Skip to main content

ailake_file/
reader.rs

1// SPDX-License-Identifier: MIT OR Apache-2.0
2use crate::footer::Precision;
3use ailake_core::{AilakeError, AilakeResult, Centroid, VectorMetric};
4use ailake_index::{AnyIndex, HnswIndex, IvfPqSerializer, MmapLoader};
5use ailake_parquet::ParquetVectorReader;
6use arrow_array::RecordBatch;
7use bytes::Bytes;
8
9use crate::footer::{
10    parquet_footer_start, AilakeHeader, AilakeTrailer, DistanceMetric, AILK_FTS_HEADER_SIZE,
11    AILK_FTS_MAGIC, FLAG_INDEX_IVF_PQ, HEADER_SIZE, KV_FTS_OFFSET, TRAILER_SIZE,
12};
13
14pub struct AilakeFileReader {
15    bytes: Bytes,
16    vector_column: String,
17    #[allow(dead_code)]
18    dim: u32,
19}
20
21impl AilakeFileReader {
22    pub fn new(bytes: Bytes, vector_column: &str, dim: u32) -> Self {
23        Self {
24            bytes,
25            vector_column: vector_column.to_string(),
26            dim,
27        }
28    }
29
30    /// Returns the absolute byte offset of the primary AILK section.
31    ///
32    /// Tries `ailake.footer_offset` from Parquet KV metadata first (files written
33    /// by `write()` / `write_multi()`). Falls back to `AilakeTrailer` bootstrap for
34    /// files produced by `write_single_pass()` / `write_multi_single_pass()`.
35    pub fn ailk_offset(&self) -> AilakeResult<u64> {
36        let reader = ParquetVectorReader::new(self.bytes.clone(), &self.vector_column);
37        if let Some(val) = reader.kv_metadata("ailake.footer_offset")? {
38            return val.parse::<u64>().map_err(|_| AilakeError::NotAnAilakeFile);
39        }
40        self.ailk_offset_from_trailer()
41    }
42
43    /// Returns the absolute byte offset of the AILK section for a named vector column.
44    ///
45    /// Resolution order:
46    ///   1. `ailake.{column}.footer_offset` KV (extra columns in multi-column files)
47    ///   2. `ailake.footer_offset` KV (primary column or single-column files)
48    ///   3. `AilakeTrailer` scan (streaming / single-pass files without KV injection)
49    pub fn ailk_offset_for_column(&self, column: &str) -> AilakeResult<u64> {
50        let reader = ParquetVectorReader::new(self.bytes.clone(), column);
51        let col_key = format!("ailake.{column}.footer_offset");
52        if let Some(val) = reader.kv_metadata(&col_key)? {
53            return val.parse::<u64>().map_err(|_| AilakeError::NotAnAilakeFile);
54        }
55        if let Some(val) = reader.kv_metadata("ailake.footer_offset")? {
56            return val.parse::<u64>().map_err(|_| AilakeError::NotAnAilakeFile);
57        }
58        self.ailk_offset_from_trailer()
59    }
60
61    /// Bootstrap AILK offset from the `AilakeTrailer` embedded just before the Parquet footer.
62    ///
63    /// The `AilakeTrailer` (24 bytes) ends immediately before the Parquet footer thrift
64    /// in every AI-Lake file. On S3, the trailer bytes are already present in the initial
65    /// footer range-GET (same GET that fetches the Parquet footer), so this bootstrap path
66    /// costs no additional I/O compared to the KV path.
67    fn ailk_offset_from_trailer(&self) -> AilakeResult<u64> {
68        let buf = self.bytes.as_ref();
69        let footer_start = parquet_footer_start(buf)?;
70        if footer_start < TRAILER_SIZE {
71            return Err(AilakeError::NotAnAilakeFile);
72        }
73        let trailer_start = footer_start - TRAILER_SIZE;
74        let trailer_bytes: &[u8; TRAILER_SIZE] = buf[trailer_start..footer_start]
75            .try_into()
76            .map_err(|_| AilakeError::NotAnAilakeFile)?;
77        let trailer = AilakeTrailer::from_bytes(trailer_bytes)?;
78        Ok(trailer.footer_offset)
79    }
80
81    /// Returns true if the file contains an embedded AILK section.
82    pub fn is_ailake_file(&self) -> bool {
83        self.ailk_offset().is_ok()
84    }
85
86    /// Parse the 64-byte AI-Lake header from the embedded AILK section.
