frankensearch-storage 0.2.0

FrankenSQLite-backed metadata and embedding job storage for frankensearch
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
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
424
425
426
427
428
429
430
431
432
433
434
435
436
437
438
439
440
441
442
443
444
445
446
447
448
449
450
451
452
453
454
455
456
457
458
459
460
461
462
463
464
465
466
467
468
469
470
471
472
473
474
475
476
477
478
479
480
481
482
483
484
485
486
487
488
489
490
491
492
493
494
495
496
497
498
499
500
501
502
503
504
505
506
507
508
509
510
511
512
513
514
515
516
517
518
519
520
521
522
523
524
525
526
527
528
529
530
531
532
533
534
535
536
537
538
539
540
541
542
543
544
545
546
547
548
549
550
551
552
553
554
555
556
557
558
559
560
561
562
563
564
565
566
567
568
569
570
571
572
573
574
575
576
577
578
579
580
581
582
583
584
585
586
587
588
589
590
591
592
593
594
595
596
597
598
599
600
601
602
603
604
605
606
607
608
609
610
611
612
613
614
615
616
617
618
619
620
621
622
623
624
625
626
627
628
629
630
631
632
633
634
635
636
637
638
639
640
641
642
643
644
645
646
647
648
649
650
651
652
653
654
655
656
657
658
659
660
661
662
663
664
665
666
667
668
669
670
671
672
673
674
675
676
677
678
679
680
681
682
683
684
685
686
687
688
689
690
691
692
693
694
695
696
697
698
699
700
701
702
703
704
705
706
707
708
709
710
711
712
713
714
715
716
717
718
719
720
721
722
723
724
725
726
727
728
729
730
731
732
733
734
735
736
737
738
739
740
741
742
743
744
745
746
747
748
749
750
751
752
753
754
755
756
757
758
759
760
761
762
763
764
765
766
767
768
769
770
771
772
773
774
775
776
777
778
779
780
781
782
783
784
785
786
787
788
789
790
791
792
793
794
795
796
797
798
799
800
801
802
803
804
805
806
807
808
809
810
811
812
813
814
815
816
817
818
819
820
821
822
823
824
825
826
827
828
829
830
831
832
833
834
835
836
837
838
839
840
841
842
843
844
845
846
847
848
849
850
851
852
853
854
855
856
857
858
859
860
861
862
863
864
865
866
867
868
869
870
871
872
873
874
875
876
877
878
879
880
881
882
883
884
885
886
887
888
889
890
891
892
893
894
895
896
897
898
899
900
901
902
903
use std::panic::{AssertUnwindSafe, catch_unwind, resume_unwind};
use std::path::PathBuf;

use frankensearch_core::{SearchError, SearchResult};
use fsqlite::Connection;
use serde::{Deserialize, Serialize};

use crate::metrics::{StorageMetrics, StorageMetricsSnapshot};
use crate::schema;

#[derive(Debug, Clone, Serialize, Deserialize, PartialEq, Eq)]
pub struct StorageConfig {
    pub db_path: PathBuf,
    pub wal_mode: bool,
    pub busy_timeout_ms: u64,
    pub raptorq_repair_symbols: u32,
    pub cache_size_pages: i32,
    pub concurrent_transactions: bool,
}

impl StorageConfig {
    #[must_use]
    pub fn in_memory() -> Self {
        Self {
            db_path: PathBuf::from(":memory:"),
            ..Self::default()
        }
    }

    fn begin_sql(&self) -> &'static str {
        if self.concurrent_transactions {
            "BEGIN CONCURRENT;"
        } else {
            "BEGIN;"
        }
    }
}

impl Default for StorageConfig {
    fn default() -> Self {
        Self {
            db_path: PathBuf::from("storage.sqlite3"),
            wal_mode: true,
            busy_timeout_ms: 5_000,
            // Disabled by default until upstream frankensqlite reserve-byte
            // balancing issues are resolved.
            raptorq_repair_symbols: 0,
            cache_size_pages: 2_000,
            concurrent_transactions: true,
        }
    }
}

pub struct Storage {
    conn: Connection,
    config: StorageConfig,
    metrics: StorageMetrics,
}

static FILE_BOOTSTRAP_LOCK: std::sync::Mutex<()> = std::sync::Mutex::new(());

