gbif-name-parser 0.3.0

A faithful Rust port of the scientific name parser from the Global Biodiversity Information Facility (GBIF): parses biological scientific names into their structured components (genus, epithets, authorship, rank, nomenclatural code).
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
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
1566
1567
1568
1569
1570
1571
1572
1573
1574
1575
1576
1577
1578
1579
1580
1581
1582
1583
1584
1585
1586
1587
1588
1589
1590
1591
1592
1593
1594
1595
1596
1597
1598
1599
1600
1601
1602
1603
1604
1605
1606
1607
1608
1609
1610
1611
1612
1613
1614
1615
1616
1617
1618
1619
1620
1621
1622
1623
1624
1625
1626
1627
1628
1629
1630
1631
1632
1633
1634
1635
1636
1637
1638
1639
1640
1641
1642
1643
1644
1645
1646
1647
1648
1649
1650
1651
1652
1653
1654
1655
1656
1657
1658
1659
1660
1661
1662
1663
1664
1665
1666
1667
1668
1669
1670
1671
1672
1673
1674
1675
1676
1677
1678
1679
1680
1681
1682
1683
1684
1685
1686
1687
1688
1689
1690
1691
1692
1693
1694
1695
1696
1697
1698
1699
1700
1701
1702
1703
1704
1705
1706
// SPDX-License-Identifier: Apache-2.0

//! Java `org.gbif.nameparser.pipeline.NameTokens` (712 lines) — walks the token range
//! `[0, boundary)` (the "name" span `AuthorshipSplit::find_boundary`, Phase 1 Slice 3 Task
//! 2, has already separated from the trailing authorship) and classifies it into the
//! structural slots of a scientific name: uninomial / genus / subgenus / infrageneric /
//! specific / infraspecific, plus the hybrid-marker (`notho`), open-nomenclature
//! (`epithetQualifier`), and aggregate/indet side signals. Entirely regex-free: a
//! hand-written token-index state machine, mirroring `AuthorshipSplit`'s own style.
//!
//! Three load-bearing quirks, preserved verbatim (see the plan's Global Constraints):
//!
//! 1. **`boundary == 0`** short-circuits immediately: `ctx.name.state = PARTIAL`,
//!    `ctx.name.unparsed = ctx.original`, return — no other field is touched.
//! 2. **Notho parts add up.** A `HYBRID_MARK` token anywhere in the name adds its part to
//!    the notho set (`ParsedName::add_notho`, additive, deduped), and so does a
//!    notho-prefixed infraspecific rank marker like "nothosubsp."/"nvar." after the loop:
//!    `Aconitum ×teppneri nothosubsp. goetzii` is notho SPECIFIC and INFRASPECIFIC.
//!    Deliberate divergence from Java 4.2.0, whose `setNotho(INFRASPECIFIC)` replaced the
//!    whole set and erased the hybrid sign before the genus or species.
//! 3. **[`skip_paren_author_block`] here is NameTokens's OWN copy**, textually different
//!    from `authorship_split::skip_paren_author_block` (Task 2) of the same name: that one
//!    gates its match on `has_epithet_after_marker` before returning the marker index; this
//!    one returns as soon as it finds ANY infraspecific-marker-shaped word, with no such
//!    follow-up check. The two are intentionally NOT unified — this matches the Java
//!    source, which keeps two separate private methods of the same name in two different
//!    classes (see `authorship_split`'s own module doc for the full rationale).
//!
//! **Adapting Java's `boundary` parameter.** Several Java helpers
//! (`hasInfraspecificEpithetAfter`, `skipParenAuthorBlock`, and the
//! `AuthorshipSplit.midNameAuthorEnd` bridge) take an explicit `int boundary` third
//! argument distinct from `ts.size()` (`ts` is always the FULL token list; `boundary` is
//! the name/author split Task 2 already computed). Rather than threading a redundant
//! `boundary`/`to` parameter through every helper, [`classify`] slices once, up front —
//! `let ts: &[Token] = &ctx.tokens[..boundary];` — and every helper below (including the
//! two `authorship_split` bridges, [`authorship_split::mid_name_author_end`] and
//! [`authorship_split::is_apostrophe_particle`]) is then called with THIS cropped slice,
//! so a plain `tokens.len()` inside each helper reproduces Java's `boundary` bound exactly
//! (slicing from the front never changes any in-range index, so every recorded token index,
//! e.g. `ctx.mid_author_from`/`ctx.mid_author_to`, stays numerically identical to the
//! corresponding index into the untruncated `ctx.tokens`).
//!
//! Ported helpers not among the Java source's own 8 private statics (`render_author_span`,
//! `skip_paren_author_block`, `recover_case`, `is_all_upper_letters`,
//! `has_infraspecific_epithet_after`, `is_strain_code`, `is_all_letter_case`, `strip_dot`):
//! `starts_upper`/`starts_lower`/`starts_digit_epithet` duplicate `authorship_split.rs`'s
//! own identically-named, identically-behaved private bridges (`Token` doesn't carry these
//! as methods yet — see that module's doc comment for why they're local free functions
//! rather than additions to `token.rs`). `Rank.isInfraspecific()`/`Rank.getMarker()` are
//! called directly as [`crate::model::Rank::is_infraspecific`]/
//! [`crate::model::Rank::marker`] — Phase 1 Slice 4 Task 1 made the full `Rank` model the
//! single source of truth, replacing this file's former ad-hoc `rank_is_infraspecific`/
//! `rank_marker` free functions.

use std::sync::LazyLock;

use regex::Regex;

use crate::model::{warnings, NamePart, NameType, Rank, State};
use crate::pipeline::authorship_split;
use crate::pipeline::rank_markers;
use crate::pipeline::stripandstash;
use crate::pipeline::ParseContext;
use crate::token::{self, Token, TokenKind};
use crate::unicode::java_trim;

/// Java `NameTokens.AGG_HYPHEN_SUFFIXES` (`NameTokens.java:21`).
const AGG_HYPHEN_SUFFIXES: [&str; 3] = ["-group", "-complex", "-aggregate"];

/// An authorship or a source citation with its year, as the whole tail after an indet marker:
/// "Forster, 1968", "N. Bruce, 2008", "of Zhuravlev & Gravestock 1994", "[of Sokolov et al.,
/// 2025]". Capitalised words, connectors and particles only, then the year — a designation
/// ("RMCC TR1811", "Olinda (R.Coveny 6616)", "Bunney Road") never ends in a separate year, and a
/// lone letter ("sp. A Soto-Adames, 2010") is a designation, not an initial.
static AUTHOR_YEAR_TAIL: LazyLock<Regex> = LazyLock::new(|| {
    Regex::new(
        r"^\[?(?:of\s+)?(?:\p{Lu}\.|\p{Lu}[\p{L}'\-]+\.?)(?:\s+(?:\p{Lu}[\p{L}'.\-]*|&|et|and|al\.?|de|van|von|der|den|du|le|la|da|del|di))*,?\s+\d{4}[a-z]?\]?$",
    )
    .unwrap()
});

/// Java `NameTokens.classify(ParseContext, int)` (`NameTokens.java:25-536`). Walks
/// `ctx.tokens[0, boundary)`, classifying it into the structural name-part fields on
/// `ctx.name`, plus `ctx.mid_author_from`/`ctx.mid_author_to` (the species-level author
/// span for a later autonym check) and `ctx.aggregate`. See this module's own doc comment
/// for the 3 load-bearing quirks preserved verbatim.
pub(crate) fn classify(ctx: &mut ParseContext, boundary: usize) {
    if boundary == 0 {
        ctx.name.state = State::Partial;
        ctx.name.unparsed = Some(ctx.original.clone());
        return;
    }

    // See this module's doc comment: slicing once up front lets every helper below use
    // `tokens.len()` as Java's separately-threaded `boundary` parameter.
    let ts: &[Token] = &ctx.tokens[..boundary];

    // True when AuthorshipSplit left NO authorship tail — the name section spans the whole
    // input. Gates the trailing-tag phrase capture below: only flatten a supraspecific indet
    // tail into a phrase when find_boundary already decided the whole string is the name (it
    // returns `n` for a yearless "Genus sp. <tag>"); when it kept an author+year tail out of
    // the name section (`boundary < len`, e.g. "Bacterium sp. (serotype) aboney Dräger 1951"),
    // the in-name tokens are structural (subgenus, epithet) and must parse normally, not flatten.
    let name_section_covers_all = boundary == ctx.tokens.len();

    let mut genus: Option<String> = None;
    let mut subgenus: Option<String> = None;
    let mut infrageneric: Option<Rank> = None;
    let mut infragen_epithet: Option<String> = None;
    let mut lower_epithets: Vec<String> = Vec::with_capacity(4);
    let mut marker_idx_in_epithets: i32 = -1;
    let mut inline_rank: Option<Rank> = None;
    let mut inline_rank_notho = false;
    let mut indet = false;
    // a lone-letter species designation (`Collettea a`), see step 8 below
    let mut letter_designation = false;
    // A bare supraspecific indet ("Genus sp." with no distinguishing tail) — stays flagged
    // INDETERMINED even though its phrase now carries the verbatim marker (see the top-check
    // in the indet branch and the warning block near the end of this fn).
    let mut indet_bare = false;
    let mut cf_aff_qualifier: Option<String> = None;
    // A cf./aff./nr./near with no epithet after it opens the informal phrase (see below).
    let mut qualifier_phrase = false;
    // how many epithets came before the qualifier: it qualifies the next one, or the last if none
    // follows ("Arctostaphylos preglauca cf.")
    let mut qualifier_at = 0usize;
    // Tracks the most recently skipped mid-name author span so that, when a second
    // infraspecific marker overrides the first, we can describe the dropped middle
    // classification ("Intermediate classification removed: subsp.X Author") in a warning.
    let mut pending_mid_name_author: Option<String> = None;

    let mut i = 0usize;
    while i < ts.len() {
        let t = &ts[i];

        if t.kind == TokenKind::HybridMark {
            if genus.is_some() && lower_epithets.is_empty() {
                ctx.name.add_notho(NamePart::Specific);
            } else if genus.is_none() {
                ctx.name.add_notho(NamePart::Generic);
            } else {
                ctx.name.add_notho(NamePart::Infraspecific);
            }
            i += 1;
            continue;
        }

