lance-nd 0.1.0

N-dimensional datasets stored in Lance, one dataset or chunk per row
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
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262

//! Delete rows, compact the table and clean up old versions.

use std::sync::Arc;
use std::time::Duration;

use arrow::array::{ArrayRef, AsArray, Float64Array, Int64Array};
use arrow::datatypes::{Field, Int64Type, Schema};
use futures::stream;
use lance::Dataset;
use lance_nd::datafusion::functions::core::expr_fn::get_field;
use lance_nd::datafusion::logical_expr::{col, lit};
use lance_nd::{
    CompactOptions, NdWriter, WriteOptions, cleanup_old_versions, compact, delete, delete_expr,
    scan,
};
use nd_arrow_array::{Attributes, Dimensions, NdArrowArray, NdRecordBatch};

fn variable(name: &str, array: ArrayRef) -> (Field, NdArrowArray) {
    let dims = Dimensions::try_new([(format!("{name}_dim"), array.len())]).unwrap();
    let field = Field::new(name, array.data_type().clone(), true);
    (field, NdArrowArray::try_new(array, dims).unwrap())
}

/// Row `id` with the global attribute `id = r<id>`. Even rows have a large
/// `temp` with units `degF` in rows divisible by 4, else `degC`.
fn row(id: i64) -> NdRecordBatch {
    let mut variables = vec![variable("id", Arc::new(Int64Array::from(vec![id])))];
    if id % 2 == 0 {
        let values = Float64Array::from_iter_values((0..5000).map(|i| (id * 10_000 + i) as f64));
        let (field, temp) = variable("temp", Arc::new(values));
        let units = if id % 4 == 0 { "degF" } else { "degC" };
        variables.push((
            field,
            temp.with_attributes(Attributes::new().with("units", units)),
        ));
    }
    let (fields, arrays): (Vec<_>, Vec<_>) = variables.into_iter().unzip();
    NdRecordBatch::try_new_inferred(Arc::new(Schema::new(fields)), arrays)
        .unwrap()
        .with_attributes(Attributes::new().with("id", format!("r{id}")))
}

/// A table of `rows` rows, one fragment per row.
async fn table(dir: &tempfile::TempDir, rows: i64, stats: bool) -> Dataset {
    let uri = dir.path().join("table.lance").to_str().unwrap().to_string();
    let options = WriteOptions {
        stats,
        max_rows_per_group: 1,
        target_fragment_bytes: 1,
        ..Default::default()
    };
    let mut writer = NdWriter::create(&uri, &options).await.unwrap();
    writer
        .write(stream::iter((0..rows).map(|id| Ok(row(id)))))
        .await
        .unwrap();
    writer.finish().await.unwrap()
}

async fn ids(dataset: &Dataset) -> Vec<i64> {
    let rows = scan(dataset).try_collect().await.unwrap();
    rows.iter()
        .map(|row| {
            let id = row.column_by_name("id").expect("an id");
            id.values().as_primitive::<Int64Type>().value(0)
        })
        .collect()
}

#[tokio::test]
async fn delete_by_global_attribute() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 4, false).await;
    let version = dataset.version().version;
    let deleted = delete(&mut dataset, "__global_attrs__.id = 'r1'")
        .await
        .unwrap();
    assert_eq!(deleted, 1);
    assert_eq!(ids(&dataset).await, [0, 2, 3]);
    assert!(dataset.version().version > version);
    let reopened = Dataset::open(dataset.uri()).await.unwrap();
    assert_eq!(ids(&reopened).await, [0, 2, 3]);
}

#[tokio::test]
async fn delete_expr_by_variable_attribute() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 6, false).await;
    let units = get_field(get_field(col("temp"), "attrs"), "units");
    let deleted = delete_expr(&mut dataset, units.eq(lit("degF")))
        .await
        .unwrap();
    assert_eq!(deleted, 2);
    assert_eq!(ids(&dataset).await, [1, 2, 3, 5]);
}

#[tokio::test]
async fn delete_without_match() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 3, false).await;
    let deleted = delete(&mut dataset, "__global_attrs__.id = 'nope'")
        .await
        .unwrap();
    assert_eq!(deleted, 0);
    assert_eq!(ids(&dataset).await, [0, 1, 2]);
}

#[tokio::test]
async fn bad_predicate_fails() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 3, false).await;
    assert!(delete(&mut dataset, "nope = 1").await.is_err());
    assert_eq!(ids(&dataset).await, [0, 1, 2]);
}

#[tokio::test]
async fn compact_merges_fragments_and_drops_deleted_rows() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 12, true).await;
    assert_eq!(dataset.get_fragments().len(), 12);
    delete(&mut dataset, "__global_attrs__.id IN ('r3', 'r4', 'r7')")
        .await
        .unwrap();
    // Lance drops a fragment when each of its rows is deleted.
    assert_eq!(dataset.get_fragments().len(), 9);
    let expected: Vec<NdRecordBatch> = scan(&dataset).try_collect().await.unwrap();

    let metrics = compact(&mut dataset, &CompactOptions::default())
        .await
        .unwrap();
    assert_eq!(metrics.fragments_removed, 9);
    assert_eq!(dataset.get_fragments().len(), 1);
    assert_eq!(dataset.count_deleted_rows().await.unwrap(), 0);
    let rows: Vec<NdRecordBatch> = scan(&dataset).try_collect().await.unwrap();
    assert_eq!(rows.len(), expected.len());
    for (row, want) in rows.iter().zip(&expected) {
        assert_eq!(row.schema(), want.schema());
        assert_eq!(row.columns(), want.columns());
        assert_eq!(row.attributes(), want.attributes());
    }

