ara-core 0.1.7

Shared parsing, normalization, and DAG layout for the ARA viewer runtime.
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
//! Lenient reader for the `evidence/` layer plus node→exhibit / node→RW
//! resolution.
//!
//! The evidence layer is an index (`evidence/README.md`) plus body files under
//! `evidence/figures/*.md` and `evidence/tables/*.md`. The corpus index tables
//! drift heavily — eight distinct header shapes across the 34 real artifacts,
//! some with no claims column, one reordering columns, others using `Key refs`
//! or `Used by` in place of `Claims`. So [`parse_index`] is **column-name
//! tolerant**: it identifies columns by header substring, never by position.
//!
//! Everything is warn-never-fatal and source-order preserving. Bodies are stored
//! verbatim (`body`); rendering (tables, images) is a client concern. The two
//! resolution passes ([`resolve_node_exhibits`], [`resolve_built_on`]) are pure
//! and deterministic, iterating nodes in manifest order and exhibits / related
//! work in source order.

use std::collections::BTreeSet;

use crate::manifest::{
    Binding, BindingRole, BuiltOn, ClaimId, Exhibit, ExhibitKind, Node, NodeExhibit, NodeId,
    RelatedWork, is_canonical_id,
};

// ── index (`evidence/README.md`) ─────────────────────────────────────────────

/// One parsed index row, keyed by a normalized basename `id`. Fields are
/// whatever the row's columns carried; any may be absent.
#[derive(Debug, Clone, PartialEq)]
pub(crate) struct IndexRow {
    /// Normalized basename id (file stem), used to match a body file.
    pub id: String,
    /// The raw file-cell path, when the row carried a resolvable file cell.
    pub file: Option<String>,
    /// Origin string from a `source`-like column.
    pub source: Option<String>,
    /// Description from a `description`/`what`-like column.
    pub description: Option<String>,
    /// `C##` tokens from a `claims`/`key ref`/`used by` column.
    pub claims: Vec<ClaimId>,
}

/// Parses an `evidence/README.md` into index rows, in source order.
///
/// A README may hold several GFM tables (commonly `## Tables` and `## Figures`);
/// each is parsed independently and its rows appended. Columns are identified by
/// header name, so reordered or renamed columns still resolve. Non-table prose
/// is ignored.
pub(crate) fn parse_index(md: &str) -> Vec<IndexRow> {
    let lines: Vec<&str> = md.lines().collect();
    let mut rows = Vec::new();

    let mut i = 0;
    while i < lines.len() {
        // A table starts where a `|`-row is followed by a delimiter row.
        if is_table_row(lines[i]) && i + 1 < lines.len() && is_delimiter_row(lines[i + 1]) {
            let headers = split_row(lines[i]);
            let cols = ColumnMap::from_headers(&headers);
            let mut j = i + 2;
            while j < lines.len() && is_table_row(lines[j]) {
                let cells = split_row(lines[j]);
                if let Some(row) = cols.build_row(&cells) {
                    rows.push(row);
                }
                j += 1;
            }
            i = j;
        } else {
            i += 1;
        }
    }

    rows
}

/// Which column carries which field, resolved by header name.
struct ColumnMap {
    file: Option<usize>,
    claims: Option<usize>,
    source: Option<usize>,
    description: Option<usize>,
}

impl ColumnMap {
    /// Resolves columns from a header row by case-insensitive substring match.
    fn from_headers(headers: &[String]) -> Self {
        let lower: Vec<String> = headers.iter().map(|h| h.to_ascii_lowercase()).collect();
        let find = |pred: &dyn Fn(&str) -> bool| lower.iter().position(|h| pred(h));

        // File: first header containing `file`; fall back to column 0.
        let file = find(&|h| h.contains("file")).or(if lower.is_empty() { None } else { Some(0) });
        // Claims: `claim` | `key ref` | `used by`.
        let claims =
            find(&|h| h.contains("claim") || h.contains("key ref") || h.contains("used by"));
        let source = find(&|h| h.contains("source"));
        let description = find(&|h| h.contains("desc") || h.contains("what"));

