orion-server 1.0.0

Turn business logic into live REST/Kafka services. Declare workflows as JSON and Orion runs them, with rate limiting, circuit breakers, versioning, and observability built in
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
904
905
906
907
908
909
910
911
912
913
914
915
916
917
918
919
920
921
922
923
924
925
926
927
928
929
930
931
932
933
934
935
936
937
938
939
940
941
942
943
944
945
946
947
948
949
950
951
952
953
954
955
956
957
958
959
960
961
962
963
964
965
966
967
968
969
970
971
972
973
974
975
976
977
978
979
980
981
982
983
984
985
986
987
988
989
990
991
992
993
994
995
996
997
998
999
1000
1001
1002
1003
1004
1005
1006
1007
1008
1009
1010
1011
1012
1013
1014
1015
1016
1017
1018
1019
1020
1021
1022
1023
1024
1025
1026
1027
1028
1029
1030
1031
1032
1033
1034
1035
1036
1037
1038
1039
1040
1041
1042
1043
1044
1045
1046
1047
1048
1049
1050
1051
1052
1053
1054
1055
1056
1057
1058
1059
1060
1061
1062
1063
1064
1065
1066
1067
1068
1069
1070
1071
1072
1073
1074
1075
1076
1077
1078
1079
1080
1081
1082
1083
1084
1085
1086
1087
1088
1089
1090
1091
1092
1093
1094
1095
1096
1097
1098
1099
1100
1101
1102
1103
1104
1105
1106
1107
1108
1109
1110
1111
1112
1113
1114
1115
1116
1117
1118
1119
1120
1121
1122
1123
1124
1125
1126
1127
1128
1129
1130
1131
1132
1133
1134
1135
1136
1137
1138
1139
1140
1141
1142
1143
1144
1145
1146
1147
1148
1149
1150
1151
1152
1153
1154
1155
1156
1157
1158
1159
1160
1161
1162
1163
1164
1165
1166
1167
1168
1169
1170
1171
1172
1173
1174
1175
1176
1177
1178
1179
1180
1181
1182
1183
1184
1185
1186
1187
1188
1189
1190
1191
1192
1193
1194
1195
1196
1197
1198
1199
1200
1201
1202
1203
1204
1205
1206
1207
1208
1209
1210
1211
1212
1213
1214
1215
1216
1217
1218
1219
1220
1221
1222
1223
1224
1225
1226
1227
1228
1229
1230
1231
1232
1233
1234
1235
1236
1237
1238
1239
1240
1241
1242
1243
1244
1245
1246
1247
1248
1249
1250
1251
1252
1253
1254
1255
1256
1257
1258
1259
1260
1261
1262
1263
1264
1265
1266
1267
1268
1269
1270
1271
1272
1273
1274
1275
1276
1277
1278
1279
1280
1281
1282
1283
1284
1285
1286
1287
1288
1289
1290
1291
1292
1293
1294
1295
1296
1297
1298
1299
1300
1301
1302
1303
1304
1305
1306
1307
1308
1309
1310
1311
1312
1313
1314
1315
1316
1317
1318
1319
1320
1321
1322
1323
1324
1325
1326
1327
1328
1329
1330
1331
1332
1333
1334
1335
1336
1337
1338
1339
1340
1341
1342
1343
1344
1345
1346
1347
1348
1349
1350
1351
1352
1353
1354
1355
1356
1357
1358
1359
1360
1361
1362
1363
1364
1365
1366
1367
1368
1369
1370
1371
1372
1373
1374
1375
1376
1377
1378
1379
1380
1381
1382
1383
1384
1385
1386
1387
1388
1389
1390
1391
1392
1393
1394
1395
1396
1397
1398
1399
1400
1401
1402
1403
1404
1405
1406
1407
1408
1409
1410
1411
1412
1413
1414
1415
1416
1417
1418
1419
1420
1421
1422
1423
1424
1425
1426
1427
1428
1429
1430
1431
1432
1433
1434
1435
1436
1437
1438
1439
1440
1441
1442
1443
1444
1445
1446
1447
1448
1449
1450
1451
1452
1453
1454
1455
1456
1457
1458
1459
1460
1461
1462
1463
1464
1465
1466
1467
1468
1469
1470
1471
1472
1473
1474
1475
1476
1477
1478
1479
1480
1481
1482
1483
1484
1485
1486
1487
1488
1489
1490
1491
1492
1493
1494
1495
1496
1497
1498
1499
1500
1501
1502
1503
1504
1505
1506
1507
1508
1509
1510
1511
1512
1513
1514
1515
1516
1517
1518
1519
1520
1521
1522
1523
1524
1525
1526
1527
1528
1529
1530
1531
1532
1533
1534
1535
1536
1537
1538
1539
1540
1541
1542
1543
1544
1545
1546
1547
1548
1549
1550
1551
1552
1553
1554
1555
1556
1557
1558
1559
1560
1561
1562
1563
1564
1565
//! SQL rendering over sea-query.
//!
//! Walks a [`Cond`] + [`QuerySpec`] into a `sea_query::SelectStatement`, then
//! [`build_for`] produces the dialect-specific `(sql, values)` bound to the
//! external connector's `sqlx::AnyPool`. Identifiers are dynamic
//! (`sea_query::Alias`) and every literal is a bound value, so the output is
//! injection-safe and quoted per dialect.

use sea_query::{
    Alias, Asterisk, Condition, Expr, ExprTrait, Func, LikeExpr, MysqlQueryBuilder, NullOrdering,
    OnConflict, Order, OrderedStatement, PostgresQueryBuilder, Query, SelectStatement, SimpleExpr,
    SqliteQueryBuilder, Value as SeaValue, WindowStatement,
};
use sea_query_sqlx::{SqlxBinder, SqlxValues};

use crate::config::QueryConfig;
use crate::query::IncludePlan;
use crate::query::backend::SqlDialect;
use crate::query::error::QueryError;
use crate::query::ir::{CmpOp, Cond, FieldRef, Quant, RelRef, TextOp, Value};
use crate::query::spec::{QuerySpec, SortKey};
use crate::query::write::{ConflictAction, ResolvedConflict, ResolvedWrite, WriteError};

/// Column the include's window function writes its per-parent row number into.
/// Prefixed so it cannot collide with a real column; stripped from the output.
pub const INCLUDE_RANK_COLUMN: &str = "__orion_include_rank";

/// Alias for the include's inner (windowed) sub-select. A window function is not
/// usable in `WHERE`, so the rank is computed in a subquery and filtered outside.
const INCLUDE_SUBQUERY_ALIAS: &str = "__orion_include";

/// Build a `SelectStatement` from the envelope and lowered condition, enforcing
/// the page bounds (`LimitExceeded` / `SkipExceeded` when over the configured
/// caps). `root_table` is the physical table the query selects from (and the
/// correlation base for any relation subqueries).
pub fn render(
    spec: &QuerySpec,
    cond: &Cond,
    root_table: &str,
    dialect: SqlDialect,
    limits: &QueryConfig,
) -> Result<SelectStatement, QueryError> {
    let limit = resolve_limit(spec.limit, limits)?;
    let skip = resolve_skip(spec.skip, limits)?;

    let mut stmt = Query::select();
    match super::plan_projection(&spec.fields) {
        None => {
            stmt.column(Asterisk);
        }
        Some(fields) => {
            for f in fields {
                stmt.column(Alias::new(f.as_str()));
            }
        }
    }
    stmt.from(Alias::new(root_table));
    // Skip the WHERE clause entirely when the filter is unconditionally true, so
    // a filterless query does not render a redundant `WHERE TRUE`.
    if !matches!(cond, Cond::True) {
        stmt.cond_where(render_expr(cond, root_table)?);
    }
    apply_sort_keys(&mut stmt, &spec.sort, dialect);
    stmt.limit(limit);
    if let Some(skip) = skip {
        stmt.offset(skip);
    }
    Ok(stmt)
}