impl std::fmt::Debug for Storage {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.debug_struct("Storage")
            .field("path", &self.config.db_path)
            .field("wal_mode", &self.config.wal_mode)
            .field("busy_timeout_ms", &self.config.busy_timeout_ms)
            .finish_non_exhaustive()
    }
}

impl Storage {
    pub fn open(config: StorageConfig) -> SearchResult<Self> {
        tracing::debug!(
            target: "frankensearch.storage",
            path = %config.db_path.display(),
            wal_mode = config.wal_mode,
            busy_timeout_ms = config.busy_timeout_ms,
            cache_size_pages = config.cache_size_pages,
            concurrent_transactions = config.concurrent_transactions,
            "opening storage connection"
        );

        let file_bootstrap_guard = if config.db_path.as_os_str() == ":memory:" {
            None
        } else {
            Some(
                FILE_BOOTSTRAP_LOCK
                    .lock()
                    .map_err(|_| SearchError::SubsystemError {
                        subsystem: "storage",
                        source: Box::new(std::io::Error::other("file bootstrap lock poisoned")),
                    })?,
            )
        };

        let path = config.db_path.to_string_lossy().to_string();
        let conn = Connection::open(path).map_err(map_storage_error)?;

        let storage = Self {
            conn,
            config,
            metrics: StorageMetrics::default(),
        };

        storage.metrics.record_open();
        storage.apply_pragmas()?;
        schema::bootstrap(storage.connection())?;
        storage.metrics.record_schema_bootstrap();
        drop(file_bootstrap_guard);

        if let Ok(version) = schema::current_version(storage.connection()) {
            tracing::debug!(
                target: "frankensearch.storage",
                schema_version = version,
                "storage bootstrap complete"
            );
        }

        Ok(storage)
    }

    pub fn open_in_memory() -> SearchResult<Self> {
        Self::open(StorageConfig::in_memory())
    }

    #[must_use]
    pub fn connection(&self) -> &Connection {
        &self.conn
    }

    #[must_use]
    pub fn config(&self) -> &StorageConfig {
        &self.config
    }

    #[must_use]
    pub fn metrics_snapshot(&self) -> StorageMetricsSnapshot {
        self.metrics.snapshot()
    }

    pub fn transaction<F, T>(&self, f: F) -> SearchResult<T>
    where
        F: FnOnce(&Connection) -> SearchResult<T>,
    {
        self.transaction_with_mode(self.config.begin_sql(), f)
    }

    /// Run a closure inside a `BEGIN IMMEDIATE` transaction.
    ///
    /// Unlike [`Storage::transaction`], this acquires a write lock immediately,
    /// preventing concurrent writers from interleaving reads and writes.
    /// Use this when correctness depends on serialized read-then-write
    /// (e.g. `claim_batch`).
    pub fn immediate_transaction<F, T>(&self, f: F) -> SearchResult<T>
    where
        F: FnOnce(&Connection) -> SearchResult<T>,
    {
        self.transaction_with_mode("BEGIN IMMEDIATE;", f)
    }

    fn transaction_with_mode<F, T>(&self, begin_sql: &str, f: F) -> SearchResult<T>
    where
        F: FnOnce(&Connection) -> SearchResult<T>,
    {
        tracing::trace!(
            target: "frankensearch.storage",
            begin_sql,
            "starting storage transaction"
        );

        self.conn.execute(begin_sql).map_err(map_storage_error)?;

        let outcome = catch_unwind(AssertUnwindSafe(|| f(&self.conn)));

        match outcome {
            Ok(Ok(value)) => {
                self.conn.execute("COMMIT;").map_err(|commit_err| {
                    if let Err(rollback_err) = self.conn.execute("ROLLBACK;") {
                        tracing::warn!(
                            target: "frankensearch.storage",
                            error = %rollback_err,
                            "rollback failed after commit error"
                        );
                    }
                    map_storage_error(commit_err)
                })?;
                self.metrics.record_commit();
                tracing::trace!(target: "frankensearch.storage", "storage transaction committed");
                Ok(value)
            }
            Ok(Err(err)) => {
                if let Err(rollback_err) = self.conn.execute("ROLLBACK;") {
                    tracing::warn!(
                        target: "frankensearch.storage",
                        error = %rollback_err,
                        "rollback failed after closure error"
                    );
                }
                self.metrics.record_rollback();
                tracing::debug!(
                    target: "frankensearch.storage",
                    ?err,
                    "storage transaction rolled back due to closure error"
                );
                Err(err)
            }
            Err(payload) => {
                if let Err(rollback_err) = self.conn.execute("ROLLBACK;") {
                    tracing::error!(
                        target: "frankensearch.storage",
                        error = %rollback_err,
                        "critical: rollback failed during panic recovery"
                    );
                }
                self.metrics.record_rollback();
                tracing::error!(
                    target: "frankensearch.storage",
                    "storage transaction rolled back after panic"
                );
                resume_unwind(payload);
            }
        }
    }