        ColumnMap {
            file,
            claims,
            source,
            description,
        }
    }

    /// Builds an [`IndexRow`] from a data row's cells. Returns `None` when the
    /// row has no usable id (empty first/file cell).
    fn build_row(&self, cells: &[String]) -> Option<IndexRow> {
        let cell = |idx: Option<usize>| idx.and_then(|k| cells.get(k)).map(|s| s.trim());

        let file_cell = cell(self.file)?;
        if file_cell.is_empty() {
            return None;
        }
        let (id, file) = normalize_file_cell(file_cell);
        if id.is_empty() {
            return None;
        }

        let claims = self
            .claims
            .and_then(|k| cells.get(k))
            .map(|c| extract_claim_ids(c))
            .unwrap_or_default();

        Some(IndexRow {
            id,
            file,
            source: cell(self.source).and_then(non_empty),
            description: cell(self.description).and_then(non_empty),
            claims,
        })
    }
}

/// True for a line that looks like a GFM table row (`| ... |`).
fn is_table_row(line: &str) -> bool {
    line.trim_start().starts_with('|')
}

/// True for a GFM delimiter row: only `|`, `-`, `:`, and spaces, with at least
/// one `-`.
fn is_delimiter_row(line: &str) -> bool {
    let t = line.trim();
    if !t.starts_with('|') {
        return false;
    }
    let mut saw_dash = false;
    for c in t.chars() {
        match c {
            '-' => saw_dash = true,
            '|' | ':' | ' ' | '\t' => {}
            _ => return false,
        }
    }
    saw_dash
}

/// Splits a `| a | b |` row into trimmed cell strings, dropping the empty edges
/// created by the leading/trailing pipes.
fn split_row(line: &str) -> Vec<String> {
    let t = line.trim();
    let t = t.strip_prefix('|').unwrap_or(t);
    let t = t.strip_suffix('|').unwrap_or(t);
    t.split('|').map(|c| c.trim().to_string()).collect()
}

/// Normalizes a file cell to `(id, file)`. Handles a markdown link
/// `[x](path/foo.md)`, a backtick `` `table1.md` ``, a bare path `figures/foo.md`,
/// and a dual-ext `foo (png/md)` / `foo.(png/md)`. `id` is the basename stem;
/// `file` is the raw relative path when one is recoverable (a prose "fact" cell
/// yields `id` = slug, `file` = `None`).
fn normalize_file_cell(cell: &str) -> (String, Option<String>) {
    let cell = cell.trim();

    // Markdown link: take the link target.
    if let Some(path) = markdown_link_target(cell) {
        return (file_stem_id(&path), Some(path));
    }
    // Backtick-fenced path.
    let unticked = cell.trim_matches('`').trim();
    // A path-like cell contains a `/` or ends in a recognizable extension.
    if looks_like_path(unticked) {
        return (file_stem_id(unticked), Some(unticked.to_string()));
    }
    // Prose cell (e.g. a "Fact" column): no file, id is a slug for warnings.
    (slug(cell), None)
}

/// The `(target)` of a `[text](target)` markdown link, if the cell is one.
fn markdown_link_target(cell: &str) -> Option<String> {
    let open = cell.find("](")?;
    let rest = &cell[open + 2..];
    let close = rest.find(')')?;
    let target = rest[..close].trim();
    if target.is_empty() {
        None
    } else {
        Some(target.to_string())
    }
}

