proef-core 0.5.3

Engine-agnostic core of proef: parsing, binding, lowering, IR, emit, dispatch, World, events, errors
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
//! Collect-all suite analysis: the LSP's whole-suite recompute in one function.
//!
//! Where `front::run` fails fast and emits artifacts, `analyze_suite`
//! accumulates every diagnostic and emits the relations editors need
//! (`bindings` for go-to-def/references, `macros` for completion/def targets).
//! A parse-failed unit reports its own diagnostic and is skipped downstream —
//! no cascade of bogus follow-on errors.

use std::collections::BTreeMap;
use std::sync::Arc;

use crate::bind;
use crate::diag::{Diag, Span};
use crate::emit;
use crate::engine::StepKindSpec;
use crate::feature;
use crate::lower::{self, LowerCtx};
use crate::pack::{self, MacroBody, MacroStepKind, PackSet, PackSource};
use crate::provider::SourceProvider;
use crate::world::{GlobalStore, World};

/// One prose step bound to a macro — powers go-to-definition and references.
#[derive(Debug, Clone)]
pub struct Binding {
    /// Source name of the feature file the step lives in.
    pub feature: String,
    /// Byte span of the step text in the *normalized* feature source.
    pub step_span: Span,
    /// The macro this step resolved to.
    pub macro_name: String,
}

/// A macro definition — powers completion and is the go-to-definition target.
#[derive(Debug, Clone)]
pub struct MacroRef {
    /// Macro name (globally unique across the loaded packs).
    pub name: String,
    /// The `match:` pattern (`None` for `use:`-only macros).
    pub pattern: Option<String>,
    /// Declared params, in declaration order.
    pub params: Vec<String>,
    /// Source name of the pack the macro is defined in.
    pub pack: String,
    /// Byte span of the macro's name key in the *normalized* pack source, when
    /// locatable. This is the definition anchor go-to-definition jumps to.
    pub def_span: Option<Span>,
    /// Byte span of the macro's `match:` line, when locatable — the preferred
    /// go-to-definition landing anchor (falls back to `def_span`).
    pub match_span: Option<Span>,
}

/// One `use:` reference inside a pack → the macro it resolves to. Powers
/// go-to-definition from a `use:` line to the target macro's definition.
#[derive(Debug, Clone)]
pub struct UseRef {
    /// Source name of the pack the `use:` line lives in.
    pub pack: String,
    /// Byte span of the `use:` line in the *normalized* pack source.
    pub span: Span,
    /// The macro the reference resolves to (globally unique name).
    pub target_macro: String,
}

/// The product of one wholesale recompute: every feature's read from here.
#[derive(Debug, Default)]
pub struct SuiteAnalysis {
    /// source name → its diagnostics (features and packs alike).
    pub diagnostics: BTreeMap<String, Vec<Diag>>,
    /// Every prose-step-to-macro binding across the suite.
    pub bindings: Vec<Binding>,
    /// Every macro definition across the loaded packs.
    pub macros: Vec<MacroRef>,
    /// Every `use:` reference across the loaded packs, resolved to its target.
    pub use_refs: Vec<UseRef>,
}

/// Everything `analyze_suite` needs, injected at the IO edge (sans-IO core).
pub struct AnalyzeCtx<'a> {
    /// The source of feature and pack bytes (the IO edge lives behind it).
    pub provider: &'a dyn SourceProvider,
    /// Registered engine step kinds (drives pack validation and artifact probes).
    pub kinds: &'a [StepKindSpec],
    /// Step-kind prefix → engine id, the lowering routing table.
    pub kind_to_engine: &'a BTreeMap<String, String>,
    /// Injected environment snapshot (`${env:…}`).
    pub env: &'a BTreeMap<String, String>,
    /// Injected `proef.toml` config scope (`${url:…}` / `${vars:…}`), with the
    /// active `[env.<name>]` already deep-merged in.
    pub config_vars: &'a BTreeMap<String, String>,
    /// Injected run identifier (`${run:id}`).
    pub run_id: &'a str,
}