    fn apply_pragmas(&self) -> SearchResult<()> {
        tracing::trace!(
            target: "frankensearch.storage",
            wal_mode = self.config.wal_mode,
            busy_timeout_ms = self.config.busy_timeout_ms,
            cache_size_pages = self.config.cache_size_pages,
            raptorq_repair_symbols = self.config.raptorq_repair_symbols,
            "applying storage pragmas"
        );

        self.conn
            .execute("PRAGMA foreign_keys=ON;")
            .map_err(map_storage_error)?;

        if self.config.wal_mode {
            self.conn
                .execute("PRAGMA journal_mode=WAL;")
                .map_err(map_storage_error)?;
        } else if let Err(error) = self.conn.execute("PRAGMA journal_mode=DELETE;") {
            tracing::warn!(
                target: "frankensearch.storage",
                ?error,
                "journal_mode=DELETE was not accepted by backend; falling back to WAL"
            );
            self.conn
                .execute("PRAGMA journal_mode=WAL;")
                .map_err(map_storage_error)?;
        }

        self.conn
            .execute(&format!(
                "PRAGMA busy_timeout={};",
                self.config.busy_timeout_ms
            ))
            .map_err(map_storage_error)?;

        self.conn
            .execute(&format!(
                "PRAGMA cache_size={};",
                self.config.cache_size_pages
            ))
            .map_err(map_storage_error)?;

        if self.config.raptorq_repair_symbols > 0 {
            tracing::warn!(
                target: "frankensearch.storage",
                symbols = self.config.raptorq_repair_symbols,
                "raptorq FEC repair symbols requested but currently disabled: \
                 upstream frankensqlite reserve-byte balancing is unresolved; \
                 configure raptorq_repair_symbols=0 to suppress this warning"
            );
        }

        Ok(())
    }
}

pub(crate) fn map_storage_error<E>(source: E) -> SearchError
where
    E: std::error::Error + Send + Sync + 'static,
{
    SearchError::SubsystemError {
        subsystem: "storage",
        source: Box::new(source),
    }
}

#[cfg(test)]
mod tests {
    use std::panic::{self, AssertUnwindSafe};
    use std::path::PathBuf;
    use std::process;
    use std::sync::{Arc, Barrier, mpsc};
    use std::thread;
    use std::time::{SystemTime, UNIX_EPOCH};

    use frankensearch_core::{SearchError, SearchResult};
    use fsqlite_types::value::SqliteValue;
    use serde_json::json;

    use crate::document::{DocumentRecord, count_documents, upsert_document};
    use crate::schema::{self, SCHEMA_VERSION};

    use super::{Storage, StorageConfig};

    struct TempDbPath {
        path: PathBuf,
    }

    impl TempDbPath {
        fn new(tag: &str) -> Self {
            let nanos = SystemTime::now()
                .duration_since(UNIX_EPOCH)
                .expect("system clock should be after unix epoch")
                .as_nanos();
            let path = std::env::temp_dir().join(format!(
                "frankensearch-storage-{tag}-{}-{nanos}.sqlite3",
                process::id()
            ));
            Self { path }
        }

        fn config(&self) -> StorageConfig {
            StorageConfig {
                db_path: self.path.clone(),
                ..StorageConfig::default()
            }
        }
    }

    impl Drop for TempDbPath {
        fn drop(&mut self) {
            for suffix in ["", "-wal", "-shm"] {
                let candidate = if suffix.is_empty() {
                    self.path.clone()
                } else {
                    PathBuf::from(format!("{}{}", self.path.display(), suffix))
                };
                let _ = std::fs::remove_file(candidate);
            }
        }
    }