/// True when a cell looks like a file path rather than free prose: it has a `/`
/// separator or a short alphanumeric-ish extension, and no interior spaces
/// beyond an optional dual-ext marker.
fn looks_like_path(s: &str) -> bool {
    if s.is_empty() {
        return false;
    }
    if s.contains('/') {
        return true;
    }
    // A dual-ext marker like `foo (png/md)` or a trailing `.md`.
    if s.contains("(png") || s.contains("(md") || s.contains("/md)") || s.contains("/png)") {
        return true;
    }
    // A single trailing extension with no spaces in the name.
    match s.rsplit_once('.') {
        Some((stem, ext)) => {
            !stem.is_empty()
                && !stem.contains(' ')
                && (1..=5).contains(&ext.len())
                && ext.chars().all(|c| c.is_ascii_alphanumeric())
        }
        None => false,
    }
}

/// The basename stem of a path, stripping directories, a dual-ext `(png/md)`
/// marker, and a single trailing extension.
fn file_stem_id(path: &str) -> String {
    // Trim once up front so every byte offset below indexes the same string
    // (a leading-whitespace mismatch could otherwise slice mid-char and panic).
    let path = path.trim();
    // Strip a trailing dual-ext group like `.(png/md)` or ` (png/md)` FIRST — it
    // may itself contain a `/` that would otherwise corrupt the basename split.
    let path = match path.rfind('(') {
        Some(open) => {
            let tail = &path[open..];
            if tail.ends_with(')')
                && (tail.contains("md") || tail.contains("png") || tail.contains('/'))
            {
                path[..open].trim_end_matches([' ', '.'])
            } else {
                path
            }
        }
        None => path,
    };

    let base = path
        .rsplit(['/', '\\'])
        .next()
        .unwrap_or(path)
        .trim()
        .to_string();

    // Strip a single trailing extension (`.md`, `.png`, …).
    match base.rsplit_once('.') {
        Some((stem, ext))
            if !stem.is_empty()
                && !ext.is_empty()
                && ext.chars().all(|c| c.is_ascii_alphanumeric()) =>
        {
            stem.to_string()
        }
        _ => base,
    }
}

/// A conservative slug of a prose cell, used only as a warning id when a row
/// carries no file. Lowercased, non-alphanumerics collapsed to `-`.
fn slug(s: &str) -> String {
    let mut out = String::new();
    let mut prev_dash = false;
    for c in s.chars().take(48) {
        if c.is_ascii_alphanumeric() {
            out.push(c.to_ascii_lowercase());
            prev_dash = false;
        } else if !prev_dash && !out.is_empty() {
            out.push('-');
            prev_dash = true;
        }
    }
    out.trim_matches('-').to_string()
}

/// Extracts every `C\d+` token from a cell, ignoring `H##`/`N##`/prose tokens.
fn extract_claim_ids(value: &str) -> Vec<ClaimId> {
    let mut seen = BTreeSet::new();
    value
        .split(|c: char| !c.is_ascii_alphanumeric())
        .filter(|tok| is_canonical_id(tok, 'C'))
        .filter(|tok| seen.insert(tok.to_string()))
        .map(ClaimId::new)
        .collect()
}

/// Trims and returns `None` for empty values.
fn non_empty(s: &str) -> Option<String> {
    let t = s.trim();
    if t.is_empty() {
        None
    } else {
        Some(t.to_string())
    }
}

// ── body files + assembly (native) ───────────────────────────────────────────

/// Reads the `evidence/` layer of `dir` into exhibits, appending warnings.
///
/// Enumerates `evidence/figures/*.md` then `evidence/tables/*.md` (each sorted),
/// building one [`Exhibit`] per body file. Index rows enrich the matching body
/// by basename id (index wins over body for source/description). An index row
/// with no matching body, or a body with no index row, warns but never errors.
/// An absent `evidence/` dir or absent `README.md` is a silent skip.
#[cfg(feature = "native")]
pub(crate) fn read_evidence(
    dir: &std::path::Path,
    report: &mut crate::report::ParseReport,
) -> Vec<Exhibit> {
    let evidence_dir = dir.join("evidence");
    if !evidence_dir.is_dir() {
        return Vec::new();
    }