87    pub fn read_header(&self) -> AilakeResult<AilakeHeader> {
88        self.read_header_at_offset(self.ailk_offset()?)
89    }
90
91    /// Parse the 64-byte AI-Lake header for a named vector column.
92    ///
93    /// Uses `ailake.{column}.footer_offset` for extra columns and falls back to
94    /// `ailake.footer_offset` for the primary column (single-column files).
95    pub fn read_header_for_column(&self, column: &str) -> AilakeResult<AilakeHeader> {
96        self.read_header_at_offset(self.ailk_offset_for_column(column)?)
97    }
98
99    fn read_header_at_offset(&self, offset: u64) -> AilakeResult<AilakeHeader> {
100        let offset = offset as usize;
101        if offset + HEADER_SIZE > self.bytes.len() {
102            return Err(AilakeError::NotAnAilakeFile);
103        }
104        let header_bytes: &[u8; HEADER_SIZE] = self.bytes[offset..offset + HEADER_SIZE]
105            .try_into()
106            .map_err(|_| AilakeError::NotAnAilakeFile)?;
107        AilakeHeader::from_bytes(header_bytes)
108    }
109
110    /// Read centroid + radius from the AILK section.
111    pub fn get_centroid(&self) -> AilakeResult<Centroid> {
112        let ailk_start = self.ailk_offset()? as usize;
113        let header = self.read_header()?;
114        let centroid_start = ailk_start + header.centroid_offset as usize;
115        let centroid_end = centroid_start + header.centroid_len as usize;
116
117        if centroid_end > self.bytes.len() {
118            return Err(AilakeError::NotAnAilakeFile);
119        }
120
121        let centroid_data = &self.bytes[centroid_start..centroid_end];
122        let dim = header.dim as usize;
123        let expected_len = dim.checked_mul(4).and_then(|v| v.checked_add(4)).ok_or(
124            AilakeError::InvalidCentroidLength {
125                expected_dim: header.dim,
126                actual: centroid_data.len(),
127            },
128        )?;
129        if centroid_data.len() != expected_len {
130            return Err(AilakeError::InvalidCentroidLength {
131                expected_dim: header.dim,
132                actual: centroid_data.len(),
133            });
134        }
135
136        let values: Vec<f32> = centroid_data[..dim * 4]
137            .chunks_exact(4)
138            .map(|b| f32::from_le_bytes(b.try_into().unwrap()))
139            .collect();
140        let radius = f32::from_le_bytes(centroid_data[dim * 4..].try_into().unwrap());
141        let metric = distance_metric_to_vector_metric(header.distance_metric);
142
143        Ok(Centroid {
144            values,
145            radius,
146            metric,
147        })
148    }
149
150    /// Load the HNSW index from the primary AILK section.
151    pub fn load_index(&self) -> AilakeResult<HnswIndex> {
152        self.load_index_for_column(&self.vector_column.clone())
153    }
154
155    /// Load the HNSW index for a specific vector column.
156    ///
157    /// Works for both single-column files (falls back to primary AILK) and
158    /// multi-column files written with `AilakeFileWriter::write_multi`.
159    pub fn load_index_for_column(&self, column: &str) -> AilakeResult<HnswIndex> {
160        let ailk_start = self.ailk_offset_for_column(column)? as usize;
161
162        if ailk_start + HEADER_SIZE > self.bytes.len() {
163            return Err(AilakeError::NotAnAilakeFile);
164        }
165        let header_bytes: &[u8; HEADER_SIZE] = self.bytes[ailk_start..ailk_start + HEADER_SIZE]
166            .try_into()
167            .map_err(|_| AilakeError::NotAnAilakeFile)?;
168        let header = AilakeHeader::from_bytes(header_bytes)?;
169
170        let hnsw_start = ailk_start + header.hnsw_offset as usize;
171        let hnsw_end = hnsw_start + header.hnsw_len as usize;
172
173        if hnsw_end > self.bytes.len() {
174            return Err(AilakeError::NotAnAilakeFile);
175        }
176        let mut idx = MmapLoader::from_bytes(&self.bytes[hnsw_start..hnsw_end])?;
177        if header.precision == Precision::F16 {
178            idx.quantize_to_f16();
179        }
180        Ok(idx)
181    }
182
183    /// Load primary index as `AnyIndex`, dispatching on header flags.