        // "(Word)" -> subgenus, but only before any species epithet. After a species
        // epithet "(Klatt)" is a basionym author span (e.g. the autonym "Trimezia
        // spathata (Klatt) Baker subsp. spathata"), handled by the
        // skip_paren_author_block branch below.
        if t.kind == TokenKind::OpenParen
            && lower_epithets.is_empty()
            && subgenus.is_none()
            && i + 2 < ts.len()
            && ts[i + 1].kind == TokenKind::Word
            && (starts_upper(&ts[i + 1]) || starts_lower(&ts[i + 1]))
            && ts[i + 2].kind == TokenKind::CloseParen
        {
            let raw = ts[i + 1].text.clone();
            // A lower-case parenthesised subgenus ("(acanthoderes)") is malformed —
            // capitalise it to the conventional form and flag the name doubtful.
            let sub = if starts_lower(&ts[i + 1]) {
                ctx.name.doubtful = true;
                let mut chars = raw.chars();
                match chars.next() {
                    Some(c) => {
                        let mut s: String = c.to_uppercase().collect();
                        s.push_str(chars.as_str());
                        s
                    }
                    None => raw,
                }
            } else {
                raw
            };
            subgenus = Some(sub);
            i += 3;
            continue;
        }
        // Abbreviated subgenus: "(Tin.)" / "(G.)" — Title-cased word + DOT inside parens.
        // Only fires before any species epithet, otherwise "(Aubl.)" after a species is a
        // basionym author span and the next block (skip_paren_author_block) handles it.
        if t.kind == TokenKind::OpenParen
            && lower_epithets.is_empty()
            && subgenus.is_none()
            && i + 3 < ts.len()
            && ts[i + 1].kind == TokenKind::Word
            && starts_upper(&ts[i + 1])
            && ts[i + 2].kind == TokenKind::Dot
            && ts[i + 3].kind == TokenKind::CloseParen
        {
            subgenus = Some(format!("{}.", ts[i + 1].text));
            ctx.name.add_warning(warnings::ABBREVIATED_SUBGENUS);
            i += 4;
            continue;
        }
        // After the species epithet, "(BasionymAuth) CombAuth …" sits between species and
        // an infraspecific rank marker. Skip the parens + comb-author span so the rank
        // marker that follows can be classified normally.
        if t.kind == TokenKind::OpenParen && genus.is_some() && !lower_epithets.is_empty() {
            if let Some(after) = skip_paren_author_block(ts, i) {
                // Record the species-level author span [i, after) so the pipeline can
                // attach it to an autonym (ICN Art. 22.1/26.1). Only the first span, right
                // after the species epithet and before any infraspecific marker, is the
                // species author.
                if ctx.mid_author_from < 0
                    && lower_epithets.len() == 1
                    && marker_idx_in_epithets < 0
                {
                    ctx.mid_author_from = i as i32;
                    ctx.mid_author_to = after as i32;
                }
                i = after;
                continue;
            }
        }
        // Abbreviated genus: Title-cased word of 1-4 chars then DOT, only when no genus
        // yet. The single-letter form ("M. alpium") is unambiguous; 2-4 letter forms
        // ("Mo. alpium", "Phl. guttella", "Pseud. dendrobii") are recognised when the next
        // non-DOT token is a lowercase epithet (so an authorship-sequence like
        // "Mo.J.Wong, 1990" doesn't trip).
        if genus.is_none()
            && t.kind == TokenKind::Word
            && (1..=4).contains(&t.text.chars().count())
            && starts_upper(t)
            && i + 1 < ts.len()
            && ts[i + 1].kind == TokenKind::Dot
        {
            let next_is_lower_epithet =
                i + 2 < ts.len() && ts[i + 2].kind == TokenKind::Word && starts_lower(&ts[i + 2]);
            if t.text.chars().count() == 1 || next_is_lower_epithet {
                genus = Some(format!("{}.", t.text));
                ctx.name.type_ = NameType::Informal;
                ctx.name.add_warning(warnings::ABBREVIATED_GENUS);
                i += 2;
                continue;
            }
        }
        // Missing-genus placeholder: "?" as the genus stand-in. Only at the very start.
        if genus.is_none() && t.kind == TokenKind::Other && t.text == "?" {
            genus = Some("?".to_string());
            i += 1;
            continue;
        }
        // Open-nomenclature doubtful-identification marker: "?" between epithets, like
        // "Ferganoconcha? oblonga" or "Buteo borealis ? ventralis". Treat exactly like
        // cf./aff.: skip the token, attach the qualifier to the next epithet (specific when
        // no species yet, infraspecific when one exists), set type=INFORMAL, doubtful, and
        // emit QUESTION_MARKS_REMOVED.
        if genus.is_some()
            && cf_aff_qualifier.is_none()
            && t.kind == TokenKind::Other
            && t.text == "?"
            && lower_epithets.len() < 2
        {
            qualifier_at = lower_epithets.len();
            cf_aff_qualifier = Some("?".to_string());
            ctx.name.type_ = NameType::Informal;
            ctx.name.doubtful = true;
            ctx.name.add_warning(warnings::QUESTION_MARKS_REMOVED);
            i += 1;
            continue;
        }