/// Produce the dialect-specific SQL string + bound values for execution on an
/// `AnyPool`. The dialect must match the connector's real driver (see
/// [`SqlDialect`]).
pub fn build_for(dialect: SqlDialect, stmt: &SelectStatement) -> (String, SqlxValues) {
    match dialect {
        SqlDialect::Postgres => stmt.build_sqlx(PostgresQueryBuilder),
        SqlDialect::Mysql => stmt.build_sqlx(MysqlQueryBuilder),
        SqlDialect::Sqlite => stmt.build_sqlx(SqliteQueryBuilder),
    }
}

/// Build the child query for an `include`, with the per-parent page cut **in
/// SQL** (F27):
///
/// ```sql
/// SELECT <projection> FROM (
///   SELECT <projection>,
///          ROW_NUMBER() OVER (PARTITION BY <foreign> ORDER BY <sort>) AS <rank>
///   FROM <target> WHERE <foreign> IN (<keys>)
/// ) AS <alias>
/// WHERE <rank> <= <limit>
/// ORDER BY <sort>
/// ```
///
/// This used to be a bare `SELECT … WHERE fk IN (…)` with no `LIMIT` and no
/// `ORDER BY`, and the handler truncated afterwards: 1000 parents × 10 000
/// children materialised 10M rows to return 5 each, and because nothing ordered
/// them, `include.limit` returned an *arbitrary* subset that could differ run to
/// run. The window function is supported by every backend the dialect renders
/// for (SQLite ≥ 3.25 — the bundled build is far newer, PostgreSQL ≥ 8.4,
/// MySQL ≥ 8.0).
///
/// `<projection>` is [`IncludePlan::projection`] on **both** levels: the
/// requested fields plus the foreign key (for grouping) plus the sort keys,
/// because the outer `ORDER BY` can only name columns the sub-select emits. The
/// handler drops the extras again ([`IncludePlan::strip`]). Callers must skip
/// this when `keys` is empty (an empty `IN` is not built here).
pub fn build_include_select(
    inc: &IncludePlan,
    keys: &[SeaValue],
    dialect: SqlDialect,
) -> (String, SqlxValues) {
    let foreign = inc.foreign.as_str();
    let projection = inc.projection();

    // Inner: the child rows for this parent page, ranked within each parent.
    let mut inner = Query::select();
    project_child(&mut inner, &projection);
    let mut window = WindowStatement::partition_by(Alias::new(foreign));
    apply_sort_keys(&mut window, &inc.sort, dialect);
    inner.expr_window_as(
        Func::cust(Alias::new("ROW_NUMBER")),
        window,
        Alias::new(INCLUDE_RANK_COLUMN),
    );
    inner.from(Alias::new(inc.target_table.as_str()));
    inner.cond_where(Expr::col(Alias::new(foreign)).is_in(keys.to_vec()));

    // Outer: keep the first `limit` per parent, in the requested order.
    let mut stmt = Query::select();
    project_child(&mut stmt, &projection);
    stmt.from_subquery(inner, Alias::new(INCLUDE_SUBQUERY_ALIAS));
    // Bound as `i64`: the `sqlx-any` binder converts a `BigUnsigned` with an
    // unchecked `try_from`, which panics above `i64::MAX`.
    let cap = i64::try_from(inc.limit).unwrap_or(i64::MAX);
    stmt.cond_where(Expr::col(Alias::new(INCLUDE_RANK_COLUMN)).lte(cap));
    apply_sort_keys(&mut stmt, &inc.sort, dialect);
    build_for(dialect, &stmt)
}

/// Project a child row from [`IncludePlan::projection`]. An empty projection is
/// the unprojected case (`include` with no `fields`) and means every column —
/// the same [`super::plan_projection`] rule the top-level select follows.
fn project_child(stmt: &mut SelectStatement, projection: &[String]) {
    match super::plan_projection(projection) {
        None => {
            stmt.column(Asterisk);
        }
        Some(fields) => {
            for f in fields {
                stmt.column(Alias::new(f.as_str()));
            }
        }
    }
}

/// Convert a JSON scalar (a parent key value) into a bound `sea_query::Value` for
/// an `IN` list. Returns `None` for null / non-scalar values (skipped as keys).
pub fn json_key_to_sea(v: &serde_json::Value) -> Option<SeaValue> {
    match v {
        serde_json::Value::String(s) => Some(s.clone().into()),
        serde_json::Value::Bool(b) => Some((*b).into()),
        serde_json::Value::Number(n) => n
            .as_i64()
            .map(Into::into)
            .or_else(|| n.as_f64().map(Into::into)),
        _ => None,
    }
}

use super::{resolve_limit, resolve_skip};

/// Render a `Cond` as a single boolean `SimpleExpr`. Composing on `SimpleExpr`
/// (rather than `Condition`) keeps `and`/`or`/`not`, EXISTS subqueries, and the
/// the `all` null rule uniformly expressible. `current_table` is the physical
/// table the bare columns belong to (the correlation base for relations).
fn render_expr(cond: &Cond, current_table: &str) -> Result<SimpleExpr, QueryError> {
    Ok(match cond {
        Cond::True => Expr::val(1).eq(1),
        Cond::False => Expr::val(1).eq(0),
        Cond::And(cs) => fold_bool(cs, current_table, true)?,
        Cond::Or(cs) => fold_bool(cs, current_table, false)?,
        Cond::Not(inner) => render_expr(inner, current_table)?.not(),
        Cond::Compare { field, op, value } => compare_expr(field, *op, value),
        Cond::In {
            field,
            values,
            negated,
        } => in_expr(field, values, *negated),
        Cond::IsNull { field, negated } => {
            let col = col_expr(field);
            if *negated {
                col.is_not_null()
            } else {
                col.is_null()
            }
        }
        Cond::Between {
            field,
            low,
            high,
            low_incl,
            high_incl,
            negated,
        } => between_expr(field, low, high, *low_incl, *high_incl, *negated),
        Cond::Text { field, op, pattern } => text_expr(field, *op, pattern),
        Cond::Rel { quant, rel, cond } => render_rel(*quant, rel, cond, current_table)?,
    })
}

/// Fold a non-empty list of conditions with AND (`and = true`) or OR.
fn fold_bool(cs: &[Cond], current_table: &str, and: bool) -> Result<SimpleExpr, QueryError> {
    let mut iter = cs.iter();
    let mut acc = match iter.next() {
        Some(first) => render_expr(first, current_table)?,
        // Empty groups are folded to True/False at lowering; be defensive anyway.
        None => return Ok(Expr::val(1).eq(if and { 1 } else { 0 })),
    };
    for c in iter {
        let e = render_expr(c, current_table)?;
        acc = if and { acc.and(e) } else { acc.or(e) };
    }
    Ok(acc)
}

/// Render a relation predicate as an EXISTS-style semi/anti join.
fn render_rel(
    quant: Quant,
    rel: &RelRef,
    inner: &Cond,
    current_table: &str,
) -> Result<SimpleExpr, QueryError> {
    // Inner columns belong to the target entity's table.
    let target = rel.target_table.as_str();
    match quant {
        Quant::Any => {
            let inner_e = render_expr(inner, target)?;
            Ok(Expr::exists(rel_subquery(
                rel,
                current_table,
                Some(inner_e),
            )?))
        }
        Quant::None => {
            let inner_e = render_expr(inner, target)?;
            Ok(Expr::exists(rel_subquery(rel, current_table, Some(inner_e))?).not())
        }
        Quant::All => {
            // Reference semantics: false on an empty relation, and a null inner
            // predicate counts as a counterexample. Rendered as
            //   EXISTS(rel) AND NOT EXISTS(rel WHERE NOT c OR c IS NULL)
            //   (the `all` null rule)
            let nonempty = Expr::exists(rel_subquery(rel, current_table, None)?);
            let ie = render_expr(inner, target)?;
            let violates = ie.clone().not().or(Expr::expr(ie).is_null());
            let no_violation =
                Expr::exists(rel_subquery(rel, current_table, Some(violates))?).not();
            Ok(nonempty.and(no_violation))
        }
    }
}