184    pub fn load_any_index(&self) -> AilakeResult<AnyIndex> {
185        self.load_any_index_for_column(&self.vector_column.clone())
186    }
187
188    /// Load index for a specific vector column as `AnyIndex`.
189    pub fn load_any_index_for_column(&self, column: &str) -> AilakeResult<AnyIndex> {
190        let ailk_start = self.ailk_offset_for_column(column)? as usize;
191
192        if ailk_start + HEADER_SIZE > self.bytes.len() {
193            return Err(AilakeError::NotAnAilakeFile);
194        }
195        let header_bytes: &[u8; HEADER_SIZE] = self.bytes[ailk_start..ailk_start + HEADER_SIZE]
196            .try_into()
197            .map_err(|_| AilakeError::NotAnAilakeFile)?;
198        let header = AilakeHeader::from_bytes(header_bytes)?;
199
200        let index_start = ailk_start + header.hnsw_offset as usize;
201        let index_end = index_start + header.hnsw_len as usize;
202
203        if index_end > self.bytes.len() {
204            return Err(AilakeError::NotAnAilakeFile);
205        }
206        let index_bytes = &self.bytes[index_start..index_end];
207
208        if header.flags & FLAG_INDEX_IVF_PQ != 0 {
209            let idx = IvfPqSerializer::from_bytes(index_bytes)?;
210            Ok(AnyIndex::IvfPq(idx))
211        } else {
212            let mut idx = MmapLoader::from_bytes(index_bytes)?;
213            if header.precision == Precision::F16 {
214                idx.quantize_to_f16();
215            }
216            Ok(AnyIndex::Hnsw(idx))
217        }
218    }
219
220    /// Read the Parquet section (tabular data + decoded embeddings).
221    /// The full file is valid Parquet; the AILK section is invisible to standard readers.
222    pub fn read_parquet(&self) -> AilakeResult<(RecordBatch, Vec<Vec<f32>>)> {
223        let reader = ParquetVectorReader::new(self.bytes.clone(), &self.vector_column);
224        reader.read_all()
225    }
226
227    /// Load the raw FTS blob from the AILK_FTS section, if present.
228    ///
229    /// Returns `Ok(None)` when the file has no FTS section (opt-in feature).
230    /// The returned bytes can be passed directly to `ailake_fts::FtsSearcher::from_blob`.
231    pub fn load_fts_blob(&self) -> AilakeResult<Option<Bytes>> {
232        let pq_reader = ParquetVectorReader::new(self.bytes.clone(), &self.vector_column);
233        let fts_offset_str = match pq_reader.kv_metadata(KV_FTS_OFFSET)? {
234            Some(s) => s,
235            None => return Ok(None),
236        };
237        let fts_abs: usize = fts_offset_str
238            .parse::<u64>()
239            .map_err(|e| {
240                AilakeError::Fts(format!(
241                    "invalid FTS section offset '{fts_offset_str}': {e}"
242                ))
243            })?
244            .try_into()
245            .map_err(|_| AilakeError::Fts("FTS section offset exceeds address space".into()))?;
246
247        // Read AILK_FTS header: magic(4) | version(2) | reserved(2) | blob_len(8)
248        if fts_abs + AILK_FTS_HEADER_SIZE > self.bytes.len() {
249            return Err(AilakeError::Fts("AILK_FTS header out of bounds".into()));
250        }
251        let hdr = &self.bytes[fts_abs..fts_abs + AILK_FTS_HEADER_SIZE];
252        if hdr[0..4] != AILK_FTS_MAGIC {
253            return Err(AilakeError::Fts(format!(
254                "bad AILK_FTS magic: {:?}",
255                &hdr[0..4]
256            )));
257        }
258        let blob_len = u64::from_le_bytes(hdr[8..16].try_into().unwrap()) as usize;
259        let blob_start = fts_abs + AILK_FTS_HEADER_SIZE;
260        let blob_end = blob_start + blob_len;
261        if blob_end > self.bytes.len() {
262            return Err(AilakeError::Fts("AILK_FTS blob out of bounds".into()));
263        }
264        Ok(Some(self.bytes.slice(blob_start..blob_end)))
265    }
266
267    /// Verify the positional invariant: Parquet record_count == HNSW node_count.