        if t.kind == TokenKind::Word {
            // Mid-name author span (uppercase Author abbreviations followed by a rank
            // marker, or particle-starting authors like "d'Urv. subsp.") — silently
            // skipped so that downstream classification operates only on the structural
            // tokens.
            // an epithet capitalised in the old style ("Delias Abnormis var. euryxantha")
            if genus.is_some()
                && lower_epithets.is_empty()
                && authorship_split::is_capitalised_old_epithet(&ctx.tokens, i, genus.as_deref())
            {
                lower_epithets.push(t.text.to_lowercase());
                i += 1;
                continue;
            }
            let can_start_author = starts_upper(t)
                || (starts_lower(t)
                    && (token::is_particle(&t.text)
                        || authorship_split::is_apostrophe_particle(&t.text)));
            if genus.is_some() && can_start_author {
                if let Some(after) = authorship_split::mid_name_author_end(ts, i) {
                    // Record the species-level author span so the pipeline can attach it
                    // to an autonym (ICN Art. 22.1/26.1): only the first span, sitting
                    // directly after the species epithet and before any infraspecific
                    // marker.
                    if ctx.mid_author_from < 0
                        && lower_epithets.len() == 1
                        && marker_idx_in_epithets < 0
                    {
                        ctx.mid_author_from = i as i32;
                        ctx.mid_author_to = after as i32;
                    }
                    pending_mid_name_author = Some(render_author_span(ts, i, after));
                    i = after;
                    continue;
                }
            }
            if genus.is_none() {
                // For a regular capital-letter genus we keep the input as-is; for
                // all-caps or all-lowercase we recover the canonical form.
                let mut g = recover_case(&t.text, true);
                i += 1;
                // pull abbreviated-genus dot through
                if g.chars().count() == 2 && g.ends_with('.') {
                    // Dead in practice: a WORD token's text can never itself contain a
                    // literal '.' (the tokenizer always splits it into its own DOT
                    // token), so a freshly case-recovered genus can never already end in
                    // a dot. Ported verbatim for branch-for-branch fidelity with
                    // `NameTokens.java:187-188` (its own "already a dot — nothing to do"
                    // comment), not because it's reachable.
                } else if g.chars().count() == 1 && i < ts.len() && ts[i].kind == TokenKind::Dot {
                    g.push('.');
                    ctx.name.type_ = NameType::Informal;
                    i += 1;
                }
                genus = Some(g);
                continue;
            }
            // upper-case epithets (e.g. all-caps "ELEVATA") — treat as lower-case epithets
            // after recovery. A diacritic is not a discriminator: "ELEVÄTA" is lower-cased
            // to an epithet exactly like its ASCII twin "ELEVATA". Only an abbreviation dot
            // ("ELEV." -> "ELEV" + ".") still routes the token to author-recovery, since
            // that shape is an abbreviated author surname.
            if starts_upper(t)
                && genus.is_some()
                && t.text.chars().count() > 1
                && is_all_upper_letters(&t.text)
            {
                let is_abbrev = i + 1 < ts.len() && ts[i + 1].kind == TokenKind::Dot;
                if !is_abbrev {
                    lower_epithets.push(t.text.to_lowercase());
                    i += 1;
                    continue;
                }
                // Otherwise fall through to author-recovery handling below.
            }
            if genus.is_some() && starts_digit_epithet(t) {
                // lower-case like the ordinary-epithet branch below ("11-Punctata" ->
                // "11-punctata")
                lower_epithets.push(t.text.to_lowercase());
                i += 1;
                continue;
            }
            if starts_lower(t) {
                let w = strip_dot(&t.text);
                // 0. Single Greek letter or short ASCII letter between two lowercase
                // epithets ("Agaricus collinitus β mucosus", "Cyphelium disseminatum c
                // subsessile") is an informal infra-rank marker, not an epithet. Skip
                // silently when there's already a species epithet AND another lower-case
                // epithet follows — otherwise the letter IS the epithet (e.g. "var. β").
                if w.chars().count() == 1 && !lower_epithets.is_empty() {
                    let cp = w.chars().next().unwrap() as u32;
                    let is_greek = (0x03B1..=0x03C9).contains(&cp);
                    let is_fungi_ascii =
                        cp == 0x237A || (u32::from(b'a')..=u32::from(b'g')).contains(&cp);
                    if is_greek || is_fungi_ascii {
                        let peek = i + 1;
                        if peek < ts.len()
                            && ts[peek].kind == TokenKind::Word
                            && starts_lower(&ts[peek])
                        {
                            i += 1;
                            continue;
                        }
                    }
                }
                // 1a. "ex gr." (ex grege, of the species group of): the paleontologists' qualifier
                if w == "ex" && is_ex_grege(ts, i) && lower_epithets.len() < 2 {
                    qualifier_at = lower_epithets.len();
                    cf_aff_qualifier = Some("ex gr.".to_string());
                    ctx.name.type_ = NameType::Informal;
                    i += 2;
                    if i < ts.len() && ts[i].kind == TokenKind::Dot {
                        i += 1;
                    }
                    continue;
                }
                // 1. cf./aff./near/nr. open-nomenclature qualifier
                if (w.eq_ignore_ascii_case("cf")
                    || w.eq_ignore_ascii_case("aff")
                    || w.eq_ignore_ascii_case("nr")
                    || w.eq_ignore_ascii_case("near"))
                    && lower_epithets.len() < 2
                {
                    // cf./aff. are abbreviations, rendered with a trailing dot ("cf."/"aff."); "near"
                    // is a full English word synonymous with aff., so it is stored verbatim, no dot.
                    qualifier_at = lower_epithets.len();
                    cf_aff_qualifier = Some(if w.eq_ignore_ascii_case("near") {
                        w.to_string()
                    } else {
                        format!("{w}.")
                    });
                    ctx.name.type_ = NameType::Informal;
                    let marker_start = t.start;
                    i += 1;
                    if i < ts.len() && ts[i].kind == TokenKind::Dot {
                        i += 1;
                    }
                    // With no epithet before or after it, the qualifier compares an unnamed
                    // species to the taxon, like "sp.": it opens that indetermined name's phrase,
                    // as written, running to the end ("Formicidae cf.", "Acroceridae aff. Terphis
                    // sp. SLW-2002"). Java kept a bare one as the qualifier of a species epithet
                    // the name does not have. A repeated genus ("Sorex cf. S. shinto") leads to
                    // the epithet instead.
                    let another_taxon = ts.get(i).is_some_and(|t| {
                        t.kind == TokenKind::Word
                            && starts_upper(t)
                            && authorship_split::skip_repeated_genus(ts, i, genus.as_deref())
                                .is_none()
                    });
                    if lower_epithets.is_empty()
                        && genus.is_some()
                        && ctx.name.phrase.is_none()
                        && (i >= ts.len() || another_taxon)
                    {
                        indet_bare = i >= ts.len();
                        ctx.name.phrase =
                            Some(ctx.working[marker_start..ts[ts.len() - 1].end].to_string());
                        qualifier_phrase = true;
                        indet = true;
                        i = ts.len();
                    }
                    continue;
                }
                // 2. indet markers — sp./spec./indet
                // A number right after the marker (past an optional dot) is a placeholder tag:
                // "Genus sp. 1" (monomial) or "Genus epithet species 12" (after a binomial). Preserve
                // it as an informal phrase rather than reading "species"/"sp" as a (blacklisted)
                // infraspecific epithet and dropping the number — so `species N` round-trips verbatim.
                let number_follows_marker = {
                    let mut j = i + 1;
                    if j < ts.len() && ts[j].kind == TokenKind::Dot {
                        j += 1;
                    }
                    // a strain code, a digit-led word or a final capital letter is a designation
                    // too: "Euxoa idahoensis sp. 1clay", "Abies alba sp. JGP0404", "… sp. E"
                    let word = |k: usize| ts.get(k).filter(|t| t.kind == TokenKind::Word);
                    j < ts.len()
                        && (ts[j].kind == TokenKind::Number
                            || word(j).is_some_and(|t| {
                                t.text.starts_with(|c: char| c.is_ascii_digit())
                                    || (starts_upper(t)
                                        && (is_strain_code(&t.text)
                                            || ts.get(j + 1).is_some_and(|nx| {
                                                nx.kind == TokenKind::Number && j + 2 == ts.len()
                                            })
                                            || (t.text.chars().count() == 1 && j + 1 == ts.len())))
                            }))
                };
                // …but `spec` is also a genuine published epithet ("Hemicloeina spec Platnick,
                // 2002", "Lampona spec Platnick, 2000", "Gobiosoma spec (Ginsburg, 1939)"; COL
                // carries nine). Two signals together separate those from a provisional
                // `Genus spec.`: the word carries NO abbreviation dot, and an authorship follows
                // it. `i + 1 == ts.len()` covers both — a dot would sit in the name section as its
                // own `Dot` token right after the marker (the tokenizer never folds it into the
                // word), and the marker can only be the section's last token when
                // AuthorshipSplit kept a tail out of it, which `!name_section_covers_all`
                // re-asserts. Only `spec` is rescued: a bare `sp` is overwhelmingly a dot-less
                // `sp.`, not an epithet. Mirrors AuthorshipSplit's own guard of the same name. A
                // separately supplied authorship is the same signal (`Hemicloeina spec` +
                // `Platnick, 2002`, how ChecklistBank passes it).
                let separate_authorship = ctx.authorship_input.as_deref().is_some_and(|a| {
                    !java_trim(stripandstash::without_authorship_placeholder(a)).is_empty()
                });
                let is_published_spec_epithet = w.eq_ignore_ascii_case("spec")
                    && genus.is_some()
                    && lower_epithets.is_empty()
                    && (!name_section_covers_all || separate_authorship)
                    && i + 1 == ts.len();
                if (w.eq_ignore_ascii_case("sp")
                    || w.eq_ignore_ascii_case("spp")
                    || w.eq_ignore_ascii_case("spec")
                    || w.eq_ignore_ascii_case("species")
                    || w.eq_ignore_ascii_case("indet"))
                    && !is_published_spec_epithet
                    && (lower_epithets.is_empty()
                        || marker_idx_in_epithets >= 0
                        || number_follows_marker)
                {
                    let marker_start = ts[i].start;
                    indet = true;
                    ctx.name.type_ = NameType::Informal;
                    i += 1;
                    if i < ts.len() && ts[i].kind == TokenKind::Dot {
                        i += 1;
                    }
                    // 5.0.0 Informal-phrase contract: a supraspecific indet (no species epithet yet)
                    // keeps the WHOLE verbatim tail FROM the marker as its phrase — marker included,
                    // original spelling preserved — so it round-trips exactly and the formatter's
                    // `phrase_leads_with_species_marker` guard emits it as-is instead of re-synthesising
                    // "sp." from the rank. With a trailing designation that is "sp. RMCC TR1811" /
                    // "species 1" / "spec. 3"; a bare "Genus sp." captures just the marker itself
                    // ("sp."), so every Informal name's phrase is uniformly the printable text after
                    // the anchor taxon (no None special-case, no synth-from-rank). `ts[ts.len() - 1]`
                    // is the last consumed token — the trailing tag when present, else the marker/dot.
                    // The marker-stripping sub-branches below now serve only the binomial indet case (a
                    // species epithet is already present, so it stays Parsed, not Informal).
                    // Only the species-level markers (sp/spec/species) are round-tripped by the
                    // formatter's `phrase_leads_with_species_marker` guard; a fully-generic "indet"
                    // is NOT, so keeping it in the phrase would double under the synthesised "sp."
                    // ("Aster indet." -> "Aster sp. indet."). Restrict the marker capture to
                    // sp/spec/species; "indet" falls through and renders via the synthesised rank.
                    let is_species_marker = w.eq_ignore_ascii_case("sp")
                        || w.eq_ignore_ascii_case("spp")
                        || w.eq_ignore_ascii_case("spec")
                        || w.eq_ignore_ascii_case("species");
                    // A cultivar epithet (extracted upstream from "Genus sp. cv. 'Name'") is the
                    // operative designation, not the "sp." — so it is NOT an informal indet and the
                    // marker must not be captured as a phrase (assemble step 17 clears INFORMAL for
                    // it); fall through and leave the phrase to the cultivar rendering. Likewise, an
                    // already-set phrase means an upstream step (the voucher `stash_phrase_name`,
                    // which rewrites working to a bare "Genus sp." while stashing the full "sp. <voucher>"
                    // on the phrase) owns it — don't clobber it with the bare marker here.
                    if lower_epithets.is_empty()
                        && ctx.name.cultivar_epithet.is_none()
                        && ctx.name.phrase.is_none()
                        && is_species_marker
                    {
                        // A distinguishing tail after the marker (a specimen tag, number, voucher)
                        // makes the name determinable; a bare "Genus sp." does not, so it stays
                        // flagged INDETERMINED below even though its phrase carries the marker. Nor
                        // does an authorship or source citation with its year ("Cantuaria sp.
                        // Forster, 1968", "[of Sokolov et al., 2025]"), as when it comes separately.
                        let tail = &ctx.working
                            [ts.get(i).map_or(marker_start, |t| t.start)..ts[ts.len() - 1].end];
                        indet_bare = i >= ts.len() || AUTHOR_YEAR_TAIL.is_match(tail);
                        ctx.name.phrase =
                            Some(ctx.working[marker_start..ts[ts.len() - 1].end].to_string());
                        i = ts.len();
                        continue;
                    }
                    // A number immediately following the indet marker becomes the
                    // informal phrase. When the source spelled out the marker as the
                    // full word "species" we keep it verbatim in the phrase ("Allium
                    // species 1" -> phrase "species 1") rather than collapsing it to the
                    // synthetic "sp." marker; the formatter then renders the phrase
                    // as-is. Abbreviated "sp."/"spec." keep the number-only phrase on a
                    // genus, where the formatter supplies the marker ("Allium sp. 1"), but not
                    // after a species epithet, where nothing would: "Dichanthelium
                    // chrysopsidifolium sp. 12" keeps "sp. 12" (Java rendered "… 12").
                    if i < ts.len() && ts[i].kind == TokenKind::Number {
                        let number = ts[i].text.clone();
                        let number_end = ts[i].end;
                        i += 1;
                        // Rule: anything after "(sp|spec|species) N" belongs to the phrase — once a
                        // phrase starts it runs to the end of the input. So when tokens follow the
                        // number, capture the whole VERBATIM tail from the marker ("Dichanthelium
                        // species 12 (=chrysopsidifolium)" -> phrase "species 12 (=chrysopsidifolium)")
                        // rather than leaving "(=chrysopsidifolium)" to be misread as a subgenus /
                        // epithet. `ctx.working` is the source string (token offsets index into it), so
                        // the slice keeps the original spacing.
                        if i < ts.len() {
                            ctx.name.phrase =
                                Some(ctx.working[marker_start..ts[ts.len() - 1].end].to_string());
                            i = ts.len();
                        } else if w.eq_ignore_ascii_case("species") {
                            ctx.name.phrase = Some(format!("species {number}"));
                        } else if !lower_epithets.is_empty() {
                            ctx.name.phrase =
                                Some(ctx.working[marker_start..number_end].to_string());
                        } else {
                            ctx.name.phrase = Some(number);
                        }
                    } else if i < ts.len()
                        && ts[i].kind == TokenKind::Word
                        && ts[i].text.chars().count() == 1