    // Index is optional: bodies still yield exhibits, just without index enrichment.
    let index: Vec<IndexRow> = std::fs::read_to_string(evidence_dir.join("README.md"))
        .ok()
        .map(|md| parse_index(&md))
        .unwrap_or_default();
    // First index row wins per id (deterministic).
    let mut index_by_id: std::collections::BTreeMap<String, IndexRow> =
        std::collections::BTreeMap::new();
    for row in &index {
        index_by_id
            .entry(row.id.clone())
            .or_insert_with(|| row.clone());
    }

    let mut exhibits = Vec::new();
    let mut consumed: BTreeSet<String> = BTreeSet::new();

    for (subdir, kind) in [
        ("figures", ExhibitKind::Figure),
        ("tables", ExhibitKind::Table),
    ] {
        for path in sorted_md_files(&evidence_dir.join(subdir)) {
            let Ok(body) = std::fs::read_to_string(&path) else {
                continue;
            };
            let id = path
                .file_stem()
                .map(|s| s.to_string_lossy().into_owned())
                .unwrap_or_default();
            let file = format!(
                "evidence/{subdir}/{}",
                path.file_name()
                    .map(|s| s.to_string_lossy())
                    .unwrap_or_default()
            );

            let row = index_by_id.get(&id);
            if let Some(row) = row {
                consumed.insert(row.id.clone());
            } else {
                report.warn(
                    format!("evidence/{subdir}/{id}"),
                    "body file has no index row in evidence/README.md",
                );
            }

            let exhibit = assemble_exhibit(id, file, kind.clone(), row, &body);
            exhibits.push(exhibit);
        }
    }

    // Any index row that never matched a body file (and pointed at one) warns.
    for row in &index {
        if row.file.is_some() && !consumed.contains(&row.id) {
            report.warn(
                format!("evidence[{}]", row.id),
                "index row references a file with no body under evidence/",
            );
        }
    }

    exhibits
}

/// Merges an index row (if any) and a body into one exhibit. `source`/
/// `description` prefer the index, falling back to the body's `- **Source**:` /
/// `- **Caption**:` bullets. `claims` are the index row's `C##`, or — when the
/// index has none — the `Supports: C##` refs scanned from the body.
fn assemble_exhibit(
    id: String,
    file: String,
    kind: ExhibitKind,
    row: Option<&IndexRow>,
    body: &str,
) -> Exhibit {
    let source = row
        .and_then(|r| r.source.clone())
        .or_else(|| body_bullet(body, "source"));
    let description = row
        .and_then(|r| r.description.clone())
        .or_else(|| body_bullet(body, "caption"));

    let claims = match row.map(|r| r.claims.clone()).unwrap_or_default() {
        c if !c.is_empty() => c,
        _ => body_supports(body),
    };

    Exhibit {
        id,
        file,
        kind,
        source,
        description,
        claims,
        body: body.to_string(),
    }
}

/// The value of a `- **Label**: value` bullet in a body, matched by
/// case-insensitive label. Returns the first such bullet's value.
fn body_bullet(body: &str, label: &str) -> Option<String> {
    for line in body.lines() {
        let t = line.trim_start();
        let Some(rest) = t.strip_prefix("- ").or_else(|| t.strip_prefix("* ")) else {
            continue;
        };
        let Some(after) = rest.trim_start().strip_prefix("**") else {
            continue;
        };
        let Some((key, tail)) = after.split_once("**") else {
            continue;
        };
        if key.trim().eq_ignore_ascii_case(label) {
            let tail = tail.trim_start();
            let value = tail.strip_prefix(':').unwrap_or(tail).trim();
            if let Some(v) = non_empty(value) {
                return Some(v);
            }
        }
    }
    None
}

/// `C##` refs from any body line mentioning `Supports` (the nanogpt_ara
/// convention, where claim linkage lives in the body, not the index).
fn body_supports(body: &str) -> Vec<ClaimId> {
    let mut out = Vec::new();
    let mut seen = BTreeSet::new();
    for line in body.lines() {
        if !line.contains("Supports") {
            continue;
        }
        for id in extract_claim_ids(line) {
            if seen.insert(id.as_str().to_string()) {
                out.push(id);
            }
        }
    }
    out
}