/// Build `SELECT 1 FROM <rel> WHERE <correlation> [AND <inner>]`, joining through
/// the junction for a many-to-many relation.
fn rel_subquery(
    rel: &RelRef,
    current_table: &str,
    inner: Option<SimpleExpr>,
) -> Result<SelectStatement, QueryError> {
    let mut sub = Query::select();
    sub.expr(Expr::val(1));
    let mut where_cond = Condition::all();
    match &rel.through {
        None => {
            // Direct: target.foreign = current.local
            sub.from(Alias::new(rel.target_table.as_str()));
            where_cond = where_cond.add(
                Expr::col((
                    Alias::new(rel.target_table.as_str()),
                    Alias::new(rel.foreign.as_str()),
                ))
                .equals((Alias::new(current_table), Alias::new(rel.local.as_str()))),
            );
        }
        Some(j) => {
            // M:M: junction.local = current.local, junction.foreign = target.foreign
            sub.from(Alias::new(j.table.as_str()));
            sub.inner_join(
                Alias::new(rel.target_table.as_str()),
                Expr::col((Alias::new(j.table.as_str()), Alias::new(j.foreign.as_str()))).equals((
                    Alias::new(rel.target_table.as_str()),
                    Alias::new(rel.foreign.as_str()),
                )),
            );
            where_cond = where_cond.add(
                Expr::col((Alias::new(j.table.as_str()), Alias::new(j.local.as_str())))
                    .equals((Alias::new(current_table), Alias::new(rel.local.as_str()))),
            );
        }
    }
    if let Some(e) = inner {
        where_cond = where_cond.add(e);
    }
    sub.cond_where(where_cond);
    Ok(sub)
}

fn compare_expr(field: &FieldRef, op: CmpOp, value: &Value) -> SimpleExpr {
    let col = col_expr(field);
    let v = to_sea_value(value);
    match op {
        CmpOp::Eq => col.eq(v),
        CmpOp::Ne => col.ne(v),
        CmpOp::Lt => col.lt(v),
        CmpOp::Le => col.lte(v),
        CmpOp::Gt => col.gt(v),
        CmpOp::Ge => col.gte(v),
    }
}

fn in_expr(field: &FieldRef, values: &[Value], negated: bool) -> SimpleExpr {
    let col = col_expr(field);
    let vals: Vec<SeaValue> = values.iter().map(to_sea_value).collect();
    if negated {
        col.is_not_in(vals)
    } else {
        col.is_in(vals)
    }
}

fn between_expr(
    field: &FieldRef,
    low: &Value,
    high: &Value,
    low_incl: bool,
    high_incl: bool,
    negated: bool,
) -> SimpleExpr {
    let lo = to_sea_value(low);
    let hi = to_sea_value(high);
    // Native BETWEEN is inclusive-only; use it only when both bounds are
    // inclusive, else render explicit per-bound comparisons (the
    // chained-range rule).
    let e = if low_incl && high_incl {
        col_expr(field).between(lo, hi)
    } else {
        let lo_e = if low_incl {
            col_expr(field).gte(lo)
        } else {
            col_expr(field).gt(lo)
        };
        let hi_e = if high_incl {
            col_expr(field).lte(hi)
        } else {
            col_expr(field).lt(hi)
        };
        lo_e.and(hi_e)
    };
    if negated { e.not() } else { e }
}

fn text_expr(field: &FieldRef, op: TextOp, pattern: &str) -> SimpleExpr {
    let escaped = escape_like(pattern);
    let like = match op {
        TextOp::StartsWith => format!("{escaped}%"),
        TextOp::EndsWith => format!("%{escaped}"),
        TextOp::Contains => format!("%{escaped}%"),
    };
    // W13: `LIKE` with `\` as the escape character. Case sensitivity is the one
    // thing the dialect does **not** normalise: it is a property of the stored
    // data (a SQL collation, an ES analyzer), not of the query, and no query-time
    // flag can restore case-sensitive matching against an Elasticsearch `text`
    // field whose analyzer already folded the tokens. So the per-backend truth is
    // stated in the parity table of `docs/src/reference/data-dialect.md`
    // ("Text-match case sensitivity") rather than papered over here: PostgreSQL
    // `LIKE` is case-sensitive, SQLite's folds ASCII, MySQL's follows the
    // column's collation (`_ci` by default).
    col_expr(field).like(LikeExpr::new(like).escape('\\'))
}

/// Escape the LIKE metacharacters in user-provided text so they match literally.
fn escape_like(pattern: &str) -> String {
    pattern
        .replace('\\', "\\\\")
        .replace('%', "\\%")
        .replace('_', "\\_")
}

/// Apply the shared sort plan ([`super::plan_sort`], which owns the W8
/// null-ordering rule) to anything that takes an `ORDER BY` — the statement
/// itself, or the `OVER (…)` clause of the include's window (F27).
///
/// Only MySQL is special-cased here: it has no `NULLS FIRST`/`NULLS LAST`
/// syntax and needs none, because its native ordering already places nulls
/// first on ASC and last on DESC — exactly the planned placement. SQLite's
/// native order agrees too but accepts the clause, so it gets it, redundantly
/// but harmlessly — one rendering path rather than a second exception to keep
/// in step. PostgreSQL (which sorts nulls as the largest value) is the backend
/// the explicit clause exists for.
fn apply_sort_keys<S: OrderedStatement>(stmt: &mut S, sort: &[SortKey], dialect: SqlDialect) {
    for p in super::plan_sort(sort) {
        let order = if p.ascending { Order::Asc } else { Order::Desc };
        let nulls = if p.nulls_first {
            NullOrdering::First
        } else {
            NullOrdering::Last
        };
        match dialect {
            SqlDialect::Mysql => {
                stmt.order_by(Alias::new(p.field), order);
            }
            _ => {
                stmt.order_by_with_nulls(Alias::new(p.field), order, nulls);
            }
        }
    }
}

fn col_expr(field: &FieldRef) -> Expr {
    Expr::col(Alias::new(field.physical.as_str()))
}

/// Convert an IR value to a `sea_query::Value`, restricted to the variants the
/// `sqlx-any` binder accepts (Bool/BigInt/Double/String) — never Decimal/Json,
/// which panic under `AnyArguments`. IR values are always scalars (nested lists
/// are rejected during lowering), so this conversion is total.
fn to_sea_value(v: &Value) -> SeaValue {
    match v {
        // A scalar null is already lowered to `IsNull`; this placeholder only
        // arises for an explicit null inside a list, binding as SQL NULL.
        Value::Null => SeaValue::String(None),
        Value::Bool(b) => (*b).into(),
        Value::Int(i) => (*i).into(),
        Value::Float(f) => (*f).into(),
        Value::Str(s) => s.clone().into(),
    }
}