impl SuiteAnalysis {
    fn push_diags(&mut self, name: &str, diags: impl IntoIterator<Item = Diag>) {
        // Ensure the primary source has a bucket even with no diagnostics, so a
        // now-clean file still surfaces an empty set that clears stale marks.
        self.diagnostics.entry(name.to_owned()).or_default();
        for d in diags {
            // Prefer the diagnostic's own source name when it carries one, so a
            // pack error raised while analyzing a feature lands on the pack.
            let target = d.source_name.clone().unwrap_or_else(|| name.to_owned());
            self.diagnostics.entry(target).or_default().push(d);
        }
    }
}

/// Recompute the whole suite in one pass: read every pack and feature through
/// the provider, accumulate every diagnostic per source name, and record the
/// binding and macro relations editors need. A broken pack contributes its own
/// diagnostic and is excluded from the loaded set, but does not stop the rest
/// of the suite from binding; a parse-failed feature is skipped, not fatal.
pub fn analyze_suite(ctx: &AnalyzeCtx<'_>) -> SuiteAnalysis {
    let mut out = SuiteAnalysis::default();

    // Packs first: a broken pack contributes its own diagnostic below and is
    // excluded from the loaded set, but its siblings still load.
    let mut sources = pack::builtin_sources();
    let pack_names = ctx.provider.discover_packs().unwrap_or_default();
    for name in &pack_names {
        match ctx.provider.read(name) {
            Ok(text) => sources.push(PackSource {
                name: name.clone(),
                text,
            }),
            Err(e) => out.push_diags(name, [read_error_diag(name, &e.0)]),
        }
    }

    // Collect-all load: a broken pack contributes its diagnostic and is
    // excluded from the set, but its siblings still load — the editor keeps
    // binding against the good packs instead of going dark (v0.5.1 fix).
    let (loaded, pack_diags) = pack::load_collecting(&sources, ctx.kinds);
    for d in pack_diags {
        let name = d.source_name.clone().unwrap_or_default();
        out.push_diags(&name, [d]);
    }
    let packs: Arc<PackSet> = Arc::new(loaded);

    // Macro vocabulary for completion / go-to-def targets.
    for m in packs.macros.values() {
        out.macros.push(MacroRef {
            name: m.name.clone(),
            pattern: m.pattern.clone(),
            params: m.params.clone(),
            pack: m.pack.clone(),
            def_span: m.span,
            match_span: m.match_span,
        });
    }

    out.use_refs = index_use_refs(&packs);

    let world = World::new(GlobalStore::default());

    let feature_names = ctx.provider.discover_features().unwrap_or_default();
    for name in &feature_names {
        let text = match ctx.provider.read(name) {
            Ok(t) => t,
            Err(e) => {
                out.push_diags(name, [read_error_diag(name, &e.0)]);
                continue;
            }
        };
        let file = match feature::parse(name, &text) {
            Ok(f) => f,
            Err(errs) => {
                out.push_diags(name, errs);
                continue; // parse failed → skip downstream, no cascade
            }
        };

        let (bound, bind_diags) = bind::bind_collect(&file, &packs);
        out.push_diags(name, bind_diags);

        for scenario in &bound {
            for step in &scenario.steps {
                out.bindings.push(Binding {
                    feature: name.clone(),
                    step_span: step.defn.span,
                    macro_name: step.macro_name.clone(),
                });
            }
        }