                        && starts_upper(&ts[i])
                        && i + 1 == ts.len()
                    {
                        // Single uppercase letter following sp. — informal phrase
                        // identifier ("Bryozoan sp. E"). Stored as phrase, leaves
                        // indet=true; after a species epithet with its marker.
                        ctx.name.phrase = Some(if lower_epithets.is_empty() {
                            ts[i].text.clone()
                        } else {
                            ctx.working[marker_start..ts[i].end].to_string()
                        });
                        i += 1;
                    } else if i < ts.len()
                        && ts[i].kind == TokenKind::Word
                        && ts[i].text.chars().count() >= 2
                        && (i + 1 == ts.len()
                            || (i + 1 < ts.len()
                                && ts[i + 1].kind == TokenKind::Number
                                && i + 2 == ts.len()))
                    {
                        // "Genus sp. JYr4" / "Genus sp. JGP0404" — strain code
                        // immediately after the marker with nothing else following.
                        // Capture as informal phrase (species stays indet). Allow WORD
                        // or WORD+NUMBER (when the tokenizer splits a mixed
                        // letters-and-digits code like "JGP" + "0404").
                        let mut code = ts[i].text.clone();
                        i += 1;
                        if i < ts.len() && ts[i].kind == TokenKind::Number {
                            code.push_str(&ts[i].text);
                            i += 1;
                        }
                        if is_strain_code(&code) {
                            // after a species epithet the marker stays in the phrase, as with a
                            // number ("Euxoa idahoensis sp. 1clay")
                            ctx.name.phrase = Some(if lower_epithets.is_empty() {
                                code
                            } else {
                                ctx.working[marker_start..ts[i - 1].end].to_string()
                            });
                        }
                    }
                    // 5.0.0 enhancement (deliberately BEYOND Java 4.2.0): a supraspecific indet
                    // marker followed by a trailing tag the narrow branches above did not capture —
                    // a multi-token specimen/culture/BOLD code like "Rhizobium sp. RMCC TR1811" or
                    // "Ichneumonidae sp. UAM Ento 145060" — captures the whole VERBATIM tail as the
                    // informal phrase instead of silently dropping it token-by-token at the
                    // "upper word … skip" fallthrough below. Fires only for the pure
                    // supraspecific-provisional band (no species epithet, so it stays out of the
                    // binomial-core names that remain Parsed) and only when nothing was captured
                    // above. Rescues the ~382k "tag not captured" rows the 67.5M verbatim-corpus
                    // study found. Token offsets index into
                    // `ctx.working` (`ctx.tokens = tokenize(&ctx.working)`, unmodified until here),
                    // so the slice is the exact source substring with its original spacing.
                    if name_section_covers_all
                        && ctx.name.phrase.is_none()
                        && lower_epithets.is_empty()
                        && i < ts.len()
                    {
                        let start = ts[i].start;
                        let end = ts[ts.len() - 1].end;
                        ctx.name.phrase = Some(ctx.working[start..end].to_string());
                        i = ts.len();
                    }
                    continue;
                }
                // 2b. "sp." between species and infraspecific epithet is almost always a
                // misspelling of "ssp." (subspecies). Only triggers when there's already
                // a species epithet and a lower epithet follows.
                if (w.eq_ignore_ascii_case("sp") || w.eq_ignore_ascii_case("spp"))
                    && lower_epithets.len() == 1
                    && marker_idx_in_epithets < 0
                    && has_infraspecific_epithet_after(ts, i)
                {
                    inline_rank = Some(Rank::Subspecies);
                    marker_idx_in_epithets = lower_epithets.len() as i32;
                    i += 1;
                    if i < ts.len() && ts[i].kind == TokenKind::Dot {
                        i += 1;
                    }
                    continue;
                }
                // 2c. A bare trailing "sp."/"spec."/"species" after a species epithet,
                // with nothing following, is a redundant leftover marker — drop it and
                // keep the binomial at SPECIES rather than reading "sp" as an
                // infraspecific epithet (which yielded INFRASPECIFIC_NAME with epithet
                // "sp"). The sp.->ssp. case (2b) already handled a following epithet.
                if (w.eq_ignore_ascii_case("sp")
                    || w.eq_ignore_ascii_case("spp")
                    || w.eq_ignore_ascii_case("spec")
                    || w.eq_ignore_ascii_case("species"))
                    && !lower_epithets.is_empty()
                    && marker_idx_in_epithets < 0
                {
                    let mut j = i + 1;
                    if j < ts.len() && ts[j].kind == TokenKind::Dot {
                        j += 1;
                    }
                    if j >= ts.len() {
                        i = j;
                        continue;
                    }
                }
                // 3. infraspecific rank marker (with notho-prefix support)
                if let Some((rm_infra, notho_flag)) =
                    rank_markers::match_infraspecific_allow_notho(w)
                {
                    // The capitalised final epithet of an autonym ("var. Tenuicaulis").
                    let autonym_at = i
                        + 1
                        + usize::from(ts.get(i + 1).is_some_and(|t| t.kind == TokenKind::Dot));
                    let capitalised_autonym = !lower_epithets.is_empty()
                        && super::authorship_split::is_capitalised_autonym(ts, autonym_at);
                    if has_infraspecific_epithet_after(ts, i) || capitalised_autonym {
                        // Second marker overriding the first: the previous
                        // classification (oldRank.epithet + author) was an intermediate
                        // level the model can't hold, so warn about the drop.
                        if let Some(old_rank) = inline_rank.filter(|_| {
                            marker_idx_in_epithets >= 0
                                && (marker_idx_in_epithets as usize) < lower_epithets.len()
                        }) {
                            let old_marker = old_rank.marker();
                            let old_epithet = &lower_epithets[marker_idx_in_epithets as usize];
                            let mut sb = String::from(warnings::REMOVED_PREFIX);
                            if let Some(m) = old_marker {
                                sb.push_str(m);
                                sb.push(' ');
                            }
                            sb.push_str(old_epithet);
                            if let Some(pma) = &pending_mid_name_author {
                                if !pma.is_empty() {
                                    sb.push(' ');
                                    sb.push_str(pma);
                                }
                            }
                            ctx.name.add_warning(&sb);
                            ctx.name.add_warning(warnings::QUADRINOMIAL);
                        }
                        inline_rank = Some(rm_infra);
                        inline_rank_notho = notho_flag;
                        marker_idx_in_epithets = lower_epithets.len() as i32;
                        pending_mid_name_author = None;
                        i += 1;
                        if i < ts.len() && ts[i].kind == TokenKind::Dot {
                            i += 1;
                        }
                        // microbial "f. sp." -> forma specialis: skip the "sp" + dot too
                        if i + 1 < ts.len()
                            && ts[i].kind == TokenKind::Word
                            && ts[i].text.eq_ignore_ascii_case("sp")
                        {
                            inline_rank = Some(Rank::FormaSpecialis);
                            i += 1;
                            if i < ts.len() && ts[i].kind == TokenKind::Dot {
                                i += 1;
                            }
                        }
                        if capitalised_autonym {
                            lower_epithets.push(ts[i].text.to_lowercase());
                            i += 1;
                            continue;
                        }
                        // Informal infra epithet: a single letter immediately following the
                        // rank marker ("form A", "f. B", "var. a", "f. (a)") — consume it here
                        // so the normal lowercase-epithet path doesn't drop it as an "upper
                        // word" or a parenthesised remainder.
                        // The full token list, so a dotted abbreviation (`ab. n. undularia`)
                        // can see the epithet AuthorshipSplit already carved off as authorship.
                        if let Some(len) =
                            super::authorship_split::single_letter_designation(&ctx.tokens, i)
                                .filter(|len| i + len <= ts.len())
                        {
                            let letter = &ts[i + len / 2];
                            lower_epithets.push(letter.text.clone());
                            indet = true; // INFORMAL informal infra epithet
                            i += len;
                        }
                    } else if inline_rank.is_some()
                        && marker_idx_in_epithets == lower_epithets.len() as i32
                        && ctx
                            .tokens
                            .get(i + 1)
                            .is_none_or(|t| t.kind == TokenKind::Word && starts_upper(t))
                    {
                        // An undotted marker word where the previous marker's epithet belongs,
                        // last or before its author, is that epithet ("Haliotis cracherodii var.
                        // lusus Finlay, 1927"); a dotted one is a marker ("Abies alba subsp. var.").
                        lower_epithets.push(w.to_lowercase());
                        i += 1;
                    } else if !lower_epithets.is_empty() {
                        // Trailing rank marker with no following epithet = indetermined
                        // infraspecific
                        indet = true;
                        inline_rank = Some(rm_infra);
                        inline_rank_notho = notho_flag;
                        marker_idx_in_epithets = lower_epithets.len() as i32;
                        i += 1;
                        if i < ts.len() && ts[i].kind == TokenKind::Dot {
                            i += 1;
                        }
                        // gbif/name-parser-rust#16: "no following EPITHET" did not mean nothing
                        // follows — "Abies alba var. 3" / "subsp. 7" / "var. B12" carry a
                        // DESIGNATION after the marker, and it was falling out of the loop
                        // token-by-token below, leaving a name that rendered "Abies alba var."
                        // and collided with every other numbered variety of the same species.
                        // Same treatment the species-level indet marker already gets 200 lines
                        // up: once a phrase starts it runs to the end, captured verbatim from
                        // `ctx.working` so the original spelling and spacing survive. The rank
                        // marker itself stays OUT of the phrase (unlike the supraspecific case,
                        // where nothing else carries it) because `inline_rank` already holds it
                        // and the formatter re-emits it — so `Abies alba var. 3` round-trips
                        // with a phrase of just "3".
                        //
                        // `ts` is the NAME section only, so whatever is left here is by
                        // construction not the authorship — AuthorshipSplit already carved that
                        // off, and it keeps a designation in this section on purpose (see its own
                        // "a NUMBERED indeterminate infraspecific" branch). That is what lets
                        // "Abies alba subsp. 7 Mill." come back with BOTH phrase "7" and the
                        // author Mill., instead of dropping the 7 as it used to.
                        //
                        // …and never over a numeral-prefixed epithet: `var. 4-lineata` is a real
                        // infraspecific epithet that `has_infraspecific_epithet_after` does not
                        // recognise (it starts with a digit, not a letter) but the ordinary
                        // epithet path below does. See [`token::is_numeral_epithet`].
                        let numeral_epithet_follows = i < ts.len()
                            && ts[i].kind == TokenKind::Word
                            && token::is_numeral_epithet(&ts[i].text);
                        if ctx.name.phrase.is_none() && i < ts.len() && !numeral_epithet_follows {
                            let start = ts[i].start;
                            let end = ts[ts.len() - 1].end;
                            ctx.name.phrase = Some(ctx.working[start..end].to_string());
                            ctx.name.type_ = NameType::Informal;
                            i = ts.len();
                        }
                    } else {
                        // Rank marker before any lower epithet and no following epithet:
                        // treat it as the specific epithet (e.g. "Foa fo" — "fo" is the
                        // species name, not a forma rank indicator).
                        lower_epithets.push(w.to_string());
                        i += 1;
                        if i < ts.len() && ts[i].kind == TokenKind::Dot {
                            i += 1;
                        }
                    }
                    continue;
                }
                // 4. infrageneric marker (with optional "notho-" prefix)
                if let Some((rm_infragen, gnotho)) = rank_markers::match_infrageneric_allow_notho(w)
                    .filter(|_| !authorship_split::is_epithet_before_dated_author(&ctx.tokens, i))
                {
                    if lower_epithets.is_empty() && subgenus.is_none() && infragen_epithet.is_none()
                    {
                        i += 1;
                        if i < ts.len() && ts[i].kind == TokenKind::Dot {
                            i += 1;
                        }
                        if i < ts.len() && ts[i].kind == TokenKind::Word && starts_upper(&ts[i]) {
                            infragen_epithet = Some(ts[i].text.clone());
                            infrageneric = Some(rm_infragen);
                            if gnotho {
                                ctx.name.add_notho(NamePart::Infrageneric);
                            }
                            i += 1;
                        }
                        continue;
                    }
                }
                // 5. hyphenated aggregate suffix: "X-group"
                {
                    let low = t.text.to_lowercase();
                    let mut stripped: Option<String> = None;
                    for suf in AGG_HYPHEN_SUFFIXES {
                        let char_count = t.text.chars().count();
                        let suf_chars = suf.chars().count();
                        if low.ends_with(suf) && char_count > suf_chars {
                            stripped = Some(t.text.chars().take(char_count - suf_chars).collect());
                            break;
                        }
                    }
                    if let Some(s) = stripped {
                        ctx.aggregate = true;
                        lower_epithets.push(s);
                        i += 1;
                        continue;
                    }
                }
                // 6. standalone aggregate marker after a species epithet
                if !lower_epithets.is_empty()
                    && (w.eq_ignore_ascii_case("agg")
                        || w.eq_ignore_ascii_case("aggregate")
                        || w.eq_ignore_ascii_case("group")
                        || w.eq_ignore_ascii_case("complex"))
                {
                    ctx.aggregate = true;
                    i += 1;
                    if i < ts.len() && ts[i].kind == TokenKind::Dot {
                        i += 1;
                    }
                    continue;
                }
                // 7. "species group" two-word marker
                if !lower_epithets.is_empty()
                    && w.eq_ignore_ascii_case("species")
                    && i + 1 < ts.len()
                    && ts[i + 1].kind == TokenKind::Word
                    && ts[i + 1].text.eq_ignore_ascii_case("group")
                {
                    ctx.aggregate = true;
                    i += 2;
                    continue;
                }
                // 8. ordinary epithet
                // A lone letter in the species slot (`Collettea a`) is a provisional designation
                // — no species-group name has a single letter (ICZN Art. 11.9.1) — unless it is
                // an abbreviation (`c.album`, `h.-čejkai`, `Abacina s.`) or the first half of a
                // split epithet (`Drepana z nigrum`). Looks at the full token list so a following
                // word AuthorshipSplit carved off still counts.
                if lower_epithets.is_empty()
                    && t.text.len() == 1
                    && t.text.as_bytes()[0].is_ascii_lowercase()
                {
                    letter_designation = !ctx.tokens.get(i + 1).is_some_and(|nx| {
                        nx.kind == TokenKind::Dot
                            || (nx.kind == TokenKind::Word && starts_lower(nx))
                    });
                }
                lower_epithets.push(t.text.clone());
                i += 1;
                continue;
            }
            // upper word inside name section: skip
            i += 1;
            continue;
        }
        i += 1;
    }