    // The stats still prune after the compaction.
    let filter = get_field(col("temp"), "values").gt(lit(100_000.0));
    let pruned = scan(&dataset)
        .filter_expr(filter)
        .try_collect()
        .await
        .unwrap();
    let pruned: Vec<i64> = pruned
        .iter()
        .map(|row| {
            row.column_by_name("id")
                .unwrap()
                .values()
                .as_primitive::<Int64Type>()
                .value(0)
        })
        .collect();
    assert_eq!(pruned, [10]);
}

#[tokio::test]
async fn compact_splits_a_large_fragment_by_size() {
    let dir = tempfile::tempdir().unwrap();
    let uri = dir.path().join("table.lance").to_str().unwrap().to_string();
    let mut writer = NdWriter::create(&uri, &WriteOptions::default())
        .await
        .unwrap();
    writer
        .write(stream::iter((0..12).map(|id| Ok(row(id)))))
        .await
        .unwrap();
    let mut dataset = writer.finish().await.unwrap();
    assert_eq!(dataset.get_fragments().len(), 1);
    let expected: Vec<NdRecordBatch> = scan(&dataset).try_collect().await.unwrap();

    // Far below the size of one row: each row becomes a fragment.
    let options = CompactOptions {
        target_fragment_bytes: 16,
        ..Default::default()
    };
    let metrics = compact(&mut dataset, &options).await.unwrap();
    assert_eq!(metrics.fragments_removed, 1);
    assert_eq!(dataset.get_fragments().len(), 12);
    let rows: Vec<NdRecordBatch> = scan(&dataset).try_collect().await.unwrap();
    assert_eq!(rows.len(), expected.len());
    for (row, want) in rows.iter().zip(&expected) {
        assert_eq!(row.columns(), want.columns());
        assert_eq!(row.attributes(), want.attributes());
    }
}

#[tokio::test]
async fn compact_keeps_fragments_near_the_target() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 12, false).await;
    let options = CompactOptions {
        target_fragment_bytes: 1,
        ..Default::default()
    };
    let metrics = compact(&mut dataset, &options).await.unwrap();
    assert_eq!(metrics.fragments_removed, 0);
    assert_eq!(dataset.get_fragments().len(), 12);
}

#[tokio::test]
async fn cleanup_removes_old_versions() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 6, false).await;
    delete(&mut dataset, "__global_attrs__.id = 'r0'")
        .await
        .unwrap();
    compact(&mut dataset, &CompactOptions::default())
        .await
        .unwrap();
    let stats = cleanup_old_versions(&dataset, Duration::ZERO)
        .await
        .unwrap();
    assert!(stats.old_versions > 0, "{stats:?}");
    assert!(stats.bytes_removed > 0, "{stats:?}");
    assert_eq!(ids(&dataset).await, [1, 2, 3, 4, 5]);
    let reopened = Dataset::open(dataset.uri()).await.unwrap();
    assert_eq!(ids(&reopened).await, [1, 2, 3, 4, 5]);
}

#[tokio::test]
async fn compact_rewrites_a_fragment_with_deleted_rows() {
    let dir = tempfile::tempdir().unwrap();
    let uri = dir.path().join("table.lance").to_str().unwrap().to_string();
    let mut writer = NdWriter::create(&uri, &WriteOptions::default())
        .await
        .unwrap();
    writer
        .write(stream::iter((0..8).map(|id| Ok(row(id)))))
        .await
        .unwrap();
    let mut dataset = writer.finish().await.unwrap();
    assert_eq!(dataset.get_fragments().len(), 1);
    delete(&mut dataset, "__global_attrs__.id IN ('r2', 'r5')")
        .await
        .unwrap();
    assert_eq!(dataset.count_deleted_rows().await.unwrap(), 2);

    let metrics = compact(&mut dataset, &CompactOptions::default())
        .await
        .unwrap();
    assert_eq!(metrics.fragments_removed, 1);
    assert_eq!(dataset.get_fragments().len(), 1);
    assert_eq!(dataset.count_deleted_rows().await.unwrap(), 0);
    assert_eq!(ids(&dataset).await, [0, 1, 3, 4, 6, 7]);
}

#[tokio::test]
async fn delete_by_value() {
    let dir = tempfile::tempdir().unwrap();
    let mut dataset = table(&dir, 4, false).await;
    let deleted = delete(&mut dataset, "array_has(temp.values, 20001.0)")
        .await
        .unwrap();
    assert_eq!(deleted, 1);
    assert_eq!(ids(&dataset).await, [0, 1, 3]);
}