    fn sample_document(id: &str) -> DocumentRecord {
        let mut doc = DocumentRecord::new(
            id,
            "Rust ownership keeps heap usage explicit during indexing.",
            [0x2a; 32],
            128,
            1_739_499_200,
            1_739_499_200,
        );
        doc.source_path = Some("tests://fixture".to_owned());
        doc.metadata = Some(json!({"fixture": true}));
        doc
    }

    const CONCURRENT_OPEN_THREADS: usize = 4;
    const CONCURRENT_OPEN_STRESS_ROUNDS: usize = 24;

    fn run_concurrent_open_round(config: &StorageConfig) -> Vec<SearchResult<i64>> {
        let barrier = Arc::new(Barrier::new(CONCURRENT_OPEN_THREADS));
        let (tx, rx) = mpsc::channel::<SearchResult<i64>>();
        let mut handles = Vec::with_capacity(CONCURRENT_OPEN_THREADS);

        for _ in 0..CONCURRENT_OPEN_THREADS {
            let cfg = config.clone();
            let gate = Arc::clone(&barrier);
            let sender = tx.clone();
            handles.push(thread::spawn(move || {
                gate.wait();
                let open_result = Storage::open(cfg)
                    .and_then(|storage| schema::current_version(storage.connection()));
                sender
                    .send(open_result)
                    .expect("result send should succeed");
            }));
        }
        drop(tx);

        let results: Vec<SearchResult<i64>> = rx.into_iter().collect();
        for handle in handles {
            handle.join().expect("concurrent opener thread should join");
        }
        results
    }

    fn insert_document_minimal(conn: &fsqlite::Connection, id: &str) -> SearchResult<()> {
        let params = [
            SqliteValue::Text(id.to_owned().into()),
            SqliteValue::Text("fixture-content".to_owned().into()),
            SqliteValue::Blob(vec![0x33; 32].into()),
            SqliteValue::Integer(64),
            SqliteValue::Integer(1_739_499_200),
            SqliteValue::Integer(1_739_499_200),
        ];
        conn.execute_with_params(
            "INSERT INTO documents (\
                doc_id, content_preview, content_hash, content_length, created_at, updated_at\
             ) VALUES (?1, ?2, ?3, ?4, ?5, ?6);",
            &params,
        )
        .map_err(super::map_storage_error)?;
        Ok(())
    }

    fn pragma_i64(conn: &fsqlite::Connection, name: &str) -> SearchResult<i64> {
        let row = conn
            .query_row(&format!("PRAGMA {name};"))
            .map_err(super::map_storage_error)?;
        match row.get(0) {
            Some(SqliteValue::Integer(value)) => Ok(*value),
            Some(SqliteValue::Text(value)) => {
                value
                    .parse::<i64>()
                    .map_err(|parse_error| SearchError::SubsystemError {
                        subsystem: "storage",
                        source: Box::new(parse_error),
                    })
            }
            Some(other) => Err(SearchError::SubsystemError {
                subsystem: "storage",
                source: Box::new(std::io::Error::other(format!(
                    "unexpected pragma type for {name}: {:?}",
                    other
                ))),
            }),
            None => Err(SearchError::SubsystemError {
                subsystem: "storage",
                source: Box::new(std::io::Error::other(format!(
                    "missing pragma value for {name}"
                ))),
            }),
        }
    }

    fn pragma_text(conn: &fsqlite::Connection, name: &str) -> SearchResult<String> {
        let row = conn
            .query_row(&format!("PRAGMA {name};"))
            .map_err(super::map_storage_error)?;
        match row.get(0) {
            Some(SqliteValue::Text(value)) => Ok(value.to_string()),
            Some(SqliteValue::Integer(value)) => Ok(value.to_string()),
            Some(other) => Err(SearchError::SubsystemError {
                subsystem: "storage",
                source: Box::new(std::io::Error::other(format!(
                    "unexpected pragma type for {name}: {:?}",
                    other
                ))),
            }),
            None => Err(SearchError::SubsystemError {
                subsystem: "storage",
                source: Box::new(std::io::Error::other(format!(
                    "missing pragma value for {name}"
                ))),
            }),
        }
    }