    let mut specific: Option<String> = None;
    let mut infraspecific: Option<String> = None;
    let mut rank: Option<Rank> = None;

    // A quadrinomial keeps the species and the last infraspecific epithet; one dropped in between
    // is reported, as a second rank marker's is above ("Poa pratensis kewensis primula", "Acipenser
    // gueldenstaedti colchicus natio danubicus"). Java dropped it silently.
    let warn_dropped = |name: &mut crate::model::ParsedName, dropped: &[String]| {
        if dropped.is_empty() || name.warnings.iter().any(|w| w == warnings::QUADRINOMIAL) {
            return;
        }
        for epithet in dropped {
            name.add_warning(&format!("{}{epithet}", warnings::REMOVED_PREFIX));
        }
        name.add_warning(warnings::QUADRINOMIAL);
    };
    if marker_idx_in_epithets >= 0 {
        let midx = marker_idx_in_epithets as usize;
        if marker_idx_in_epithets >= 1 {
            specific = Some(lower_epithets[0].clone());
        }
        if midx >= 2 && midx <= lower_epithets.len() {
            warn_dropped(&mut ctx.name, &lower_epithets[1..midx]);
        }
        if midx < lower_epithets.len() {
            infraspecific = Some(lower_epithets[midx].clone());
        }
        rank = inline_rank;
        if lower_epithets.len() as i32 > marker_idx_in_epithets + 1 {
            let mut rem = String::new();
            for epithet in &lower_epithets[(midx + 1)..] {
                if !rem.is_empty() {
                    rem.push(' ');
                }
                rem.push_str(epithet);
            }
            ctx.name.state = State::Partial;
            ctx.name.unparsed = Some(rem);
        }
    } else if !lower_epithets.is_empty() {
        specific = Some(lower_epithets[0].clone());
        if lower_epithets.len() >= 2 {
            infraspecific = Some(lower_epithets[lower_epithets.len() - 1].clone());
            warn_dropped(&mut ctx.name, &lower_epithets[1..lower_epithets.len() - 1]);
            // A trinomial without a rank marker takes the caller's infraspecific rank hint
            // ("Abies alba alpina" + VARIETY): the source's rank column is all there is. A
            // cultivar rank needs a cultivar epithet, which a plain trinomial does not have.
            let hinted = ctx.requested_rank.filter(|r| {
                r.is_infraspecific()
                    && !matches!(r, Rank::Cultivar | Rank::CultivarGroup | Rank::Grex)
                    && lower_epithets.len() == 2
            });
            rank = Some(if let Some(r) = hinted {
                r
            } else if lower_epithets.len() == 2 {
                Rank::InfraspecificName
            } else {
                Rank::InfrasubspecificName
            });
        } else if ctx.name.phrase.is_some() && ctx.name.rank.is_infraspecific() {
            // A stashed infraspecific designation (`Farrea occa n_ssp_NIWA_SO254`) already set
            // the rank of the unnamed taxon the phrase stands for — the binomial is only its head.
            rank = Some(ctx.name.rank);
        } else {
            rank = Some(Rank::Species);
        }
    }