// ---- Write rendering (INSERT / UPDATE / DELETE / upsert) ----

/// Render a resolved mutation into `(sql, values)` for `dialect`, bound to the
/// external connector's `AnyPool`. The `filter` of an update/delete reuses the
/// query dialect's [`render_expr`]; every value is a bound parameter.
pub fn render_write(
    w: &ResolvedWrite,
    dialect: SqlDialect,
) -> Result<(String, SqlxValues), WriteError> {
    // MySQL cannot express RETURNING; surface it rather than emitting invalid SQL.
    if !w.returning().is_empty() && dialect == SqlDialect::Mysql {
        return Err(WriteError::Query(QueryError::FeatureUnsupportedByTarget {
            feature: "returning".to_string(),
            target: "mysql".to_string(),
        }));
    }
    match w {
        ResolvedWrite::Insert {
            table,
            columns,
            rows,
            returning,
        } => render_insert(table, columns, rows, None, returning, dialect),
        ResolvedWrite::Upsert {
            table,
            columns,
            rows,
            set,
            conflict,
            returning,
        } => render_insert(
            table,
            columns,
            rows,
            Some((conflict, set)),
            returning,
            dialect,
        ),
        ResolvedWrite::Update {
            table,
            set,
            cond,
            returning,
        } => render_update(table, set, cond, returning, dialect),
        ResolvedWrite::Delete {
            table,
            cond,
            returning,
        } => render_delete(table, cond, returning, dialect),
    }
}

/// Render an INSERT, with `upsert` carrying the conflict clause and on-conflict
/// assignments when the mutation is an upsert.
fn render_insert(
    table: &str,
    columns: &[String],
    rows: &[Vec<Value>],
    upsert: Option<(&ResolvedConflict, &[(String, Value)])>,
    returning: &[String],
    dialect: SqlDialect,
) -> Result<(String, SqlxValues), WriteError> {
    let mut stmt = Query::insert();
    stmt.into_table(Alias::new(table));
    stmt.columns(columns.iter().map(|c| Alias::new(c.as_str())));
    for row in rows {
        let vals: Vec<SimpleExpr> = row.iter().map(value_expr).collect();
        stmt.values(vals)
            .map_err(|e| WriteError::Query(QueryError::InvalidEnvelope(e.to_string())))?;
    }

    if let Some((c, set)) = upsert {
        let mut oc = OnConflict::columns(c.targets.iter().map(|t| Alias::new(t.as_str())));
        match c.action {
            ConflictAction::Nothing => {
                oc.do_nothing();
            }
            ConflictAction::Update => {
                if !set.is_empty() {
                    for (col, v) in set {
                        oc.value(Alias::new(col.as_str()), value_expr(v));
                    }
                } else {
                    // Default: overwrite every inserted column except the conflict keys.
                    let upd: Vec<Alias> = columns
                        .iter()
                        .filter(|c2| !c.targets.contains(c2))
                        .map(|c2| Alias::new(c2.as_str()))
                        .collect();
                    if upd.is_empty() {
                        oc.do_nothing();
                    } else {
                        oc.update_columns(upd);
                    }
                }
            }
        }
        stmt.on_conflict(oc);
    }

    if !returning.is_empty() {
        stmt.returning(
            Query::returning().columns(returning.iter().map(|c| Alias::new(c.as_str()))),
        );
    }
    Ok(build_write_for(dialect, &stmt))
}

fn render_update(
    table: &str,
    set: &[(String, Value)],
    cond: &Option<Cond>,
    returning: &[String],
    dialect: SqlDialect,
) -> Result<(String, SqlxValues), WriteError> {
    let mut stmt = Query::update();
    stmt.table(Alias::new(table));
    for (col, v) in set {
        stmt.value(Alias::new(col.as_str()), value_expr(v));
    }
    if let Some(cond) = cond
        && !matches!(cond, Cond::True)
    {
        stmt.cond_where(render_expr(cond, table).map_err(WriteError::from)?);
    }
    if !returning.is_empty() {
        stmt.returning(
            Query::returning().columns(returning.iter().map(|c| Alias::new(c.as_str()))),
        );
    }
    Ok(build_write_for(dialect, &stmt))
}

fn render_delete(
    table: &str,
    cond: &Option<Cond>,
    returning: &[String],
    dialect: SqlDialect,
) -> Result<(String, SqlxValues), WriteError> {
    let mut stmt = Query::delete();
    stmt.from_table(Alias::new(table));
    if let Some(cond) = cond
        && !matches!(cond, Cond::True)
    {
        stmt.cond_where(render_expr(cond, table).map_err(WriteError::from)?);
    }
    if !returning.is_empty() {
        stmt.returning(
            Query::returning().columns(returning.iter().map(|c| Alias::new(c.as_str()))),
        );
    }
    Ok(build_write_for(dialect, &stmt))
}

/// An IR value as a bound `SimpleExpr` (a NULL binds as SQL NULL).
fn value_expr(v: &Value) -> SimpleExpr {
    Expr::val(to_sea_value(v))
}

/// Dialect-specific `(sql, values)` for any write statement (Insert/Update/Delete).
fn build_write_for<S: SqlxBinder>(dialect: SqlDialect, stmt: &S) -> (String, SqlxValues) {
    match dialect {
        SqlDialect::Postgres => stmt.build_sqlx(PostgresQueryBuilder),
        SqlDialect::Mysql => stmt.build_sqlx(MysqlQueryBuilder),
        SqlDialect::Sqlite => stmt.build_sqlx(SqliteQueryBuilder),
    }
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::query::{EntityRegistry, plan_sql};
    use serde_json::{Value as Json, json};

    /// The default page bounds used by these goldens (limit 100/1000).
    fn limits() -> QueryConfig {
        QueryConfig::default()
    }

    /// Plan `query` through the production entry point ([`plan_sql`]) and
    /// return the main `SelectStatement`, so the goldens cannot drift from
    /// the path the handler actually takes (W23).
    fn plan_stmt(
        query: &Json,
        reg: &EntityRegistry,
        dialect: SqlDialect,
        limits: &QueryConfig,
    ) -> Result<SelectStatement, QueryError> {
        plan_sql(query, &serde_json::Map::new(), reg, dialect, limits).map(|plan| plan.main)
    }

    /// Render `query` for `dialect` with values inlined, for golden assertions.
    fn sql_for(query: Json, dialect: SqlDialect) -> String {
        let stmt = plan_stmt(&query, &EntityRegistry::identity(), dialect, &limits())
            .expect("translation should succeed");
        match dialect {
            SqlDialect::Sqlite => stmt.to_string(SqliteQueryBuilder),
            SqlDialect::Postgres => stmt.to_string(PostgresQueryBuilder),
            SqlDialect::Mysql => stmt.to_string(MysqlQueryBuilder),
        }
    }

    fn sqlite(query: Json) -> String {
        sql_for(query, SqlDialect::Sqlite)
    }

    /// users→orders (has_many) and users↔tags (many-to-many via user_tags).
    fn rel_schema() -> EntityRegistry {
        EntityRegistry::from_json(&json!({
            // These tests are about relation *rendering*, not the allowlist:
            // they declare relations and let the inner predicate's columns
            // resolve by identity, which F24 now makes an explicit opt-in.
            "unmapped": "identity",
            "entities": {
                "users": {
                    "relations": {
                        "orders": { "to": "orders", "kind": "has_many", "local": "id", "foreign": "user_id" },
                        "tags": {
                            "to": "tags", "kind": "many_to_many", "local": "id", "foreign": "id",
                            "through": { "table": "user_tags", "local": "user_id", "foreign": "tag_id" }
                        }
                    }
                }
            }
        }))
        .expect("valid schema")
    }

    fn sqlite_schema(query: Json) -> String {
        let stmt = plan_stmt(&query, &rel_schema(), SqlDialect::Sqlite, &limits())
            .expect("translation should succeed");
        stmt.to_string(SqliteQueryBuilder)
    }

    #[test]
    fn test_select_all_no_filter() {
        let sql = sqlite(json!({ "source": "users" }));
        assert_eq!(sql, r#"SELECT * FROM "users" LIMIT 100"#);
    }

