoded) => {
            assert_eq!(decoded.surrogates, tile.surrogates);
            assert_eq!(decoded.valid_from_ms, tile.valid_from_ms);
            assert_eq!(decoded.valid_until_ms, tile.valid_until_ms);
            assert_eq!(decoded.row_kinds, tile.row_kinds);
            assert_eq!(decoded.attr_cols.len(), tile.attr_cols.len());
            assert_eq!(decoded.attr_cols[0], tile.attr_cols[0]);
            assert_eq!(decoded.dim_dicts.len(), tile.dim_dicts.len());
            for (d, o) in decoded.dim_dicts.iter().zip(tile.dim_dicts.iter()) {
                assert_eq!(d.indices, o.indices);
                assert_eq!(d.values, o.values);
            }
        }
        TilePayload::Dense(_) => panic!("expected Sparse"),
    }
}

/// Test 2: Legacy tile payload (raw zerompk msgpack) is still decodable by the
/// tag-dispatch layer, even after the segment format was bumped to v4.
///
/// The reader's logic is `match peek_tag { Some(_) => decode_sparse_tile,
/// None => zerompk::from_msgpack }`. This test proves:
///   1. `peek_tag` on a real zerompk-of-`SparseTile` payload returns `None`
///      (so the reader will take the legacy fallback branch, regardless of
///      whether zerompk emits a map-header or array-header byte).
///   2. `zerompk::from_msgpack` — the body of that None branch — decodes the
///      payload back into the original `SparseTile`.
///
/// Together these are equivalent to invoking the reader's full dispatch path
/// with a v3-style payload.
#[test]
fn legacy_v3_tile_payload_readable() {
    let s = simple_schema();
    let mut b = SparseTileBuilder::new(&s);
    for i in 0..5 {
        b.push_row(SparseRow {
            coord: &[CoordValue::Int64(i)],
            attrs: &[CellValue::Int64(i * 10)],
            surrogate: Surrogate::ZERO,
            valid_from_ms: 0,
            valid_until_ms: OPEN_UPPER,
            kind: RowKind::Live,
        })
        .unwrap();
    }
    let tile = b.build();

    // Simulate the v3 writer: raw zerompk with no tag byte.
    let raw_msgpack = zerompk::to_msgpack_vec(&tile).unwrap();

    // The codec's tag dispatch should fall back to zerompk for this payload
    // (peek_tag returns None for any non-tag byte; reader then tries zerompk).
    let decoded = nodedb_array::codec::tag::peek_tag(&raw_msgpack);
    assert_eq!(
        decoded, None,
        "legacy payload should yield None from peek_tag"
    );

    // Directly decode via zerompk to confirm the legacy path produces the
    // same tile the new path would.
    let recovered: SparseTile = zerompk::from_msgpack(&raw_msgpack).unwrap();
    assert_eq!(recovered.surrogates, tile.surrogates);
    assert_eq!(recovered.attr_cols, tile.attr_cols);
}

/// Test 3: Compression ratio — structural codec ≤ 30% of raw msgpack for a
/// sorted 1k-cell coordinate-heavy tile.
#[test]
fn structural_codec_compresses_sorted_tile() {
    let s = genomic_schema();
    let mut b = SparseTileBuilder::new(&s);
    // Sorted: chrom fixed at 1, pos monotonically increasing.
    for i in 0..1_000usize {
        b.push_row(SparseRow {
            coord: &[
                CoordValue::Int64(1),
                CoordValue::Int64(100_000 + i as i64 * 10),
            ],
            attrs: &[CellValue::Float64(i as f64)],
            surrogate: Surrogate::new(i as u32),
            valid_from_ms: i as i64 * 1000,
            valid_until_ms: OPEN_UPPER,
            kind: RowKind::Live,
        })
        .unwrap();
    }
    let tile = b.build();

    let raw_size = zerompk::to_msgpack_vec(&tile).unwrap().len();
    let mut structural_buf = Vec::new();
    encode_sparse_tile(&tile, &mut structural_buf).unwrap();
    let structural_size = structural_buf.len();

    // Must compress to ≤ 30 % of raw msgpack on a sorted coord-heavy tile.
    // Coords go through delta-varint, surrogates through fastlanes,
    // valid_from_ms through gorilla, and the f64 attr column through gorilla
    // XOR — all four dominant byte sources have monotonic-aware codecs.
    let threshold = raw_size * 30 / 100;
    assert!(
        structural_size <= threshold,
        "structural ({structural_size} bytes) > 30% of raw ({raw_size} bytes, threshold {threshold})"
    );
}

/// Test 4: Mixed segment (Sparse + Dense tiles) — both decode correctly.
#[test]
fn mixed_sparse_dense_segment_decodes() {
    let s = simple_schema();

    let mut b = SparseTileBuilder::new(&s);
    for i in 0..20 {
        b.push_row(SparseRow {
            coord: &[CoordValue::Int64(i)],
            attrs: &[CellValue::Int64(i * 7)],
            surrogate: Surrogate::new(i as u32),
            valid_from_ms: 0,
            valid_until_ms: OPEN_UPPER,
            kind: RowKind::Live,
        })
        .unwrap();
    }
    let sparse = b.build();
    let dense = DenseTile::empty(&s);

    let mut w = SegmentWriter::new(0xCAFE_BABE);
    w.append_sparse(TileId::new(1, 100), &sparse).unwrap();
    w.append_dense(TileId::new(2, 100), &dense).unwrap();
    let bytes = w.finish(None).unwrap();

    let r = SegmentReader::open(&bytes).unwrap();
    assert_eq!(r.tile_count(), 2);

    match r.read_tile(0).unwrap() {
        TilePayload::Sparse(t) => {
            assert_eq!(t.surrogates.len(), 20);
            assert_eq!(t.attr_cols[0][0], CellValue::Int64(0));
            assert_eq!(t.attr_cols[0][1], CellValue::Int64(7));
        }
        TilePayload::Dense(_) => panic!("tile 0 should be Sparse"),
    }

    match r.read_tile(1).unwrap() {
        TilePayload::Dense(t) => {
            // DenseTile::empty for a 1D schema with 1000-extent = 1000 cells.
            assert_eq!(t.cell_count(), 1000);
        }
        TilePayload::Sparse(_) => panic!("tile 1 should be Dense"),
    }
}