        let ctx_lower = LowerCtx {
            feature: &file,
            packs: &packs,
            kind_to_engine: ctx.kind_to_engine,
            env: ctx.env,
            config_vars: ctx.config_vars,
            run_id: ctx.run_id,
            world: &world,
            mode: crate::resolve::ResolveMode::DryRun,
        };
        for scenario in &bound {
            match lower::lower(scenario, &ctx_lower) {
                Ok(lowered) => {
                    out.push_diags(name, lowered.warnings.iter().cloned());
                    // Emit + artifact validation is executed for its diagnostics
                    // only; the artifact text is discarded.
                    let stem = feature_stem(name);
                    if let Some(artifact) = emit::emit(&lowered, &stem, &world) {
                        let mut diags = Vec::new();
                        validate_artifact(&artifact, &lowered, ctx.kinds, &mut diags);
                        out.push_diags(name, diags);
                    }
                }
                Err(errs) => out.push_diags(name, errs),
            }
        }
    }

    out
}

/// Index every `use:` reference → its resolved target, for go-to-def from a
/// `use:` line. Each macro's parsed `MacroStepKind::Use` targets pair positionally
/// with the `use:` line spans `pack::locate::use_line_spans` finds. Because both
/// counts come from the macro's own steps and source, a mismatch means the line
/// scanner missed a step it can't see (e.g. a flow-style `- {use: base}`), so the
/// pairing is unreliable and that macro is skipped entirely rather than risk a
/// wrong `Some`.
fn index_use_refs(packs: &PackSet) -> Vec<UseRef> {
    let mut use_refs = Vec::new();
    for m in packs.macros.values() {
        let MacroBody::Steps(steps) = &m.body else {
            continue;
        };
        let targets: Vec<&str> = steps
            .iter()
            .filter_map(|step| match &step.kind {
                MacroStepKind::Use { target, .. } => Some(target.as_str()),
                MacroStepKind::Payload { .. } => None,
            })
            .collect();
        let spans = crate::pack::locate::use_line_spans(&m.source, &m.name);
        if spans.len() != targets.len() {
            continue; // line scan and parsed steps disagree → pairing unreliable, skip
        }
        for (span, target) in spans.into_iter().zip(targets) {
            if let Some(target_macro) = packs.find_use_target(target) {
                use_refs.push(UseRef {
                    pack: m.pack.clone(),
                    span,
                    target_macro: target_macro.name.clone(),
                });
            }
        }
    }
    use_refs
}

fn feature_stem(name: &str) -> String {
    std::path::Path::new(name).file_stem().map_or_else(
        || "feature".to_owned(),
        |s| s.to_string_lossy().into_owned(),
    )
}

fn read_error_diag(name: &str, msg: &str) -> Diag {
    Diag::error(
        "proef::source::unreadable",
        format!("cannot read {name}: {msg}"),
    )
    .with_source(name.to_owned(), Arc::from(""))
}

/// Parse-validate the exact emitted artifact text with the claiming engine's
/// real parser (`--dry-run` = §4.1–4.5 including artifact parse-validation).
/// The diagnostic's source is the emitted text itself, span at the broken line.
///
/// This is the single implementation of artifact parse-validation, shared by the
/// CLI's fail-fast `front::run` and the LSP's collect-all `analyze_suite`. It
/// reaches the hurl parser only through the injected [`StepKindSpec::validate`]
/// function pointer, so it stays engine-agnostic and lives in the sans-IO core.
pub fn validate_artifact(
    artifact: &emit::Artifact,
    lowered: &lower::LoweredScenario,
    kinds: &[StepKindSpec],
    diags: &mut Vec<Diag>,
) {
    let Some(kind) = lowered
        .batches
        .iter()
        .flat_map(|b| b.steps.iter())
        .find(|s| matches!(s.payload, crate::step::StepPayload::HurlEntries(_)))
        .map(|s| s.kind.as_str().to_owned())
    else {
        return;
    };
    let Some(validate) = kinds
        .iter()
        .find(|k| k.prefix == kind)
        .and_then(|k| k.validate)
    else {
        return;
    };
    if let Err(err) = validate(&artifact.hurl_text) {
        let offset: usize = artifact
            .hurl_text
            .split_inclusive('\n')
            .take(err.line.saturating_sub(1))
            .map(str::len)
            .sum();
        let line_len = artifact.hurl_text[offset..]
            .lines()
            .next()
            .unwrap_or("")
            .len();
        diags.push(
            Diag::error(
                "proef::emit::invalid_artifact",
                format!(
                    "emitted artifact `{}.hurl` does not parse: {} (line {}, column {})",
                    artifact.slug, err.message, err.line, err.column
                ),
            )
            .with_source(
                format!("{}.hurl (emitted)", artifact.slug),
                std::sync::Arc::from(artifact.hurl_text.as_str()),
            )
            .with_span(Span::clamped(
                offset,
                offset + line_len.max(1),
                artifact.hurl_text.len(),
            )),
        );
    }
}