    #[test]
    fn test_projection_and_comparison() {
        let sql = sqlite(json!({
            "source": "users",
            "fields": ["id", "name"],
            "filter": { ">": [{"field": "age"}, 18] }
        }));
        assert_eq!(
            sql,
            r#"SELECT "id", "name" FROM "users" WHERE "age" > 18 LIMIT 100"#
        );
    }

    #[test]
    fn test_and_or() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "and": [
                { "==": [{"field": "a"}, 1] },
                { "or": [ { "==": [{"field": "b"}, 2] }, { "==": [{"field": "c"}, 3] } ] }
            ] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "a" = 1 AND ("b" = 2 OR "c" = 3) LIMIT 100"#
        );
    }

    #[test]
    fn test_membership() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "in": [{"field": "status"}, ["a", "b"]] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "status" IN ('a', 'b') LIMIT 100"#
        );
    }

    #[test]
    fn test_empty_membership_is_false() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "in": [{"field": "status"}, []] }
        }));
        assert_eq!(sql, r#"SELECT * FROM "t" WHERE 1 = 0 LIMIT 100"#);
    }

    #[test]
    fn test_is_null() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "==": [{"field": "email"}, null] }
        }));
        assert_eq!(sql, r#"SELECT * FROM "t" WHERE "email" IS NULL LIMIT 100"#);
    }

    #[test]
    fn test_range_strict_is_not_between() {
        // Chained `<` is strict → explicit > AND <, NOT inclusive BETWEEN.
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "<": [1, {"field": "x"}, 10] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "x" > 1 AND "x" < 10 LIMIT 100"#
        );
    }

    #[test]
    fn test_range_inclusive_is_between() {
        // Chained `<=` is inclusive → native BETWEEN.
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "<=": [1, {"field": "x"}, 10] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "x" BETWEEN 1 AND 10 LIMIT 100"#
        );
    }

    #[test]
    fn test_text_contains_escapes_wildcards() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "in": ["50%_off", {"field": "name"}] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "name" LIKE '%50\%\_off%' ESCAPE '\' LIMIT 100"#
        );
    }

    #[test]
    fn test_starts_with() {
        let sql = sqlite(json!({
            "source": "t",
            "filter": { "starts_with": [{"field": "name"}, "sm"] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" WHERE "name" LIKE 'sm%' ESCAPE '\' LIMIT 100"#
        );
    }

    #[test]
    fn test_sort_and_paging() {
        let sql = sqlite(json!({
            "source": "t",
            "sort": [ { "created_at": "desc" } ],
            "limit": 20,
            "skip": 40
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "t" ORDER BY "created_at" DESC NULLS LAST LIMIT 20 OFFSET 40"#
        );
    }

    /// W8: the shared rule is "a null sorts as the smallest value" — nulls
    /// first on `asc`, last on `desc`. It used to be the inverse, which no
    /// MongoDB `find` can express, so Mongo silently disagreed with SQL and ES
    /// on the ordering of every page containing a null.
    #[test]
    fn test_null_ordering_is_nulls_smallest() {
        assert_eq!(
            sqlite(json!({ "source": "t", "sort": [ { "name": "asc" } ] })),
            r#"SELECT * FROM "t" ORDER BY "name" ASC NULLS FIRST LIMIT 100"#
        );
        let stmt = plan_stmt(
            &json!({ "source": "t", "sort": [ { "name": "asc" } ] }),
            &EntityRegistry::identity(),
            SqlDialect::Postgres,
            &limits(),
        )
        .expect("ok");
        assert_eq!(
            stmt.to_string(PostgresQueryBuilder),
            r#"SELECT * FROM "t" ORDER BY "name" ASC NULLS FIRST LIMIT 100"#
        );
    }

    #[test]
    fn test_postgres_placeholders_via_build() {
        // Bound-parameter form (execution path): Postgres uses $1 placeholders.
        let stmt = plan_stmt(
            &json!({ "source": "users", "filter": { "==": [{"field": "id"}, 7] } }),
            &EntityRegistry::identity(),
            SqlDialect::Postgres,
            &limits(),
        )
        .expect("ok");
        let (sql, _values) = build_for(SqlDialect::Postgres, &stmt);
        // In the bound-parameter path, LIMIT is also a placeholder ($2).
        assert_eq!(sql, r#"SELECT * FROM "users" WHERE "id" = $1 LIMIT $2"#);
    }

    /// W8: MySQL has no `NULLS FIRST/LAST` clause and now needs none — its
    /// native ordering already puts nulls first on `ASC` and last on `DESC`,
    /// which is the rule. The `IS NULL` prefix key this replaced was a second,
    /// invisible sort key emitted to emulate the old inverse rule.
    #[test]
    fn test_mysql_needs_no_null_ordering_emulation() {
        let stmt = plan_stmt(
            &json!({ "source": "t", "sort": [ { "name": "asc" } ] }),
            &EntityRegistry::identity(),
            SqlDialect::Mysql,
            &limits(),
        )
        .expect("ok");
        assert_eq!(
            stmt.to_string(MysqlQueryBuilder),
            "SELECT * FROM `t` ORDER BY `name` ASC LIMIT 100"
        );
    }

    #[test]
    fn test_limit_default_applied() {
        let stmt = plan_stmt(
            &json!({ "source": "t" }),
            &EntityRegistry::identity(),
            SqlDialect::Sqlite,
            &QueryConfig {
                default_limit: 50,
                ..QueryConfig::default()
            },
        )
        .expect("ok");
        assert_eq!(
            stmt.to_string(SqliteQueryBuilder),
            r#"SELECT * FROM "t" LIMIT 50"#
        );
    }

    #[test]
    fn test_limit_exceeds_max_rejected() {
        let err = plan_stmt(
            &json!({ "source": "t", "limit": 5000 }),
            &EntityRegistry::identity(),
            SqlDialect::Sqlite,
            &limits(),
        )
        .expect_err("over the cap");
        assert!(matches!(
            err,
            QueryError::LimitExceeded {
                requested: 5000,
                max: 1000
            }
        ));
    }

    /// W12: `skip` is bounded like `limit` — rejected over the cap, never
    /// clamped. The cap used to exist only on Elasticsearch.
    #[test]
    fn test_skip_exceeds_max_rejected() {
        let err = plan_stmt(
            &json!({ "source": "t", "skip": 51 }),
            &EntityRegistry::identity(),
            SqlDialect::Sqlite,
            &QueryConfig {
                max_skip: 50,
                ..QueryConfig::default()
            },
        )
        .expect_err("over the skip cap");
        assert!(matches!(
            err,
            QueryError::SkipExceeded {
                requested: 51,
                max: 50
            }
        ));
    }

    // ---- relations ----

    #[test]
    fn test_relation_some_exists() {
        let sql = sqlite_schema(json!({
            "source": "users",
            "filter": { "some": [{"field": "orders"}, {">": [{"field": "total"}, 100]}] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "users" WHERE EXISTS(SELECT 1 FROM "orders" WHERE "orders"."user_id" = "users"."id" AND "total" > 100) LIMIT 100"#
        );
    }

    #[test]
    fn test_relation_none_not_exists() {
        let sql = sqlite_schema(json!({
            "source": "users",
            "filter": { "none": [{"field": "orders"}, {">": [{"field": "total"}, 100]}] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "users" WHERE NOT EXISTS(SELECT 1 FROM "orders" WHERE "orders"."user_id" = "users"."id" AND "total" > 100) LIMIT 100"#
        );
    }