268    pub fn verify_integrity(&self) -> AilakeResult<()> {
269        let header = self.read_header()?;
270        let index = self.load_index()?;
271        let reader = ParquetVectorReader::new(self.bytes.clone(), &self.vector_column);
272        let parquet_count = reader.record_count()?;
273
274        if parquet_count != index.node_count() {
275            return Err(AilakeError::RowCountMismatch {
276                parquet: parquet_count,
277                hnsw: index.node_count(),
278            });
279        }
280        if parquet_count != header.record_count {
281            return Err(AilakeError::RowCountMismatch {
282                parquet: parquet_count,
283                hnsw: header.record_count,
284            });
285        }
286        Ok(())
287    }
288}
289
290fn distance_metric_to_vector_metric(dm: DistanceMetric) -> VectorMetric {
291    match dm {
292        DistanceMetric::Cosine => VectorMetric::Cosine,
293        DistanceMetric::Euclidean => VectorMetric::Euclidean,
294        DistanceMetric::DotProduct => VectorMetric::DotProduct,
295        DistanceMetric::NormalizedCosine => VectorMetric::NormalizedCosine,
296    }
297}
298
299#[cfg(test)]
300mod tests {
301    use super::*;
302    use crate::writer::AilakeFileWriter;
303    use ailake_core::{VectorMetric, VectorPrecision, VectorStoragePolicy};
304    use arrow_array::{Int32Array, RecordBatch};
305    use arrow_schema::{DataType, Field, Schema};
306    use std::sync::Arc;
307
308    fn make_policy(dim: u32) -> VectorStoragePolicy {
309        VectorStoragePolicy {
310            column_name: "embedding".to_string(),
311            dim,
312            metric: VectorMetric::Cosine,
313            precision: VectorPrecision::F16,
314            pq: None,
315            keep_raw_for_reranking: true,
316            pre_normalize: false,
317            hnsw_m: None,
318            hnsw_ef_construction: None,
319            ivf_residual: false,
320            embedding_model: None,
321            modality: None,
322            partition_by: None,
323            partition_value: None,
324            partition_column_type: None,
325            partition_fields: vec![],
326        }
327    }
328
329    fn write_file(rows: usize, dim: u32) -> Bytes {
330        let schema = Arc::new(Schema::new(vec![Field::new("id", DataType::Int32, false)]));
331        let ids: Vec<i32> = (0..rows as i32).collect();
332        let batch = RecordBatch::try_new(schema, vec![Arc::new(Int32Array::from(ids))]).unwrap();
333        let embs: Vec<Vec<f32>> = (0..rows)
334            .map(|i| {
335                let mut v = vec![0.0f32; dim as usize];
336                v[i % dim as usize] = 1.0;
337                v
338            })
339            .collect();
340        AilakeFileWriter::new(make_policy(dim))
341            .write(&batch, &embs)
342            .unwrap()
343    }
344
345    #[test]
346    fn is_ailake_file() {
347        let file = write_file(3, 4);
348        let reader = AilakeFileReader::new(file, "embedding", 4);
349        assert!(reader.is_ailake_file());
350    }
351
352    #[test]
353    fn integrity_check_passes() {
354        let file = write_file(10, 8);
355        let reader = AilakeFileReader::new(file, "embedding", 8);
356        reader.verify_integrity().unwrap();
357    }
358
359    #[test]
360    fn centroid_has_correct_dim() {
361        let file = write_file(5, 4);
362        let reader = AilakeFileReader::new(file, "embedding", 4);
363        let centroid = reader.get_centroid().unwrap();
364        assert_eq!(centroid.values.len(), 4);
365    }
366
367    #[test]
368    fn search_finds_nearest() {
369        let dim = 4u32;
370        let file = write_file(4, dim);
371        let reader = AilakeFileReader::new(file, "embedding", dim);
372        let index = reader.load_index().unwrap();
373        let query = vec![1.0f32, 0.0, 0.0, 0.0];
374        let results = index.search(&query, 1, 50);
375        assert_eq!(results.len(), 1);
376        assert_eq!(results[0].0, ailake_core::RowId::new(0));
377    }
378
379    #[test]
380    fn parquet_read_returns_tabular_data() {
381        let file = write_file(3, 4);
382        let reader = AilakeFileReader::new(file, "embedding", 4);
383        let (batch, embs) = reader.read_parquet().unwrap();
384        assert_eq!(batch.num_rows(), 3);
385        assert_eq!(embs.len(), 3);
386    }
387}