#[cfg(test)]
mod tests {
    #![allow(clippy::expect_used)]

    use super::*;
    use crate::provider::{ProviderError, SourceProvider};
    use std::collections::BTreeMap;
    use std::sync::Arc;

    /// A provider backed by an in-memory map — keeps the test sans-IO.
    struct MemProvider {
        features: Vec<String>,
        packs: Vec<String>,
        files: BTreeMap<String, Arc<str>>,
    }
    impl SourceProvider for MemProvider {
        fn discover_features(&self) -> Result<Vec<String>, ProviderError> {
            Ok(self.features.clone())
        }
        fn discover_packs(&self) -> Result<Vec<String>, ProviderError> {
            Ok(self.packs.clone())
        }
        fn read(&self, name: &str) -> Result<Arc<str>, ProviderError> {
            self.files
                .get(name)
                .cloned()
                .ok_or_else(|| ProviderError(format!("no source {name}")))
        }
    }

    // Pack validation (step-kind claim, pass 8) and lowering's routing invariant
    // both require `hurl` to be a registered kind: with an empty registry a
    // `hurl:` step is `unknown_step_kind` and an unrouted lowered step. The
    // analyzer needs no *live* engine, though — a spec with no `validate` probe
    // (artifact parse-validation is skipped) plus a `hurl → hurl` route is enough
    // to bind, lower, and extract spans.
    const KINDS: &[StepKindSpec] = &[StepKindSpec {
        prefix: "hurl",
        schema: "true",
        validate: None,
    }];

    fn hurl_kind_map() -> &'static BTreeMap<String, String> {
        use std::sync::OnceLock;
        static M: OnceLock<BTreeMap<String, String>> = OnceLock::new();
        M.get_or_init(|| BTreeMap::from([("hurl".to_owned(), "hurl".to_owned())]))
    }

    fn ctx_over<'a>(
        provider: &'a dyn SourceProvider,
        empty: &'a BTreeMap<String, String>,
    ) -> AnalyzeCtx<'a> {
        AnalyzeCtx {
            provider,
            kinds: KINDS,
            kind_to_engine: hurl_kind_map(),
            env: empty,
            config_vars: empty,
            run_id: "lsp",
        }
    }

    #[test]
    fn analyze_surfaces_bindings_and_no_errors_on_a_clean_suite() {
        let mut files = BTreeMap::new();
        files.insert(
            "packs/p.yaml".to_owned(),
            Arc::from(
                "macros:\n  greet:\n    params: [who]\n    match: \"I greet {who}\"\n    steps:\n      - hurl: |\n          GET http://x\n",
            ),
        );
        files.insert(
            "f.feature".to_owned(),
            Arc::from("Feature: F\n  Scenario: S\n    When I greet Sam\n"),
        );
        let provider = MemProvider {
            features: vec!["f.feature".to_owned()],
            packs: vec!["packs/p.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));

        let errors: usize = analysis
            .diagnostics
            .values()
            .flatten()
            .filter(|d| d.severity == crate::diag::Severity::Error)
            .count();
        assert_eq!(
            errors, 0,
            "clean suite must have zero errors: {:?}",
            analysis.diagnostics
        );

        assert!(
            analysis
                .bindings
                .iter()
                .any(|b| b.macro_name == "greet" && b.feature == "f.feature"),
            "the greet step must be recorded as a binding"
        );
        assert!(
            analysis
                .macros
                .iter()
                .any(|m| m.name == "greet" && m.pattern.is_some())
        );
    }