/// Enumerates `*.md` files in `dir`, sorted by path. Missing dir → empty.
#[cfg(feature = "native")]
fn sorted_md_files(dir: &std::path::Path) -> Vec<std::path::PathBuf> {
    let Ok(entries) = std::fs::read_dir(dir) else {
        return Vec::new();
    };
    let mut paths: Vec<std::path::PathBuf> = entries
        .flatten()
        .map(|e| e.path())
        .filter(|p| p.is_file() && p.extension().is_some_and(|ext| ext == "md"))
        .collect();
    paths.sort();
    paths
}

// ── resolution passes (pure) ─────────────────────────────────────────────────

/// Node → exhibit edges by shared claim. For each node (in `nodes` order) whose
/// evidence claim set intersects an exhibit's `claims` (exhibits in source
/// order), emits one [`NodeExhibit`]. Deterministic; each (node, exhibit) once.
pub(crate) fn resolve_node_exhibits(
    nodes: &[Node],
    bindings: &[Binding],
    exhibits: &[Exhibit],
) -> Vec<NodeExhibit> {
    let mut out = Vec::new();
    let mut seen: BTreeSet<(NodeId, String)> = BTreeSet::new();
    for node in nodes {
        let node_claims = evidence_claims(&node.id, bindings);
        if node_claims.is_empty() {
            continue;
        }
        for exhibit in exhibits {
            if exhibit.claims.iter().any(|c| node_claims.contains(c)) {
                let key = (node.id.clone(), exhibit.id.clone());
                if seen.insert(key) {
                    out.push(NodeExhibit {
                        node: node.id.clone(),
                        exhibit: exhibit.id.clone(),
                    });
                }
            }
        }
    }
    out
}

/// Node → related-work edges by shared claim. Same shape as
/// [`resolve_node_exhibits`], matching a node's evidence claims against each
/// related-work entry's `claims_affected`.
pub(crate) fn resolve_built_on(
    nodes: &[Node],
    bindings: &[Binding],
    related_work: &[RelatedWork],
) -> Vec<BuiltOn> {
    let mut out = Vec::new();
    let mut seen: BTreeSet<(NodeId, String)> = BTreeSet::new();
    for node in nodes {
        let node_claims = evidence_claims(&node.id, bindings);
        if node_claims.is_empty() {
            continue;
        }
        for rw in related_work {
            if rw.claims_affected.iter().any(|c| node_claims.contains(c)) {
                let key = (node.id.clone(), rw.id.clone());
                if seen.insert(key) {
                    out.push(BuiltOn {
                        node: node.id.clone(),
                        related_work: rw.id.clone(),
                    });
                }
            }
        }
    }
    out
}

/// The set of claims a node references via `Evidence`-role bindings.
fn evidence_claims(node: &NodeId, bindings: &[Binding]) -> BTreeSet<ClaimId> {
    bindings
        .iter()
        .filter(|b| &b.node == node && b.role == BindingRole::Evidence)
        .map(|b| b.claim.clone())
        .collect()
}

#[cfg(test)]
mod tests {
    use super::*;

    fn claims(ids: &[&str]) -> Vec<ClaimId> {
        ids.iter().map(|s| ClaimId::new(*s)).collect()
    }

    #[test]
    fn index_canonical_file_source_claims_description() {
        let md = "\
## Figures
| File | Source | Claims | Description |
|------|--------|--------|-------------|
| [figures/fig3_scalability.md](figures/fig3_scalability.md) | Figure 3, §4.3 | C01, C04 | Growth demo. |
";
        let rows = parse_index(md);
        assert_eq!(rows.len(), 1);
        assert_eq!(rows[0].id, "fig3_scalability");
        assert_eq!(rows[0].claims, claims(&["C01", "C04"]));
        assert_eq!(rows[0].source.as_deref(), Some("Figure 3, §4.3"));
        assert_eq!(rows[0].description.as_deref(), Some("Growth demo."));
    }