    #[test]
    fn test_relation_all_null_fix() {
        // all → EXISTS(rel) AND NOT EXISTS(rel WHERE NOT c OR c IS NULL)
        // (the `all` null rule)
        let sql = sqlite_schema(json!({
            "source": "users",
            "filter": { "all": [{"field": "orders"}, {">": [{"field": "total"}, 100]}] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "users" WHERE EXISTS(SELECT 1 FROM "orders" WHERE "orders"."user_id" = "users"."id") AND (NOT EXISTS(SELECT 1 FROM "orders" WHERE "orders"."user_id" = "users"."id" AND ((NOT "total" > 100) OR ("total" > 100) IS NULL))) LIMIT 100"#
        );
    }

    #[test]
    fn test_relation_many_to_many_join() {
        let sql = sqlite_schema(json!({
            "source": "users",
            "filter": { "some": [{"field": "tags"}, {"==": [{"field": "label"}, "vip"]}] }
        }));
        assert_eq!(
            sql,
            r#"SELECT * FROM "users" WHERE EXISTS(SELECT 1 FROM "user_tags" INNER JOIN "tags" ON "user_tags"."tag_id" = "tags"."id" WHERE "user_tags"."user_id" = "users"."id" AND "label" = 'vip') LIMIT 100"#
        );
    }

    #[test]
    fn test_two_sibling_relations_are_independent() {
        // Two `some`s → two independent EXISTS, never a cross product.
        let sql = sqlite_schema(json!({
            "source": "users",
            "filter": { "and": [
                { "some": [{"field": "orders"}, {">": [{"field": "total"}, 100]}] },
                { "some": [{"field": "orders"}, {"==": [{"field": "user_id"}, 7]}] }
            ] }
        }));
        assert!(
            sql.matches("EXISTS(SELECT 1 FROM \"orders\"").count() == 2,
            "sql = {sql}"
        );
    }

    // ---- include planning ----

    /// Plan an include with the given selection, in identity mode over
    /// [`rel_schema`].
    fn plan_include(
        selection: Json,
        limits: &QueryConfig,
    ) -> Result<crate::query::SqlPlan, QueryError> {
        crate::query::plan_sql(
            &json!({
                "source": "users",
                "fields": ["name"],
                "include": { "orders": selection }
            }),
            &serde_json::Map::new(),
            &rel_schema(),
            SqlDialect::Sqlite,
            limits,
        )
    }

    #[test]
    fn test_plan_sql_augments_parent_key_and_plans_include() {
        let plan = plan_include(
            json!({ "fields": ["total"], "sort": [{ "id": "asc" }], "limit": 5 }),
            &limits(),
        )
        .expect("plan");

        // The parent key `id` is added to the main select (for grouping) and
        // marked for stripping from the output.
        assert_eq!(
            plan.main.to_string(SqliteQueryBuilder),
            r#"SELECT "name", "id" FROM "users" LIMIT 100"#
        );
        assert_eq!(plan.strip, vec!["id".to_string()]);
        assert_eq!(plan.includes.len(), 1);
        let inc = &plan.includes[0];
        assert_eq!(inc.field, "orders");
        assert_eq!(inc.target_table, "orders");
        assert_eq!(inc.local, "id");
        assert_eq!(inc.foreign, "user_id");
        assert_eq!(inc.limit, 5);
        assert_eq!(inc.sort.len(), 1);
    }

    /// F27: an include without a `limit` used to fetch *every* child of every
    /// parent on the page. It now takes the envelope's own page policy —
    /// `default_limit` per parent.
    #[test]
    fn test_include_without_limit_takes_the_default_page_size() {
        let plan = plan_include(json!({ "sort": [{ "id": "asc" }] }), &limits()).expect("plan");
        assert_eq!(plan.includes[0].limit, 100);
    }

    /// F27: and one over the cap is rejected, never clamped — the same rule the
    /// envelope's own `limit` gets.
    #[test]
    fn test_include_limit_over_the_cap_is_rejected() {
        let err = plan_include(
            json!({ "sort": [{ "id": "asc" }], "limit": 5000 }),
            &limits(),
        )
        .expect_err("over the cap");
        assert!(
            matches!(
                err,
                QueryError::LimitExceeded {
                    requested: 5000,
                    max: 1000
                }
            ),
            "{err}"
        );
    }

    /// F27: the per-parent page is cut in SQL by a partitioned `ROW_NUMBER()`,
    /// with the requested order key. The old query had neither `LIMIT` nor
    /// `ORDER BY`: it materialised every child of the whole parent page and the
    /// handler truncated an arbitrary — and run-to-run unstable — prefix.
    #[test]
    fn test_include_child_query_pages_per_parent_in_sql() {
        let plan = plan_include(
            json!({ "fields": ["total"], "sort": [{ "total": "desc" }], "limit": 2 }),
            &limits(),
        )
        .expect("plan");
        let keys = vec![SeaValue::from("u1"), SeaValue::from("u2")];
        let (sql, _v) = build_include_select(&plan.includes[0], &keys, SqlDialect::Sqlite);
        assert_eq!(
            sql,
            concat!(
                r#"SELECT "total", "user_id" FROM (SELECT "total", "user_id", "#,
                r#"ROW_NUMBER() OVER ( PARTITION BY "user_id" ORDER BY "total" DESC NULLS LAST ) "#,
                r#"AS "__orion_include_rank" FROM "orders" WHERE "user_id" IN (?, ?)) "#,
                r#"AS "__orion_include" WHERE "__orion_include_rank" <= ? "#,
                r#"ORDER BY "total" DESC NULLS LAST"#
            ),
            "sql = {sql}"
        );
    }

    /// The outer `ORDER BY` names columns of the *sub-select's output*, so a
    /// sort key that is not in `fields` has to be projected anyway — and then
    /// dropped from the response. Projecting only `fields` + the foreign key
    /// produced `ORDER BY "created_at"` over a subquery that emits neither:
    /// PostgreSQL `column "created_at" does not exist`, MySQL `Unknown column
    /// 'created_at' in 'order clause'`, and on SQLite no error at all — the
    /// quoted name degrades to a string literal, so the clause is a constant and
    /// the children come back in window order. That is the undefined per-parent
    /// ordering F27 removes, reintroduced on the backend the default job runs.
    #[test]
    fn test_include_projects_a_sort_key_it_was_not_asked_for() {
        let plan = plan_include(
            json!({ "fields": ["total"], "sort": [{ "created_at": "desc" }], "limit": 5 }),
            &limits(),
        )
        .expect("plan");
        let inc = &plan.includes[0];
        assert_eq!(inc.projection(), ["total", "user_id", "created_at"]);
        // …and both plumbing columns come back out of the nested object.
        assert_eq!(inc.strip(), ["user_id", "created_at"]);

        let keys = vec![SeaValue::from("u1")];
        let (sql, _v) = build_include_select(inc, &keys, SqlDialect::Sqlite);
        assert_eq!(
            sql,
            concat!(
                r#"SELECT "total", "user_id", "created_at" FROM "#,
                r#"(SELECT "total", "user_id", "created_at", "#,
                r#"ROW_NUMBER() OVER ( PARTITION BY "user_id" ORDER BY "created_at" DESC NULLS LAST ) "#,
                r#"AS "__orion_include_rank" FROM "orders" WHERE "user_id" IN (?)) "#,
                r#"AS "__orion_include" WHERE "__orion_include_rank" <= ? "#,
                r#"ORDER BY "created_at" DESC NULLS LAST"#
            ),
            "sql = {sql}"
        );
    }

    /// A sort key that *is* projected is not duplicated, and the foreign key
    /// named in `fields` stays in the output.
    #[test]
    fn test_include_projection_does_not_duplicate_or_over_strip() {
        let plan = plan_include(
            json!({ "fields": ["total", "user_id"], "sort": [{ "total": "asc" }] }),
            &limits(),
        )
        .expect("plan");
        let inc = &plan.includes[0];
        assert_eq!(inc.projection(), ["total", "user_id"]);
        assert!(inc.strip().is_empty(), "strip = {:?}", inc.strip());
    }