    #[test]
    fn open_in_memory_bootstraps_schema() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");
        let version = schema::current_version(storage.connection()).expect("schema version row");
        assert_eq!(version, SCHEMA_VERSION);

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.opens, 1);
        assert_eq!(metrics.schema_bootstraps, 1);
    }

    #[test]
    fn open_applies_configured_pragmas_deterministically() {
        let tmp = TempDbPath::new("pragmas");
        let storage = Storage::open(StorageConfig {
            db_path: tmp.path.clone(),
            wal_mode: true,
            busy_timeout_ms: 1_234,
            cache_size_pages: 321,
            concurrent_transactions: false,
            ..StorageConfig::default()
        })
        .expect("storage should open with configured pragmas");

        assert_eq!(
            pragma_text(storage.connection(), "journal_mode")
                .expect("journal_mode pragma should be queryable")
                .to_ascii_lowercase(),
            "wal"
        );
        assert_eq!(
            pragma_i64(storage.connection(), "busy_timeout")
                .expect("busy_timeout pragma should be queryable"),
            1_234
        );
        assert_eq!(
            pragma_i64(storage.connection(), "cache_size")
                .expect("cache_size pragma should be queryable"),
            321
        );
    }

    #[test]
    fn schema_bootstrap_is_idempotent() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");
        schema::bootstrap(storage.connection()).expect("second bootstrap should succeed");
        schema::bootstrap(storage.connection()).expect("third bootstrap should succeed");

        let version = schema::current_version(storage.connection()).expect("schema version row");
        assert_eq!(version, SCHEMA_VERSION);
    }

    #[test]
    fn concurrent_open_initializes_schema_consistently() {
        let tmp = TempDbPath::new("concurrent-bootstrap");
        let config = tmp.config();
        let results = run_concurrent_open_round(&config);
        assert_eq!(results.len(), CONCURRENT_OPEN_THREADS);
        for (thread_idx, result) in results.into_iter().enumerate() {
            match result {
                Ok(version) => assert_eq!(
                    version, SCHEMA_VERSION,
                    "thread {thread_idx} should observe latest schema version"
                ),
                Err(error) => panic!("thread {thread_idx} failed concurrent open: {error}"),
            }
        }
    }

    #[test]
    fn concurrent_open_stays_stable_across_repeated_rounds() {
        let tmp = TempDbPath::new("concurrent-bootstrap-rounds");
        let config = tmp.config();

        for round in 0..CONCURRENT_OPEN_STRESS_ROUNDS {
            let results = run_concurrent_open_round(&config);
            assert_eq!(
                results.len(),
                CONCURRENT_OPEN_THREADS,
                "round {round} should return one result per thread"
            );
            for (thread_idx, result) in results.into_iter().enumerate() {
                match result {
                    Ok(version) => assert_eq!(
                        version, SCHEMA_VERSION,
                        "round {round}, thread {thread_idx} should observe latest schema version"
                    ),
                    Err(error) => {
                        panic!("round {round}, thread {thread_idx} failed concurrent open: {error}")
                    }
                }
            }
        }
    }

    #[test]
    fn transaction_rolls_back_on_error() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");
        let doc = sample_document("doc-error");

        let result: SearchResult<()> = storage.transaction(|conn| {
            upsert_document(conn, &doc)?;
            Err(SearchError::InvalidConfig {
                field: "test".to_owned(),
                value: "forced".to_owned(),
                reason: "force rollback".to_owned(),
            })
        });

        assert!(result.is_err(), "transaction should return original error");
        assert_eq!(
            count_documents(storage.connection()).expect("count should work"),
            0,
            "document insert should have been rolled back"
        );

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.tx_commits, 0);
        assert_eq!(metrics.tx_rollbacks, 1);
    }

    #[test]
    fn transaction_rolls_back_on_panic_and_connection_stays_usable() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");
        let doc = sample_document("doc-panic");

        let panic_result = panic::catch_unwind(AssertUnwindSafe(|| {
            let _: SearchResult<()> = storage.transaction(|conn| {
                upsert_document(conn, &doc).expect("insert should succeed before panic");
                panic!("forced panic");
            });
        }));

        assert!(panic_result.is_err(), "panic should propagate to caller");
        assert_eq!(
            count_documents(storage.connection()).expect("count should work"),
            0,
            "panic path should rollback transaction"
        );
        assert_eq!(
            schema::current_version(storage.connection()).expect("connection should remain usable"),
            SCHEMA_VERSION
        );

