aisle 0.2.0

Metadata-driven Parquet pruning for Rust: Skip irrelevant data before reading
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
263
264
265
266
267
268
269
270
271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
333
334
335
336
337
338
339
340
341
342
343
344
345
346
347
348
349
350
351
352
353
354
355
356
357
358
359
360
361
362
363
364
365
366
367
368
369
370
371
372
373
374
375
376
377
378
379
380
381

use std::{collections::HashMap, sync::Arc};

use aisle::{AsyncBloomFilterProvider, PruneRequest};
use arrow_array::{Int64Array, RecordBatch};
use arrow_schema::{DataType, Field, Schema};
use bytes::Bytes;
use datafusion_expr::{col, lit};
use parquet::{
    arrow::AsyncArrowWriter,
    bloom_filter::Sbbf,
    file::{
        metadata::{PageIndexPolicy, ParquetMetaData, ParquetMetaDataReader},
        properties::{EnabledStatistics, WriterProperties},
    },
};

// ============================================================================
// Helper Functions
// ============================================================================

fn make_int_batch(schema: &Schema, values: &[i64]) -> RecordBatch {
    let array = Int64Array::from(values.to_vec());
    RecordBatch::try_new(Arc::new(schema.clone()), vec![Arc::new(array)]).unwrap()
}

async fn write_parquet_async(batches: &[RecordBatch], props: WriterProperties) -> Vec<u8> {
    let mut buffer: Vec<u8> = Vec::new();
    let schema = batches[0].schema();
    let mut writer = AsyncArrowWriter::try_new(&mut buffer, schema, Some(props)).unwrap();
    for batch in batches {
        writer.write(batch).await.unwrap();
    }
    writer.close().await.unwrap();
    buffer
}

fn load_metadata(bytes: &[u8]) -> ParquetMetaData {
    let bytes = Bytes::copy_from_slice(bytes);
    ParquetMetaDataReader::new()
        .with_page_index_policy(PageIndexPolicy::Skip)
        .parse_and_finish(&bytes)
        .unwrap()
}

// ============================================================================
// Mock Bloom Filter Provider
// ============================================================================

/// Mock provider that tracks which bloom filters were requested
struct MockBloomProvider {
    calls: Vec<(usize, usize)>,
    filters: HashMap<(usize, usize), Sbbf>,
}

impl MockBloomProvider {
    fn new() -> Self {
        Self {
            calls: Vec::new(),
            filters: HashMap::new(),
        }
    }

    fn calls(&self) -> &[(usize, usize)] {
        &self.calls
    }
}

impl AsyncBloomFilterProvider for MockBloomProvider {
    async fn bloom_filter(&mut self, row_group_idx: usize, column_idx: usize) -> Option<Sbbf> {
        self.calls.push((row_group_idx, column_idx));
        self.filters.get(&(row_group_idx, column_idx)).cloned()
    }
}

// ============================================================================
// Cached Bloom Filter Provider
// ============================================================================

/// Provider that caches bloom filters after first load
struct CachedBloomProvider {
    calls: Vec<(usize, usize)>,
    cache_hits: usize,
    filters: HashMap<(usize, usize), Sbbf>,
    cache: HashMap<(usize, usize), Sbbf>,
}

impl CachedBloomProvider {
    fn new() -> Self {
        Self {
            calls: Vec::new(),
            cache_hits: 0,
            filters: HashMap::new(),
            cache: HashMap::new(),
        }
    }

    fn cache_hits(&self) -> usize {
        self.cache_hits
    }
}

impl AsyncBloomFilterProvider for CachedBloomProvider {
    async fn bloom_filter(&mut self, row_group_idx: usize, column_idx: usize) -> Option<Sbbf> {
        let key = (row_group_idx, column_idx);
        self.calls.push(key);

        // Check cache first
        if let Some(filter) = self.cache.get(&key) {
            self.cache_hits += 1;
            return Some(filter.clone());
        }

        // Load and cache
        if let Some(filter) = self.filters.get(&key).cloned() {
            self.cache.insert(key, filter.clone());
            Some(filter)
        } else {
            None
        }
    }
}

// ============================================================================
// Batched Bloom Filter Provider
// ============================================================================