    /// With no `fields` the sub-select is `SELECT *`, which already carries the
    /// foreign key and every sort key — nothing extra to project or strip.
    #[test]
    fn test_include_without_fields_projects_everything() {
        let plan =
            plan_include(json!({ "sort": [{ "created_at": "desc" }] }), &limits()).expect("plan");
        let inc = &plan.includes[0];
        assert!(inc.projection().is_empty());
        assert!(inc.strip().is_empty());
        let keys = vec![SeaValue::from("u1")];
        let (sql, _v) = build_include_select(inc, &keys, SqlDialect::Sqlite);
        assert!(
            sql.starts_with(r#"SELECT * FROM (SELECT *, ROW_NUMBER() OVER ("#),
            "sql = {sql}"
        );
    }

    /// MySQL cannot take a `NULLS …` clause anywhere, including inside `OVER`.
    #[test]
    fn test_include_child_query_renders_for_mysql() {
        let plan = plan_include(
            json!({ "fields": ["total"], "sort": [{ "total": "asc" }], "limit": 2 }),
            &limits(),
        )
        .expect("plan");
        let keys = vec![SeaValue::from("u1")];
        let (sql, _v) = build_include_select(&plan.includes[0], &keys, SqlDialect::Mysql);
        assert!(!sql.contains("NULLS"), "sql = {sql}");
        assert!(
            sql.contains("ROW_NUMBER() OVER ( PARTITION BY `user_id` ORDER BY `total` ASC )"),
            "sql = {sql}"
        );
    }

    /// F27: without an order key "the first `n` children" is not a defined
    /// answer, so the envelope must name one.
    #[test]
    fn test_include_without_sort_is_rejected() {
        let err = plan_include(json!({ "limit": 5 }), &limits()).expect_err("no order key");
        assert!(matches!(err, QueryError::InvalidEnvelope(_)), "{err}");
        assert!(err.to_string().contains("include.orders"), "{err}");
        assert!(err.to_string().contains("sort"), "{err}");
    }

    #[test]
    fn test_m2m_include_rejected() {
        let err = crate::query::plan_sql(
            &json!({ "source": "users", "include": { "tags": { "sort": [{ "id": "asc" }] } } }),
            &serde_json::Map::new(),
            &rel_schema(),
            SqlDialect::Sqlite,
            &limits(),
        )
        .expect_err("m2m include not supported");
        assert!(matches!(err, QueryError::FeatureUnsupportedByTarget { .. }));
    }

    // ---- Write rendering (INSERT / UPDATE / DELETE / upsert) ----

    /// A config that lets every envelope shape through — the guards have
    /// their own tests in `write.rs`.
    fn permissive_writes() -> crate::config::WriteConfig {
        crate::config::WriteConfig {
            max_rows: 1000,
            allow_unfiltered: true,
        }
    }

    /// Resolve `input` (identity mode) and render the write SQL for `dialect`.
    fn write_sql(input: Json, dialect: SqlDialect) -> String {
        let resolved = crate::query::write::resolve_write(
            &input,
            &serde_json::Map::new(),
            &EntityRegistry::identity(),
            &permissive_writes(),
        )
        .expect("resolve_write should succeed");
        let (sql, _values) = render_write(&resolved, dialect).expect("render should succeed");
        sql
    }

    fn sqlite_w(input: Json) -> String {
        write_sql(input, SqlDialect::Sqlite)
    }

    #[test]
    fn test_insert_single() {
        let sql = sqlite_w(json!({
            "op": "insert",
            "target": "users",
            "values": { "id": "u1", "name": "Alice" }
        }));
        assert_eq!(sql, r#"INSERT INTO "users" ("id", "name") VALUES (?, ?)"#);
    }

    #[test]
    fn test_insert_bulk() {
        let sql = sqlite_w(json!({
            "op": "insert",
            "target": "users",
            "values": [ { "id": "u1", "name": "Alice" }, { "id": "u2", "name": "Bob" } ]
        }));
        assert_eq!(
            sql,
            r#"INSERT INTO "users" ("id", "name") VALUES (?, ?), (?, ?)"#
        );
    }

    #[test]
    fn test_insert_returning() {
        let sql = sqlite_w(json!({
            "op": "insert",
            "target": "users",
            "values": { "name": "Alice" },
            "returning": ["id", "name"]
        }));
        assert_eq!(
            sql,
            r#"INSERT INTO "users" ("name") VALUES (?) RETURNING "id", "name""#
        );
    }

    #[test]
    fn test_update_with_filter() {
        let sql = sqlite_w(json!({
            "op": "update",
            "target": "users",
            "set": { "status": "inactive" },
            "filter": { "==": [{ "field": "id" }, "u1"] }
        }));
        assert_eq!(sql, r#"UPDATE "users" SET "status" = ? WHERE "id" = ?"#);
    }

    #[test]
    fn test_update_relation_filter_reuses_query_dialect() {
        // The update WHERE is the query dialect's filter, EXISTS subquery and all.
        let input = json!({
            "op": "update",
            "target": "users",
            "set": { "flagged": true },
            "filter": { "some": [{ "field": "orders" }, { ">": [{ "field": "total" }, 100] }] }
        });
        let resolved = crate::query::write::resolve_write(
            &input,
            &serde_json::Map::new(),
            &rel_schema(),
            &permissive_writes(),
        )
        .expect("resolve");
        let (sql, _v) = render_write(&resolved, SqlDialect::Sqlite).expect("render");
        // In the bound-parameter path the constant `1` in `SELECT 1` binds as `?`.
        assert_eq!(
            sql,
            r#"UPDATE "users" SET "flagged" = ? WHERE EXISTS(SELECT ? FROM "orders" WHERE "orders"."user_id" = "users"."id" AND "total" > ?)"#
        );
    }

    #[test]
    fn test_delete_with_filter() {
        let sql = sqlite_w(json!({
            "op": "delete",
            "target": "sessions",
            "filter": { "<": [{ "field": "age" }, 0] }
        }));
        assert_eq!(sql, r#"DELETE FROM "sessions" WHERE "age" < ?"#);
    }

    #[test]
    fn test_upsert_do_update() {
        let sql = sqlite_w(json!({
            "op": "upsert",
            "target": "users",
            "values": { "email": "a@x.io", "name": "Ada" },
            "on_conflict": { "target": ["email"], "action": "update" }
        }));
        assert_eq!(
            sql,
            r#"INSERT INTO "users" ("email", "name") VALUES (?, ?) ON CONFLICT ("email") DO UPDATE SET "name" = "excluded"."name""#
        );
    }

    #[test]
    fn test_upsert_do_nothing() {
        let sql = sqlite_w(json!({
            "op": "upsert",
            "target": "users",
            "values": { "email": "a@x.io", "name": "Ada" },
            "on_conflict": { "target": ["email"], "action": "nothing" }
        }));
        assert_eq!(
            sql,
            r#"INSERT INTO "users" ("email", "name") VALUES (?, ?) ON CONFLICT ("email") DO NOTHING"#
        );
    }

    #[test]
    fn test_returning_on_mysql_rejected() {
        let input = json!({
            "op": "insert",
            "target": "users",
            "values": { "name": "Ada" },
            "returning": ["id"]
        });
        let resolved = crate::query::write::resolve_write(
            &input,
            &serde_json::Map::new(),
            &EntityRegistry::identity(),
            &permissive_writes(),
        )
        .expect("resolve");
        let err = render_write(&resolved, SqlDialect::Mysql).expect_err("no RETURNING on MySQL");
        assert!(matches!(
            err,
            crate::query::write::WriteError::Query(QueryError::FeatureUnsupportedByTarget { .. })
        ));
    }
}