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.tx_rollbacks, 1);
        assert_eq!(metrics.tx_commits, 0);
    }

    #[test]
    fn commit_persists_after_reopen() {
        let tmp = TempDbPath::new("visibility");
        let config = tmp.config();

        let storage_a = Storage::open(config.clone()).expect("writer storage should open");
        let doc = sample_document("doc-visible");

        storage_a
            .transaction(|conn| {
                insert_document_minimal(conn, &doc.doc_id)?;
                assert_eq!(
                    count_documents(conn)?,
                    1,
                    "write should be visible inside writer transaction"
                );
                Ok(())
            })
            .expect("transaction should commit");

        drop(storage_a);

        let storage_c = Storage::open(config).expect("post-commit storage should open");
        assert_eq!(
            count_documents(storage_c.connection()).expect("count after commit"),
            1,
            "committed write should be visible to newly opened connection"
        );
    }

    #[test]
    fn rollback_is_not_persisted_after_reopen() {
        let tmp = TempDbPath::new("rollback");
        let config = tmp.config();

        let storage_a = Storage::open(config).expect("writer storage should open");
        let doc_id = "doc-rollback";

        let tx_result: SearchResult<()> = storage_a.transaction(|conn| {
            insert_document_minimal(conn, doc_id)?;
            Err(SearchError::InvalidConfig {
                field: "test".to_owned(),
                value: "rollback".to_owned(),
                reason: "forced rollback".to_owned(),
            })
        });
        assert!(tx_result.is_err(), "transaction should rollback");
        assert_eq!(
            count_documents(storage_a.connection()).expect("count after rollback"),
            0,
            "rolled back write should not persist in active connection state"
        );
    }

    #[test]
    fn dropped_connection_allows_reopen_and_write() {
        let tmp = TempDbPath::new("drop");
        let config = tmp.config();

        let storage_a = Storage::open(config.clone()).expect("first storage should open");
        drop(storage_a);

        let storage_b = Storage::open(config).expect("second storage should open");

        storage_b
            .transaction(|conn| {
                insert_document_minimal(conn, "doc-after-drop")?;
                Ok(())
            })
            .expect("second connection should remain fully usable after first drop");

        assert_eq!(
            count_documents(storage_b.connection()).expect("count after write"),
            1
        );
    }

    #[test]
    fn nested_transaction_error_rolls_back_outer_transaction() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");

        let result: SearchResult<()> = storage.transaction(|conn| {
            insert_document_minimal(conn, "doc-nested")?;
            storage.transaction(|_nested_conn| Ok(()))
        });

        assert!(
            result.is_err(),
            "nested transaction should fail and propagate error"
        );
        assert_eq!(
            count_documents(storage.connection()).expect("count after nested failure"),
            0,
            "outer transaction should rollback when nested begin fails"
        );

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.tx_rollbacks, 1);
        assert_eq!(metrics.tx_commits, 0);
    }

    #[test]
    fn uncommitted_writes_are_invisible_to_concurrent_reader() {
        let tmp = TempDbPath::new("snapshot");
        let config = tmp.config();
        let reader = Storage::open(config.clone()).expect("reader storage should open");
        let writer_config = config;

        let (inserted_tx, inserted_rx) = mpsc::channel::<()>();
        let (release_tx, release_rx) = mpsc::channel::<()>();

        let writer_thread = thread::spawn(move || {
            let writer = Storage::open(writer_config).expect("writer storage should open");
            writer
                .transaction(|conn| {
                    insert_document_minimal(conn, "doc-uncommitted")?;
                    inserted_tx
                        .send(())
                        .expect("writer should signal insert before commit");
                    release_rx
                        .recv()
                        .expect("writer should wait for reader inspection");
                    Ok(())
                })
                .expect("writer transaction should commit");
        });

        inserted_rx
            .recv()
            .expect("reader should observe insert signal");

        assert_eq!(
            count_documents(reader.connection()).expect("reader count during writer transaction"),
            0,
            "reader should not see uncommitted write from concurrent transaction"
        );

        release_tx
            .send(())
            .expect("reader should release writer to commit");
        writer_thread.join().expect("writer thread should join");