    // A rank marker and its epithet straight after the genus ("Navicula var. fasciata") name an
    // infraspecific taxon whose species is missing: the genus stays a genus and the name is
    // INFORMAL. Java made the genus a uninomial and dropped the epithet.
    let species_missing = genus.is_some() && specific.is_none() && infraspecific.is_some();
    if species_missing {
        ctx.name.type_ = NameType::Informal;
    }
    let treat_as_genus = ctx.name.cultivar_epithet.is_some()
        || species_missing
        || (indet && lower_epithets.is_empty() && genus.is_some());
    if !treat_as_genus
        && genus.is_some()
        && specific.is_none()
        && subgenus.is_none()
        && infragen_epithet.is_none()
    {
        ctx.name.uninomial = genus.clone();
    } else {
        ctx.name.genus = genus.clone();
        if let Some(sg) = &subgenus {
            ctx.name.infrageneric_epithet = Some(sg.clone());
        }
        if let Some(ie) = &infragen_epithet {
            ctx.name.infrageneric_epithet = Some(ie.clone());
        }
        if let Some(sp) = &specific {
            ctx.name.specific_epithet = Some(sp.clone());
        }
        if let Some(isp) = &infraspecific {
            ctx.name.infraspecific_epithet = Some(isp.clone());
        }
    }

    let requested = ctx.requested_rank;
    let cultivar_set = ctx.name.cultivar_epithet.is_some();
    // `.filter(|_| !cultivar_set)` preserves the Java `X != null && !cultivarSet` compound
    // condition exactly (as opposed to nesting `if !cultivar_set` INSIDE `if let Some(x) =
    // X`, which would wrongly skip the next `else if` whenever `X` is `Some` but
    // `cultivar_set` is true, rather than falling through to test it like Java's flat
    // `else if` chain does).
    if let Some(r) = rank.filter(|_| !cultivar_set) {
        ctx.name.rank = r;
    } else if let Some(ig) = infrageneric.filter(|_| !cultivar_set) {
        ctx.name.rank = ig;
    } else if subgenus.is_some()
        && !cultivar_set
        && specific.is_none()
        && infraspecific.is_none()
        && ctx.name.rank == Rank::Unranked
    {
        // Paren-based subgenus ("Calathus (Lindrothius)") without an explicit rank
        // marker — default to the generic infrageneric rank.
        ctx.name.rank = Rank::InfragenericName;
    }

    if inline_rank_notho {
        ctx.name.add_notho(NamePart::Infraspecific);
    }
    if let Some(q) = cf_aff_qualifier.as_ref().filter(|_| !qualifier_phrase) {
        // Qualifier applies to the epithet after it, or the last one when it trails them all.
        let followed = lower_epithets.len() > qualifier_at;
        let part = if infraspecific.is_none() || (qualifier_at == 0 && followed) {
            NamePart::Specific
        } else {
            NamePart::Infraspecific
        };
        ctx.name.set_epithet_qualifier(part, q);
    }
    if letter_designation && specific.is_some() {
        ctx.name.type_ = NameType::Informal;
    }
    if indet {
        ctx.name.type_ = NameType::Informal;
        if infraspecific.is_none() {
            if lower_epithets.is_empty() && ctx.name.rank == Rank::Unranked {
                ctx.name.rank = Rank::Species;
            }
            if ctx.name.phrase.is_none() || indet_bare {
                ctx.name.add_warning(warnings::INDETERMINED);
            }
        }
    }
    // Caller-supplied infraspecific rank on a binomial with no infraspecific epithet ->
    // treat as indeterminate infraspecific (e.g. "Lepidoptera alba DC." + SUBSPECIES)
    if !indet
        && requested.is_some_and(|r| r != Rank::Unranked && r.is_infraspecific())
        && specific.is_some()
        && infraspecific.is_none()
    {
        ctx.name.type_ = NameType::Informal;
        ctx.name.rank = requested.unwrap();
        // a stashed designation phrase (`n_ssp_NIWA_SO254`) is the missing terminal epithet
        if ctx.name.phrase.is_none() {
            ctx.name.add_warning(warnings::INDETERMINED);
        }
    }
    if requested == Some(Rank::Species) && infraspecific.is_some() {
        ctx.name.add_warning(warnings::SUBSPECIES_ASSIGNED);
    }
}

/// Java `Token.startsUpper()` (`Token.java:20-22`). Duplicates
/// `authorship_split::starts_upper` (private there) rather than importing it — see this
/// module's own doc comment for why (`Token` doesn't carry this as a method yet, and each
/// class's own copy mirrors the Java source, which also has no shared free-function form).
fn starts_upper(t: &Token) -> bool {
    t.text.chars().next().is_some_and(|c| c.is_uppercase())
}

/// Java `Token.startsLower()` (`Token.java:24-26`). See [`starts_upper`]'s doc comment.
fn starts_lower(t: &Token) -> bool {
    t.text.chars().next().is_some_and(|c| c.is_lowercase())
}

/// Java `Token.startsDigitEpithet()` (`Token.java:29-33`). See [`starts_upper`]'s doc
/// comment; same ASCII-only digit approximation as `authorship_split::starts_digit_epithet`
/// and `token.rs`'s own tokenizer-internal `is_digit`.
fn starts_digit_epithet(t: &Token) -> bool {
    t.kind == TokenKind::Word
        && t.text.chars().next().is_some_and(|c| c.is_ascii_digit())
        && t.text.chars().any(|c| c.is_alphabetic())
}

/// Renders the WORD/DOT tokens in `[from, to)` into the canonical inline author form ("B.
/// Boivin" -> "B.Boivin"). Java `NameTokens.renderAuthorSpan(List<Token>, int, int)`
/// (`NameTokens.java:543-555`). Used to record the dropped author span when an intermediate
/// classification is overridden by a later rank marker.
fn render_author_span(ts: &[Token], from: usize, to: usize) -> String {
    let mut sb = String::new();
    for t in &ts[from..to] {
        match t.kind {
            TokenKind::Word => {
                if !sb.is_empty() && !sb.ends_with('.') {
                    sb.push(' ');
                }
                sb.push_str(&t.text);
            }
            TokenKind::Dot if !sb.is_empty() && !sb.ends_with('.') => {
                sb.push('.');
            }
            _ => {}
        }
    }
    sb
}

/// Java `NameTokens.skipParenAuthorBlock(List<Token>, int, int)` (`NameTokens.java:561-594`)
/// — **NameTokens's OWN copy**, textually different from
/// `authorship_split::skip_paren_author_block` of the same name: this one returns as soon
/// as it finds a matching infraspecific-marker word, with NO `hasEpithetAfterMarker`-style
/// follow-up check — the two copies are intentionally NOT unified (see this module's own
/// doc comment, and `authorship_split`'s, for the full rationale). If a "(...) Author"
/// span sits between the species epithet and an infraspecific rank marker, returns the
/// index of the rank-marker token; otherwise `None`.
fn skip_paren_author_block(ts: &[Token], open_idx: usize) -> Option<usize> {
    let n = ts.len();
    let mut depth = 1i32;
    let mut j = open_idx + 1;
    while j < n && depth > 0 {
        let k = ts[j].kind;
        if k == TokenKind::OpenParen {
            depth += 1;
        } else if k == TokenKind::CloseParen {
            depth -= 1;
        }
        if depth == 0 {
            break;
        }
        j += 1;
    }
    if j >= n || depth != 0 {
        return None;
    }
    j += 1; // skip past the close paren
    let from = j;
    while j < n {
        if authorship_split::continues_author_span(ts, j, from) {
            j += 1;
            continue;
        }
        let t = &ts[j];
        if t.kind == TokenKind::Word {
            let w = strip_dot(&t.text);
            if rank_markers::match_infraspecific_allow_notho(w).is_some() {
                return Some(j);
            }
        }
        return None;
    }
    None
}

/// Title-cases all-caps or all-lower-case Latin words used as a genus/uninomial. Java
/// `NameTokens.recoverCase(ParseContext, String, boolean)` (`NameTokens.java:597-654`). The
/// Java signature's `ParseContext ctx` parameter is never actually read in the method body
/// (verified against the Java source — every branch touches only `text`/`isGenus`), so
/// it's dropped here rather than threaded through as an unused parameter.
///
/// `Character.toUpperCase`/`toLowerCase` (single-codepoint, Java's SIMPLE case mapping)
/// vs. Rust's `char::to_uppercase`/`to_lowercase` (FULL Unicode case mapping, which for a
/// handful of characters — e.g. German 'ß' -> "SS" — yields more than one output char):
/// genus/epithet text is near-universally Latin/ASCII, so this divergence is not expected
/// to be exercised by the corpus; any case where it is would surface as a golden-corpus
/// mismatch (same acceptable-divergence rationale already documented elsewhere in this
/// port, e.g. `token.rs`'s `is_digit`).
fn recover_case(text: &str, is_genus: bool) -> String {
    if text.chars().count() < 2 {
        return text.to_string();
    }

    // For hyphenated genera, normalize subsequent segments to lowercase under specific
    // conditions:
    // - first segment is <= 3 chars (short prefix like "Eu-", "Le-", "Uva-"), OR
    // - name has 3+ hyphenated segments (e.g. "Prunus-Lauro-Cerasus")
    // This does NOT apply when subsequent segments already start lowercase.
    if is_genus && text.contains('-') {
        let parts: Vec<&str> = text.split('-').collect();
        if parts.len() >= 2 {
            let needs_norm = parts[1..]
                .iter()
                .any(|p| p.chars().next().is_some_and(|c| c.is_uppercase()));
            if needs_norm && (parts[0].chars().count() <= 3 || parts.len() >= 3) {
                let mut sb = String::from(parts[0]);
                for p in &parts[1..] {
                    sb.push('-');
                    let mut chars = p.chars();
                    match chars.next() {
                        Some(c) if c.is_uppercase() => {
                            let lc: String = c.to_lowercase().collect();
                            sb.push_str(&lc);
                            sb.push_str(chars.as_str());
                        }
                        _ => sb.push_str(p),
                    }
                }
                return sb;
            }
        }
    }

    let upper = is_all_letter_case(text, true);
    let lower = !upper && is_all_letter_case(text, false);
    if !upper && !lower {
        return text.to_string();
    }
    let mut b = String::with_capacity(text.len());
    let mut first = true;
    for c in text.chars() {
        if c.is_alphabetic() {
            if first {
                let uc: String = c.to_uppercase().collect();
                b.push_str(&uc);
                first = false;
            } else {
                let lc: String = c.to_lowercase().collect();
                b.push_str(&lc);
            }
        } else {
            b.push(c);
            first = true;
        }
    }
    b
}