#[cfg(test)]
mod prop_tests {
    use super::*;
    use crate::config::WriteConfig;
    use crate::query::schema::EntityRegistry;
    use crate::query::write::resolve_write;
    use proptest::prelude::*;
    use serde_json::json;

    fn permissive_writes() -> WriteConfig {
        WriteConfig {
            max_rows: 1000,
            allow_unfiltered: true,
        }
    }

    /// Render an update whose `set` values and `filter` comparison all carry
    /// `value`, plus an insert row carrying it — the three channels a user
    /// scalar travels through. Returns (sql texts, all bind values).
    fn rendered(value: &str, dialect: SqlDialect) -> (Vec<String>, Vec<sea_query::Value>) {
        let update = serde_json::json!({
            "op": "update",
            "target": "users",
            "set": { "name": value },
            "filter": { "==": [{"field": "name"}, value] }
        });
        let insert = serde_json::json!({
            "op": "insert",
            "target": "users",
            "values": { "name": value }
        });
        let mut sqls = Vec::new();
        let mut binds = Vec::new();
        for input in [update, insert] {
            let resolved = resolve_write(
                &input,
                &serde_json::Map::new(),
                &EntityRegistry::identity(),
                &permissive_writes(),
            )
            .expect("resolve");
            let (sql, values) = render_write(&resolved, dialect).expect("render");
            sqls.push(sql);
            binds.extend(values.0.0);
        }
        (sqls, binds)
    }

    /// Identifier candidates: arbitrary unicode, strings built around the
    /// quoting/escaping metacharacters, and plain well-formed names so the
    /// accept path stays exercised too.
    fn arb_ident() -> impl Strategy<Value = String> {
        prop_oneof![
            ".{0,12}",
            r#"[a-z]{0,3}["'`\\$.][a-z]{0,3}["'`\\$.]?"#,
            "[A-Za-z_][A-Za-z0-9_]{0,8}",
        ]
    }

    /// Whether `ident` could interfere with identifier quoting or change
    /// meaning on any backend — exactly the set `validate_identifier` rejects.
    fn is_hostile(ident: &str) -> bool {
        ident.is_empty()
            || ident.starts_with('$')
            || ident.contains('.')
            || ident
                .chars()
                .any(|c| c.is_control() || matches!(c, '"' | '\'' | '`' | '\\'))
    }

    /// The dialect's quoted form of an accepted identifier.
    fn quoted(ident: &str, dialect: SqlDialect) -> String {
        match dialect {
            SqlDialect::Mysql => format!("`{ident}`"),
            _ => format!("\"{ident}\""),
        }
    }

    proptest! {
        /// The injection-resistance invariant SECURITY.md puts in scope,
        /// stated as value-independence: the rendered SQL text is a function
        /// of the envelope SHAPE only. If any user scalar — quotes, SQL
        /// fragments, unicode, control bytes — ever leaked into the SQL
        /// string instead of the binds, the text would differ from the
        /// baseline rendering of the same shape.
        #[test]
        fn sql_text_is_independent_of_user_values(value in ".*") {
            for dialect in [SqlDialect::Postgres, SqlDialect::Sqlite, SqlDialect::Mysql] {
                let (sqls, binds) = rendered(&value, dialect);
                let (baseline_sqls, _) = rendered("baseline", dialect);
                prop_assert_eq!(&sqls, &baseline_sqls, "dialect {:?}", dialect);
                // And the value must actually travel — via the binds.
                prop_assert!(
                    binds.iter().any(
                        |v| matches!(v, sea_query::Value::String(Some(s)) if s.as_str() == value)
                    ),
                    "value must appear among the binds for {:?}",
                    dialect
                );
            }
        }

        /// F25: identifiers cannot be bound parameters, so their safety must
        /// live at the resolution boundary (`validate_identifier`), not in
        /// sea-query's `Iden::quoted`. For an arbitrary identifier fed through
        /// the WRITE path's channels — insert's `target` and an inserted
        /// column, update's `set` column and a `returning` name — either
        /// resolution rejects it, or it was benign and renders inside dialect
        /// quotes it cannot close. (`on_conflict.target` is not fuzzed here;
        /// it crosses the same `resolve_write_column` boundary as `set`.)
        #[test]
        fn write_identifiers_are_rejected_or_safely_quoted(ident in arb_ident()) {
            let mut values = serde_json::Map::new();
            values.insert(ident.clone(), json!(1));
            // insert: the identifier as `target` and as an inserted column.
            let insert = json!({
                "op": "insert",
                "target": ident.clone(),
                "values": values.clone(),
            });
            // update: the identifier as a `set` column and a `returning` name
            // (`all` acknowledges the deliberately unfiltered mutation).
            let update = json!({
                "op": "update",
                "target": "users",
                "set": values,
                "returning": [ident.clone()],
                "all": true,
            });
            for input in [insert, update] {
                let resolved = resolve_write(
                    &input,
                    &serde_json::Map::new(),
                    &EntityRegistry::identity(),
                    &permissive_writes(),
                );
                match resolved {
                    Err(_) => {} // boundary rejection — the safe outcome
                    Ok(w) => {
                        prop_assert!(
                            !is_hostile(&ident),
                            "resolve_write accepted a hostile identifier {ident:?}"
                        );
                        for dialect in [SqlDialect::Postgres, SqlDialect::Sqlite, SqlDialect::Mysql] {
                            // MySQL refuses RETURNING by feature, not by
                            // identifier — drop it there so the SET-clause
                            // channel still renders.
                            let mut w = w.clone();
                            if dialect == SqlDialect::Mysql {
                                match &mut w {
                                    ResolvedWrite::Insert { returning, .. }
                                    | ResolvedWrite::Update { returning, .. }
                                    | ResolvedWrite::Delete { returning, .. }
                                    | ResolvedWrite::Upsert { returning, .. } => returning.clear(),
                                }
                            }
                            let (sql, _) = render_write(&w, dialect).expect("render");
                            prop_assert!(
                                sql.contains(&quoted(&ident, dialect)),
                                "identifier {ident:?} must render quoted in {sql:?}"
                            );
                        }
                    }
                }
            }
        }

        /// The same invariant for the READ path's identifier channels:
        /// `source`, `fields`, `sort`, and a `filter` field reference.
        #[test]
        fn query_identifiers_are_rejected_or_safely_quoted(ident in arb_ident()) {
            let mut sort_key = serde_json::Map::new();
            sort_key.insert(ident.clone(), json!("asc"));
            let query = json!({
                "source": ident.clone(),
                "fields": [ident.clone()],
                "sort": [sort_key],
                "filter": { "==": [{ "field": ident.clone() }, 1] },
            });
            for dialect in [SqlDialect::Postgres, SqlDialect::Sqlite, SqlDialect::Mysql] {
                match crate::query::plan_sql(
                    &query,
                    &serde_json::Map::new(),
                    &EntityRegistry::identity(),
                    dialect,
                    &QueryConfig::default(),
                )
                .map(|plan| plan.main)
                {
                    Err(_) => {} // boundary rejection — the safe outcome
                    Ok(stmt) => {
                        prop_assert!(
                            !is_hostile(&ident),
                            "plan_sql accepted a hostile identifier {ident:?}"
                        );
                        let sql = match dialect {
                            SqlDialect::Sqlite => stmt.to_string(SqliteQueryBuilder),
                            SqlDialect::Postgres => stmt.to_string(PostgresQueryBuilder),
                            SqlDialect::Mysql => stmt.to_string(MysqlQueryBuilder),
                        };
                        prop_assert!(
                            sql.contains(&quoted(&ident, dialect)),
                            "identifier {ident:?} must render quoted in {sql:?}"
                        );
                    }
                }
            }
        }
    }
}