        // In WAL mode without an explicit read transaction, each statement
        // sees the latest committed state.  After the writer commits, the
        // reader observes the new row immediately.
        assert_eq!(
            count_documents(reader.connection()).expect("reader count after writer commit"),
            1,
            "reader without explicit transaction should see committed write"
        );
    }

    #[test]
    fn storage_config_serde_roundtrip() {
        let config = StorageConfig {
            db_path: PathBuf::from("/tmp/test.db"),
            wal_mode: false,
            busy_timeout_ms: 10_000,
            raptorq_repair_symbols: 4,
            cache_size_pages: 500,
            concurrent_transactions: false,
        };
        let json = serde_json::to_string(&config).expect("serialize");
        let deserialized: StorageConfig = serde_json::from_str(&json).expect("deserialize");
        assert_eq!(config, deserialized);
    }

    #[test]
    fn storage_config_in_memory_path() {
        let config = StorageConfig::in_memory();
        assert_eq!(config.db_path, PathBuf::from(":memory:"));
        assert!(config.wal_mode);
        assert!(config.concurrent_transactions);
    }

    #[test]
    fn storage_config_begin_sql_concurrent_vs_standard() {
        let concurrent = StorageConfig {
            concurrent_transactions: true,
            ..StorageConfig::default()
        };
        assert_eq!(concurrent.begin_sql(), "BEGIN CONCURRENT;");

        let standard = StorageConfig {
            concurrent_transactions: false,
            ..StorageConfig::default()
        };
        assert_eq!(standard.begin_sql(), "BEGIN;");
    }

    #[test]
    fn immediate_transaction_commits_successfully() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");

        storage
            .immediate_transaction(|conn| {
                insert_document_minimal(conn, "doc-immediate")?;
                Ok(())
            })
            .expect("immediate transaction should commit");

        assert_eq!(
            count_documents(storage.connection()).expect("count after immediate tx"),
            1
        );

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.tx_commits, 1);
    }

    #[test]
    fn immediate_transaction_rolls_back_on_error() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");

        let result: SearchResult<()> = storage.immediate_transaction(|conn| {
            insert_document_minimal(conn, "doc-immediate-rollback")?;
            Err(SearchError::InvalidConfig {
                field: "test".to_owned(),
                value: "forced".to_owned(),
                reason: "force rollback".to_owned(),
            })
        });

        assert!(result.is_err());
        assert_eq!(
            count_documents(storage.connection()).expect("count after rollback"),
            0
        );

        let metrics = storage.metrics_snapshot();
        assert_eq!(metrics.tx_rollbacks, 1);
    }

    #[test]
    fn storage_debug_format() {
        let storage = Storage::open_in_memory().expect("in-memory storage should open");
        let debug = format!("{storage:?}");
        assert!(debug.contains("Storage"));
        assert!(debug.contains(":memory:"));
    }

    #[test]
    fn multi_threaded_serial_writers_commit_disjoint_documents() {
        let tmp = TempDbPath::new("concurrent-writers");
        let config = tmp.config();
        let (first_done_tx, first_done_rx) = mpsc::channel::<()>();

        let cfg_a = config.clone();
        let first_writer = thread::spawn(move || {
            let storage = Storage::open(cfg_a).expect("first writer storage should open");
            storage
                .transaction(|conn| {
                    insert_document_minimal(conn, "doc-a")?;
                    Ok(())
                })
                .expect("first writer transaction should commit");
            first_done_tx
                .send(())
                .expect("first writer should signal completion");
        });

        let cfg_b = config.clone();
        let second_writer = thread::spawn(move || {
            first_done_rx
                .recv()
                .expect("second writer should wait for first writer");
            let storage = Storage::open(cfg_b).expect("second writer storage should open");
            storage
                .transaction(|conn| {
                    insert_document_minimal(conn, "doc-b")?;
                    Ok(())
                })
                .expect("second writer transaction should commit");
        });

        first_writer
            .join()
            .expect("first writer thread should not panic");
        second_writer
            .join()
            .expect("second writer thread should not panic");

        let verifier = Storage::open(config).expect("verifier storage should open");
        assert_eq!(
            count_documents(verifier.connection()).expect("count after concurrent writes"),
            2,
            "both writer transactions should persist exactly one row"
        );
    }
}