    #[test]
    fn index_reordered_columns_claims_in_col4() {
        // `File | Description | Source | Claims` — Claims moved to the last column.
        let md = "\
| File | Description | Source | Claims |
|------|-------------|--------|--------|
| [tables/reference_scores.md](tables/reference_scores.md) | ref scores | README | C01, C05 |
| [tables/human_baselines.md](tables/human_baselines.md) | humans | README:42-52 | — |
";
        let rows = parse_index(md);
        assert_eq!(rows.len(), 2);
        assert_eq!(rows[0].id, "reference_scores");
        assert_eq!(rows[0].claims, claims(&["C01", "C05"]));
        assert_eq!(rows[0].description.as_deref(), Some("ref scores"));
        assert_eq!(rows[0].source.as_deref(), Some("README"));
        // Em-dash claims cell → no claims.
        assert_eq!(rows[1].id, "human_baselines");
        assert!(rows[1].claims.is_empty());
    }

    #[test]
    fn index_key_refs_header_carries_claims() {
        let md = "\
| File | Description | Key refs |
|------|-------------|---------|
| [tables/reference_scores.md](tables/reference_scores.md) | ref | C01, C05, C12 |
";
        let rows = parse_index(md);
        assert_eq!(rows[0].claims, claims(&["C01", "C05", "C12"]));
    }

    #[test]
    fn index_used_by_no_file_no_claims_column() {
        // `Fact | Source turns | Used by` — no File column (id falls back to col 0
        // slug, file None) and `Used by` carries the C## refs.
        let md = "\
| Fact | Source turns | Used by |
|------|--------------|---------|
| Move model A1 up, 5-cell steps | 0->4 | C01, C02, H01-H05 |
";
        let rows = parse_index(md);
        assert_eq!(rows.len(), 1);
        assert!(rows[0].file.is_none());
        assert!(!rows[0].id.is_empty());
        assert_eq!(rows[0].claims, claims(&["C01", "C02"])); // H## ignored
    }

    #[test]
    fn index_backtick_and_dual_ext_file_cells() {
        let md = "\
| File | Type | Source object | What it shows |
|---|---|---|---|
| `tables/trajectory_summary.md` | table | run index | the arc |
| figures/v1_loss_curves.(png/md) | quantitative_plot | v1 png | loss curve |
";
        let rows = parse_index(md);
        assert_eq!(rows[0].id, "trajectory_summary");
        assert_eq!(rows[0].description.as_deref(), Some("the arc")); // `what` column
        assert_eq!(rows[1].id, "v1_loss_curves"); // dual-ext stripped
        assert_eq!(rows[1].source.as_deref(), Some("v1 png")); // `source object`
    }

    #[test]
    fn index_description_only_no_claims() {
        let md = "\
| File | Description |
|---|---|
| [tables/reference_scores.md](tables/reference_scores.md) | Starting score 1.81 |
";
        let rows = parse_index(md);
        assert_eq!(rows.len(), 1);
        assert_eq!(rows[0].id, "reference_scores");
        assert!(rows[0].claims.is_empty());
        assert_eq!(rows[0].description.as_deref(), Some("Starting score 1.81"));
    }

    #[test]
    fn index_multiple_tables_parsed_independently() {
        let md = "\
## Tables
| File | Source | Claims | Description |
|------|--------|--------|-------------|
| [tables/t1.md](tables/t1.md) | T1 | C02 | table one |

## Figures
| File | Source | Claims | Description |
|------|--------|--------|-------------|
| [figures/f1.md](figures/f1.md) | F1 | C03 | fig one |
";
        let rows = parse_index(md);
        assert_eq!(rows.len(), 2);
        assert_eq!(rows[0].id, "t1");
        assert_eq!(rows[1].id, "f1");
    }