/// Java `NameTokens.isAllUpperLetters(String)` (`NameTokens.java:656-658`) — thin delegate
/// to [`is_all_letter_case`].
fn is_all_upper_letters(s: &str) -> bool {
    is_all_letter_case(s, true)
}

/// True if the marker at `marker_idx` is followed by a recognisable infraspecific epithet —
/// either a lowercase word, or a single uppercase letter (informal collector tag like "f.
/// A" / "form A"). Java `NameTokens.hasInfraspecificEpithetAfter(List<Token>, int, int)`
/// (`NameTokens.java:665-675`).
fn has_infraspecific_epithet_after(ts: &[Token], marker_idx: usize) -> bool {
    let n = ts.len();
    let mut k = marker_idx + 1;
    if k < n && ts[k].kind == TokenKind::Dot {
        k += 1;
    }
    // tolerate an interleaving hybrid mark before the epithet ("var. ×alpina")
    if k < n && ts[k].kind == TokenKind::HybridMark {
        k += 1;
    }
    if k >= n {
        return false;
    }
    let nx = &ts[k];
    if nx.kind == TokenKind::Word && starts_lower(nx) {
        return true;
    }
    if nx.kind == TokenKind::Word && nx.text.chars().count() == 1 && starts_upper(nx) {
        return true;
    }
    // a numeral epithet ("var. 4-lineata") is an epithet too, not a designation
    nx.kind == TokenKind::Word && token::is_numeral_epithet(&nx.text)
}

/// The "ex" at `i` opens "ex gr." / "ex grege" ("Acastella ex gr. rouaulti") before an epithet.
pub(crate) fn is_ex_grege(ts: &[Token], i: usize) -> bool {
    let gr = ts.get(i + 1).filter(|t| t.kind == TokenKind::Word);
    let gr_dot = usize::from(ts.get(i + 2).is_some_and(|t| t.kind == TokenKind::Dot));
    gr.is_some_and(|t| t.text == "gr" || t.text == "grege")
        && ts
            .get(i + 2 + gr_dot)
            .is_some_and(|t| t.kind == TokenKind::Word && starts_lower(t))
}

/// True for strain-code-shaped tokens — mixed letters and digits, no spaces, length >= 3
/// (e.g. "JGP0404", "ANIC_3", "Sb1"). Used to glue the code onto an "sp." indet marker so
/// the species epithet rendering keeps the strain identifier. Java
/// `NameTokens.isStrainCode(String)` (`NameTokens.java:682-693`).
fn is_strain_code(s: &str) -> bool {
    if s.chars().count() < 3 {
        return false;
    }
    let mut has_letter = false;
    let mut has_digit = false;
    for c in s.chars() {
        if c.is_alphabetic() {
            has_letter = true;
        } else if c.is_ascii_digit() {
            // Java's Character.isDigit is Unicode Nd; token.rs's own is_digit already
            // documents this crate's ASCII-only approximation for Character.isDigit —
            // matched here for consistency.
            has_digit = true;
        }
    }
    has_letter && has_digit
}

/// Java `NameTokens.isAllLetterCase(String, boolean)` (`NameTokens.java:695-707`).
fn is_all_letter_case(s: &str, upper: bool) -> bool {
    let mut have_letter = false;
    for c in s.chars() {
        if c.is_alphabetic() {
            have_letter = true;
            if upper && !c.is_uppercase() {
                return false;
            }
            if !upper && !c.is_lowercase() {
                return false;
            }
        }
    }
    have_letter
}

/// Java `NameTokens.stripDot(String)` (`NameTokens.java:709-711`) — NameTokens's OWN copy
/// (identical body to `authorship_split::strip_dot`, matching Java's own two separate
/// private statics of the same name).
fn strip_dot(s: &str) -> &str {
    s.strip_suffix('.').unwrap_or(s)
}

#[cfg(test)]
mod tests {
    use super::*;
    use crate::model::NomCode;
    use crate::token::tokenize;

    /// Runs the real pipeline stages `classify` depends on (tokenize + find_boundary) and
    /// then `classify` itself, returning the resulting `ParseContext` for inspection of
    /// both `ctx.name` and the `ctx` side-channels (`mid_author_from`/`aggregate`).
    fn run(input: &str, requested_rank: Option<Rank>) -> ParseContext {
        let mut ctx = ParseContext::new(input.to_string(), None, requested_rank, None);
        ctx.tokens = tokenize(input);
        let boundary = authorship_split::find_boundary(&ctx.tokens, &ctx);
        classify(&mut ctx, boundary);
        ctx
    }

    #[test]
    fn boundary_zero_short_circuits_to_partial_unparsed_and_touches_nothing_else() {
        // "a" alone (a single lowercase letter) makes find_boundary return 0 (Java:
        // AuthorshipSplit's own first-token handling only accepts a lower-case first
        // token when it's 2+ chars long). Testing this in isolation with `boundary`
        // passed explicitly (rather than relying on find_boundary to produce 0) is the
        // most direct way to exercise NameTokens's own early-return branch.
        let mut ctx = ParseContext::new("a".to_string(), None, None, None);
        ctx.tokens = tokenize("a");
        classify(&mut ctx, 0);
        assert_eq!(ctx.name.state, State::Partial);
        assert_eq!(ctx.name.unparsed, Some("a".to_string()));
        assert_eq!(ctx.name.genus, None);
        assert_eq!(ctx.name.uninomial, None);
        assert!(ctx.name.warnings.is_empty());
    }

    #[test]
    fn bare_uninomial_routes_to_uninomial_not_genus() {
        let ctx = run("Abies", None);
        assert_eq!(ctx.name.uninomial, Some("Abies".to_string()));
        assert_eq!(ctx.name.genus, None);
        assert_eq!(ctx.name.rank, Rank::Unranked);
    }

    #[test]
    fn binomial_routes_genus_and_specific_epithet_not_uninomial() {
        let ctx = run("Abies alba", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.uninomial, None);
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.rank, Rank::Species);
    }

    #[test]
    fn cultivar_epithet_present_forces_genus_routing_even_for_a_bare_uninomial() {
        let mut ctx = ParseContext::new("Acer".to_string(), None, None, None);
        ctx.name.cultivar_epithet = Some("Elsrijk".to_string());
        ctx.tokens = tokenize("Acer");
        let boundary = authorship_split::find_boundary(&ctx.tokens, &ctx);
        classify(&mut ctx, boundary);
        assert_eq!(ctx.name.genus, Some("Acer".to_string()));
        assert_eq!(
            ctx.name.uninomial, None,
            "treatAsGenus (cultivarEpithet != null) must route through setGenus, not setUninomial"
        );
    }

    #[test]
    fn paren_subgenus_with_trailing_species_sets_infrageneric_epithet_and_rank_from_species() {
        let ctx = run("Amnicola (Amnicola) dubrueilliana", None);
        assert_eq!(ctx.name.genus, Some("Amnicola".to_string()));
        assert_eq!(ctx.name.infrageneric_epithet, Some("Amnicola".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("dubrueilliana".to_string()));
        assert_eq!(ctx.name.rank, Rank::Species);
    }

    #[test]
    fn paren_subgenus_alone_with_no_species_defaults_rank_to_infrageneric_name() {
        let ctx = run("Arrhoges (Antarctohoges)", None);
        assert_eq!(ctx.name.genus, Some("Arrhoges".to_string()));
        assert_eq!(
            ctx.name.infrageneric_epithet,
            Some("Antarctohoges".to_string())
        );
        assert_eq!(ctx.name.specific_epithet, None);
        assert_eq!(
            ctx.name.rank,
            Rank::InfragenericName,
            "a bare paren-subgenus with no explicit rank marker defaults to INFRAGENERIC_NAME"
        );
        assert_eq!(
            ctx.name.uninomial, None,
            "subgenus != null routes to genus even though there's no species epithet"
        );
    }

    #[test]
    fn abbreviated_subgenus_sets_infrageneric_epithet_with_dot_and_warns() {
        let ctx = run("Phalaena (Tin.) guttella", None);
        assert_eq!(ctx.name.genus, Some("Phalaena".to_string()));
        assert_eq!(ctx.name.infrageneric_epithet, Some("Tin.".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("guttella".to_string()));
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::ABBREVIATED_SUBGENUS.to_string()));
    }