/// Provider that tracks batch calls
struct BatchedBloomProvider {
    single_calls: usize,
    batch_calls: usize,
    filters: HashMap<(usize, usize), Sbbf>,
}

impl BatchedBloomProvider {
    fn new() -> Self {
        Self {
            single_calls: 0,
            batch_calls: 0,
            filters: HashMap::new(),
        }
    }

    fn single_calls(&self) -> usize {
        self.single_calls
    }

    fn batch_calls(&self) -> usize {
        self.batch_calls
    }
}

impl AsyncBloomFilterProvider for BatchedBloomProvider {
    async fn bloom_filter(&mut self, row_group_idx: usize, column_idx: usize) -> Option<Sbbf> {
        self.single_calls += 1;
        self.filters.get(&(row_group_idx, column_idx)).cloned()
    }

    async fn bloom_filters_batch<'a>(
        &'a mut self,
        requests: &'a [(usize, usize)],
    ) -> HashMap<(usize, usize), Sbbf> {
        self.batch_calls += 1;

        // Custom batch implementation
        requests
            .iter()
            .filter_map(|&key| self.filters.get(&key).map(|f| (key, f.clone())))
            .collect()
    }
}

// ============================================================================
// Tests
// ============================================================================

#[tokio::test]
async fn test_mock_provider_tracks_calls() {
    let schema = Schema::new(vec![Field::new("id", DataType::Int64, false)]);
    let batch = make_int_batch(&schema, &[1, 2]);

    let props = WriterProperties::builder()
        .set_statistics_enabled(EnabledStatistics::Chunk)
        .build();

    let bytes = write_parquet_async(&[batch], props).await;
    let metadata = load_metadata(&bytes);

    let mut provider = MockBloomProvider::new();
    let expr = col("id").eq(lit(150i64));

    // Call without bloom filters enabled - provider should not be called
    let _result = PruneRequest::new(&metadata, &schema)
        .with_predicate(&expr)
        .enable_page_index(false)
        .enable_bloom_filter(false)
        .emit_roaring(false)
        .prune_async(&mut provider)
        .await;

    assert_eq!(
        provider.calls().len(),
        0,
        "Provider should not be called when bloom filters disabled"
    );
}

#[tokio::test]
async fn test_mock_provider_called_with_bloom_enabled() {
    let schema = Schema::new(vec![Field::new("id", DataType::Int64, false)]);
    let batch = make_int_batch(&schema, &[1, 2]);

    let props = WriterProperties::builder()
        .set_statistics_enabled(EnabledStatistics::Chunk)
        .set_bloom_filter_enabled(true)
        .build();

    let bytes = write_parquet_async(&[batch], props).await;
    let metadata = load_metadata(&bytes);

    let mut provider = MockBloomProvider::new();
    let expr = col("id").eq(lit(150i64));

    let _result = PruneRequest::new(&metadata, &schema)
        .with_predicate(&expr)
        .enable_page_index(false)
        .enable_bloom_filter(true) // Enable bloom filter pruning
        .emit_roaring(false)
        .prune_async(&mut provider)
        .await;

    // Should have attempted to load bloom filter for column 0, row group 0
    assert!(
        provider.calls().contains(&(0, 0)),
        "Provider should be called for row group 0, column 0"
    );
}

#[tokio::test]
async fn test_default_batch_implementation_is_sequential() {
    let mut provider = MockBloomProvider::new();

    // Simulate batch call with default implementation
    let requests = vec![(0, 0), (0, 1), (1, 0)];
    let _results = provider.bloom_filters_batch(&requests).await;

    // Default implementation calls bloom_filter sequentially for each request
    assert_eq!(provider.calls().len(), 3);
    assert_eq!(provider.calls(), &[(0, 0), (0, 1), (1, 0)]);
}

#[tokio::test]
async fn test_cached_provider_avoids_duplicate_loads() {
    let mut provider = CachedBloomProvider::new();

    // First call - no filter available, not cached
    let result1 = provider.bloom_filter(0, 0).await;
    assert!(result1.is_none());
    assert_eq!(provider.cache_hits(), 0);
    assert_eq!(provider.calls.len(), 1);