    #[test]
    fn empty_index_is_empty() {
        assert!(parse_index("").is_empty());
        assert!(parse_index("# Just prose\n\nNo tables here.\n").is_empty());
    }

    #[test]
    fn body_bullet_source_and_caption() {
        let body = "# Fig\n- **Source**: Figure 3, Section 4.3\n- **Caption**: \"a caption\"\n";
        assert_eq!(
            body_bullet(body, "source").as_deref(),
            Some("Figure 3, Section 4.3")
        );
        assert_eq!(
            body_bullet(body, "caption").as_deref(),
            Some("\"a caption\"")
        );
        assert!(body_bullet(body, "missing").is_none());
    }

    #[test]
    fn body_supports_scans_claim_refs() {
        let body = "**Source:** `x`. Supports C06, C11; figure\n";
        assert_eq!(body_supports(body), claims(&["C06", "C11"]));
    }

    #[test]
    fn resolve_node_exhibits_matches_on_shared_claim() {
        let nodes = vec![node("N01"), node("N02")];
        let bindings = vec![binding("N01", "C01")];
        let exhibits = vec![
            exhibit("figA", &["C01", "C04"]),
            exhibit("figB", &["C02"]),
            exhibit("figC", &["C04", "C01"]),
        ];
        let out = resolve_node_exhibits(&nodes, &bindings, &exhibits);
        // N01 (claims {C01}) matches figA and figC in source order; N02 has none.
        assert_eq!(
            out.iter().map(|n| n.exhibit.as_str()).collect::<Vec<_>>(),
            vec!["figA", "figC"]
        );
        assert!(out.iter().all(|n| n.node == NodeId::new("N01")));
    }

    #[test]
    fn resolve_built_on_matches_on_shared_claim() {
        let nodes = vec![node("N07")];
        let bindings = vec![binding("N07", "C01")];
        let rw = vec![
            related_work("RW01", &["C01", "C04"]),
            related_work("RW02", &["C02", "C03"]),
            related_work("RW09", &["C01", "C02"]),
        ];
        let out = resolve_built_on(&nodes, &bindings, &rw);
        assert_eq!(
            out.iter()
                .map(|b| b.related_work.as_str())
                .collect::<Vec<_>>(),
            vec!["RW01", "RW09"]
        );
    }

    #[test]
    fn resolve_empty_when_node_has_no_matching_exhibit() {
        let nodes = vec![node("N01")];
        let bindings = vec![binding("N01", "C99")];
        let exhibits = vec![exhibit("figA", &["C01"])];
        assert!(resolve_node_exhibits(&nodes, &bindings, &exhibits).is_empty());
    }

    // ── test helpers ─────────────────────────────────────────────────────────

    fn node(id: &str) -> Node {
        Node {
            id: NodeId::new(id),
            kind: crate::manifest::NodeKind::Experiment,
            label: None,
            support_level: None,
            source_refs: Vec::new(),
            description: None,
            fields: crate::manifest::NodeFields::Experiment { result: None },
            evidence_notes: Vec::new(),
            isolated: false,
            pos: None,
        }
    }

    fn binding(node: &str, claim: &str) -> Binding {
        Binding {
            node: NodeId::new(node),
            claim: ClaimId::new(claim),
            role: BindingRole::Evidence,
        }
    }

    fn exhibit(id: &str, claim_ids: &[&str]) -> Exhibit {
        Exhibit {
            id: id.to_string(),
            file: format!("evidence/figures/{id}.md"),
            kind: ExhibitKind::Figure,
            source: None,
            description: None,
            claims: claims(claim_ids),
            body: String::new(),
        }
    }

    fn related_work(id: &str, claim_ids: &[&str]) -> RelatedWork {
        RelatedWork {
            id: id.to_string(),
            cite: String::new(),
            doi: None,
            kind: None,
            what_changed: None,
            why: None,
            adopted: None,
            claims_affected: claims(claim_ids),
        }
    }
}