    #[test]
    fn infrageneric_marker_then_paren_subgenus_the_marker_form_wins_the_overwrite() {
        // Regression for the "infragenericEpithet from subgenus then overwritten by
        // infragenEpithet" trap: NameTokens.java always applies `if (subgenus != null)
        // setInfragenericEpithet(subgenus)` BEFORE `if (infragenEpithet != null)
        // setInfragenericEpithet(infragenEpithet)`, so when both are non-null the
        // marker-based one (set second) wins. `find_boundary` would normally exclude a
        // paren immediately following a consumed infrageneric marker from the name
        // boundary (its own `after_subgenus` flag blocks the subgenus cascade there) —
        // so this test calls `classify` directly with a hand-picked boundary that
        // includes both, to exercise NameTokens's own internal transition in isolation
        // (a legitimate use of `classify`'s explicit `boundary` parameter).
        let input = "Genus sect. Foo (Bar)";
        let mut ctx = ParseContext::new(input.to_string(), None, None, None);
        ctx.tokens = tokenize(input);
        let boundary = ctx.tokens.len();
        classify(&mut ctx, boundary);
        assert_eq!(ctx.name.genus, Some("Genus".to_string()));
        assert_eq!(
            ctx.name.infrageneric_epithet,
            Some("Foo".to_string()),
            "the infrageneric-MARKER form must win over the paren-subgenus form"
        );
        assert_eq!(ctx.name.rank, Rank::SectionBotany);
    }

    #[test]
    fn subsp_rank_marker_sets_infraspecific_epithet_and_rank_with_no_warnings() {
        let ctx = run("Abies alba subsp. alba", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.infraspecific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.rank, Rank::Subspecies);
        assert!(ctx.name.warnings.is_empty());
    }

    #[test]
    fn apostrophe_particle_mid_name_author_is_skipped_and_records_mid_author_span() {
        let ctx = run("Cirsium creticum d'Urv. subsp. creticum", None);
        assert_eq!(ctx.name.genus, Some("Cirsium".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("creticum".to_string()));
        assert_eq!(ctx.name.infraspecific_epithet, Some("creticum".to_string()));
        assert_eq!(ctx.name.rank, Rank::Subspecies);
        // tokens: [Cirsium(0) creticum(1) d'Urv(2) .(3) subsp(4) .(5) creticum(6)]
        assert_eq!(ctx.mid_author_from, 2);
        assert_eq!(ctx.mid_author_to, 4);
        assert!(ctx.name.warnings.is_empty());
    }

    #[test]
    fn hybrid_mark_before_genus_adds_generic_notho() {
        let ctx = run("x Abies alba", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.notho, Some(vec![NamePart::Generic]));
    }

    #[test]
    fn hybrid_mark_after_genus_before_any_epithet_adds_specific_notho() {
        let ctx = run("Abies \u{00D7} alba", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.notho, Some(vec![NamePart::Specific]));
    }

    #[test]
    fn notho_rank_marker_adds_to_an_earlier_generic_notho() {
        // A leading HYBRID_MARK adds GENERIC; the notho-prefixed infraspecific marker
        // ("nothovar.") adds INFRASPECIFIC beside it. Java replaced the set instead.
        let ctx = run("x Abies alba nothovar. rubra", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.infraspecific_epithet, Some("rubra".to_string()));
        assert_eq!(ctx.name.rank, Rank::Variety);
        assert_eq!(
            ctx.name.notho,
            Some(vec![NamePart::Generic, NamePart::Infraspecific])
        );
    }

    #[test]
    fn cf_qualifier_sets_epithet_qualifier_on_specific_and_marks_informal() {
        let ctx = run("Abies cf. alba", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.type_, NameType::Informal);
        let mut expected = std::collections::BTreeMap::new();
        expected.insert(NamePart::Specific, "cf.".to_string());
        assert_eq!(ctx.name.epithet_qualifier, Some(expected));
    }

    #[test]
    fn open_nomenclature_question_mark_between_epithets_sets_qualifier_on_infraspecific() {
        let ctx = run("Buteo borealis ? ventralis", None);
        assert_eq!(ctx.name.genus, Some("Buteo".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("borealis".to_string()));
        assert_eq!(
            ctx.name.infraspecific_epithet,
            Some("ventralis".to_string())
        );
        assert_eq!(ctx.name.rank, Rank::InfraspecificName);
        assert!(ctx.name.doubtful);
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::QUESTION_MARKS_REMOVED.to_string()));
        let mut expected = std::collections::BTreeMap::new();
        expected.insert(NamePart::Infraspecific, "?".to_string());
        assert_eq!(ctx.name.epithet_qualifier, Some(expected));
    }

    #[test]
    fn indet_species_marker_with_single_letter_phrase_suppresses_the_indetermined_warning() {
        let ctx = run("Bryozoan sp. E", None);
        assert_eq!(ctx.name.genus, Some("Bryozoan".to_string()));
        assert_eq!(
            ctx.name.uninomial, None,
            "indet routes through genus, not uninomial"
        );
        assert_eq!(ctx.name.specific_epithet, None);
        assert_eq!(ctx.name.phrase, Some("sp. E".to_string()));
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert_eq!(ctx.name.rank, Rank::Species);
        assert!(
            ctx.name.warnings.is_empty(),
            "a phrase identifier suppresses the generic INDETERMINED warning"
        );
    }

    #[test]
    fn bare_indet_species_marker_captures_the_marker_and_keeps_the_indetermined_warning() {
        // 5.0.0 Informal-phrase contract: a bare "Genus sp." captures the verbatim marker as its
        // phrase ("sp.") for a uniform taxon+phrase round-trip, but — unlike a name with a
        // distinguishing tail — it is still genuinely indeterminate, so the INDETERMINED warning
        // stays.
        let ctx = run("Bryozoan sp.", None);
        assert_eq!(ctx.name.genus, Some("Bryozoan".to_string()));
        assert_eq!(ctx.name.phrase, Some("sp.".to_string()));
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert_eq!(ctx.name.rank, Rank::Species);
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::INDETERMINED.to_string()));
    }

    #[test]
    fn informal_single_uppercase_letter_after_rank_marker_is_captured_as_infraspecific() {
        let ctx = run("Foo bar var. A", None);
        assert_eq!(ctx.name.genus, Some("Foo".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("bar".to_string()));
        assert_eq!(ctx.name.infraspecific_epithet, Some("A".to_string()));
        assert_eq!(ctx.name.rank, Rank::Variety);
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert!(
            ctx.name.warnings.is_empty(),
            "infraspecific is set, so the indet-with-no-infraspecific INDETERMINED path must not fire"
        );
    }

    #[test]
    fn quadrinomial_overflow_drops_the_first_marker_and_flags_partial_unparsed() {
        let ctx = run("Abies alba subsp. nigra var. rubra pallens", None);
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(
            ctx.name.infraspecific_epithet,
            Some("rubra".to_string()),
            "the SECOND marker (var.) wins; the first (subsp. nigra) is dropped"
        );
        assert_eq!(ctx.name.rank, Rank::Variety);
        assert_eq!(ctx.name.state, State::Partial);
        assert_eq!(ctx.name.unparsed, Some("pallens".to_string()));
        assert!(ctx
            .name
            .warnings
            .contains(&"Removed: subsp. nigra".to_string()));
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::QUADRINOMIAL.to_string()));
    }

    #[test]
    fn abbreviated_genus_sets_informal_type_and_warns() {
        let ctx = run("M. alpium", None);
        assert_eq!(ctx.name.genus, Some("M.".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alpium".to_string()));
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::ABBREVIATED_GENUS.to_string()));
    }

    #[test]
    fn digit_leading_epithet_is_lowercased_into_specific_epithet() {
        let ctx = run("Adalia 11-punctata", None);
        assert_eq!(ctx.name.genus, Some("Adalia".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("11-punctata".to_string()));
        assert_eq!(ctx.name.rank, Rank::Species);
    }

    #[test]
    fn standalone_aggregate_marker_sets_ctx_aggregate() {
        let ctx = run("Poecilia sphenops agg.", None);
        assert_eq!(ctx.name.genus, Some("Poecilia".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("sphenops".to_string()));
        assert!(ctx.aggregate);
    }

    #[test]
    fn hyphenated_aggregate_suffix_strips_the_suffix_and_sets_ctx_aggregate() {
        let ctx = run("Poecilia sphenops-group", None);
        assert_eq!(ctx.name.genus, Some("Poecilia".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("sphenops".to_string()));
        assert!(ctx.aggregate);
    }

    #[test]
    fn requested_species_rank_with_an_infraspecific_epithet_warns_subspecies_assigned() {
        let ctx = run("Abies alba subsp. alba", Some(Rank::Species));
        assert_eq!(ctx.name.infraspecific_epithet, Some("alba".to_string()));
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::SUBSPECIES_ASSIGNED.to_string()));
    }

    #[test]
    fn requested_infraspecific_rank_on_a_bare_binomial_is_marked_informal_indetermined() {
        let ctx = run("Abies alba", Some(Rank::Subspecies));
        assert_eq!(ctx.name.genus, Some("Abies".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("alba".to_string()));
        assert_eq!(ctx.name.infraspecific_epithet, None);
        assert_eq!(
            ctx.name.rank,
            Rank::Subspecies,
            "the caller-requested rank overwrites the post-loop SPECIES default"
        );
        assert_eq!(ctx.name.type_, NameType::Informal);
        assert!(ctx
            .name
            .warnings
            .contains(&warnings::INDETERMINED.to_string()));
    }

    #[test]
    fn missing_genus_placeholder_question_mark_is_kept_as_genus() {
        let ctx = run("? gryphoides", None);
        assert_eq!(ctx.name.genus, Some("?".to_string()));
        assert_eq!(ctx.name.specific_epithet, Some("gryphoides".to_string()));
    }

    #[test]
    fn caller_supplied_code_is_unaffected_by_classify() {
        // Sanity check that classify doesn't clobber a field it never touches.
        let mut ctx = ParseContext::new(
            "Abies alba".to_string(),
            None,
            None,
            Some(NomCode::Botanical),
        );
        ctx.tokens = tokenize(&ctx.original.clone());
        let boundary = authorship_split::find_boundary(&ctx.tokens, &ctx);
        classify(&mut ctx, boundary);
        assert_eq!(ctx.name.code, Some(NomCode::Botanical));
    }
}