    // Second call to same key - still tries to load (None not cached)
    let result2 = provider.bloom_filter(0, 0).await;
    assert!(result2.is_none());
    assert_eq!(provider.cache_hits(), 0); // Still 0 - None results aren't cached
    assert_eq!(provider.calls.len(), 2);

    // Note: Cache only stores successful filter loads (Some values).
    // This is correct behavior - we don't cache "not found" results.
}

#[tokio::test]
async fn test_batched_provider_uses_custom_batch_method() {
    let mut provider = BatchedBloomProvider::new();

    let requests = vec![(0, 0), (0, 1), (1, 0)];
    let _results = provider.bloom_filters_batch(&requests).await;

    // Custom batch method should be called once, not individual calls
    assert_eq!(provider.batch_calls(), 1);
    assert_eq!(provider.single_calls(), 0);
}

#[tokio::test]
async fn test_provider_returns_none_for_missing_filters() {
    let mut provider = MockBloomProvider::new();

    let result = provider.bloom_filter(0, 0).await;
    assert!(result.is_none());

    let result = provider.bloom_filter(999, 999).await;
    assert!(result.is_none());
}

#[tokio::test]
async fn test_batch_returns_only_available_filters() {
    // Note: We can't easily create real Sbbf instances in tests without
    // actual Parquet files with bloom filters, so this test demonstrates
    // the structure even though filters will be None

    let mut provider = MockBloomProvider::new();

    let requests = vec![(0, 0), (0, 1), (1, 0)];
    let results = provider.bloom_filters_batch(&requests).await;

    // Should return empty map (no filters available)
    assert_eq!(results.len(), 0);

    // But all requests should have been attempted
    assert_eq!(provider.calls().len(), 3);
}

#[tokio::test]
async fn test_provider_called_for_multiple_row_groups() {
    let schema = Schema::new(vec![Field::new("id", DataType::Int64, false)]);
    let batch1 = make_int_batch(&schema, &[1, 2]);
    let batch2 = make_int_batch(&schema, &[3, 4]);
    let batch3 = make_int_batch(&schema, &[5, 6]);

    let props = WriterProperties::builder()
        .set_statistics_enabled(EnabledStatistics::Chunk)
        .set_max_row_group_size(2)
        .set_bloom_filter_enabled(true)
        .build();

    let bytes = write_parquet_async(&[batch1, batch2, batch3], props).await;
    let metadata = load_metadata(&bytes);

    let mut provider = MockBloomProvider::new();
    let expr = col("id").eq(lit(150i64));

    let _result = PruneRequest::new(&metadata, &schema)
        .with_predicate(&expr)
        .enable_page_index(false)
        .enable_bloom_filter(true)
        .emit_roaring(false)
        .prune_async(&mut provider)
        .await;

    // Should attempt to load bloom filter for all row groups
    assert!(provider.calls().contains(&(0, 0)));
    assert!(provider.calls().contains(&(1, 0)));
    assert!(provider.calls().contains(&(2, 0)));
}

#[tokio::test]
async fn test_cached_provider_with_multiple_columns() {
    let mut provider = CachedBloomProvider::new();

    // Load filters for different columns (all return None, so no caching occurs)
    provider.bloom_filter(0, 0).await;
    provider.bloom_filter(0, 1).await;
    provider.bloom_filter(0, 0).await; // No cache hit - None not cached
    provider.bloom_filter(0, 1).await; // No cache hit - None not cached

    assert_eq!(provider.cache_hits(), 0); // None results don't get cached
    assert_eq!(provider.calls.len(), 4);
}

#[tokio::test]
async fn test_batch_method_with_empty_requests() {
    let mut provider = MockBloomProvider::new();

    let requests: Vec<(usize, usize)> = vec![];
    let results = provider.bloom_filters_batch(&requests).await;

    assert_eq!(results.len(), 0);
    assert_eq!(provider.calls().len(), 0);
}

#[tokio::test]
async fn test_batch_method_with_duplicate_requests() {
    let mut provider = MockBloomProvider::new();

    // Same request multiple times
    let requests = vec![(0, 0), (0, 0), (0, 0)];
    let results = provider.bloom_filters_batch(&requests).await;

    // Default implementation will call bloom_filter 3 times
    assert_eq!(provider.calls().len(), 3);
    assert_eq!(results.len(), 0); // No filters available
}