    #[test]
    fn analyze_collects_unbound_without_cascade() {
        let mut files = BTreeMap::new();
        files.insert("packs/p.yaml".to_owned(), Arc::from("macros: {}\n"));
        files.insert(
            "f.feature".to_owned(),
            Arc::from("Feature: F\n  Scenario: S\n    When nothing matches this\n"),
        );
        let provider = MemProvider {
            features: vec!["f.feature".to_owned()],
            packs: vec!["packs/p.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));
        let feature_diags = analysis
            .diagnostics
            .get("f.feature")
            .expect("feature bucket");
        assert!(
            feature_diags
                .iter()
                .any(|d| d.code == "proef::bind::unbound_step")
        );
        let errors: Vec<_> = feature_diags
            .iter()
            .filter(|d| d.severity == crate::diag::Severity::Error)
            .collect();
        assert_eq!(
            errors.len(),
            1,
            "the unbound step must be the only error-severity diagnostic, no spurious extras: {feature_diags:?}"
        );
        assert_eq!(errors[0].code, "proef::bind::unbound_step");
    }

    // §9's headline robustness property: a parse failure suppresses only its own
    // file's downstream diagnostics — it must not cascade into a sibling feature's
    // binding. Two features share one valid pack; only one of them fails to parse.
    #[test]
    fn analyze_parse_failed_feature_does_not_cascade_to_sibling() {
        let mut files = BTreeMap::new();
        files.insert(
            "packs/p.yaml".to_owned(),
            Arc::from(
                "macros:\n  greet:\n    params: [who]\n    match: \"I greet {who}\"\n    steps:\n      - hurl: |\n          GET http://x\n",
            ),
        );
        // Whitespace-only text: feature::parse's own empty-file guard
        // (`normalized.trim().is_empty()`) returns `Err` before the gherkin
        // parser even runs — a guaranteed, deliberate parse failure.
        files.insert("bad.feature".to_owned(), Arc::from("   \n"));
        files.insert(
            "good.feature".to_owned(),
            Arc::from("Feature: F\n  Scenario: S\n    When I greet Sam\n"),
        );
        let provider = MemProvider {
            features: vec!["bad.feature".to_owned(), "good.feature".to_owned()],
            packs: vec!["packs/p.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));

        // The parse-failed feature carries its own parse-error diagnostic —
        // proof the `feature::parse` `Err` branch was actually hit.
        let bad_diags = analysis
            .diagnostics
            .get("bad.feature")
            .expect("bad.feature bucket");
        assert!(
            bad_diags
                .iter()
                .any(|d| d.code == "proef::feature::empty_file"),
            "the parse-failed feature must carry its parse-error diagnostic: {bad_diags:?}"
        );

        // The sibling feature is untouched: no error diagnostics, and its
        // binding still made it through — proof there was no cascade.
        let good_diags = analysis
            .diagnostics
            .get("good.feature")
            .expect("good.feature bucket");
        assert!(
            good_diags
                .iter()
                .all(|d| d.severity != crate::diag::Severity::Error),
            "the valid sibling feature must have no error diagnostics despite the parse failure next to it: {good_diags:?}"
        );
        assert!(
            analysis
                .bindings
                .iter()
                .any(|b| b.macro_name == "greet" && b.feature == "good.feature"),
            "the valid sibling feature must still produce its binding — no cascade from the parse-failed feature"
        );
    }

    // A broken pack degrades gracefully: it reports its own error, but the good
    // pack still loads, so a feature binding against a good-pack macro survives.
    // One broken pack must never zero the whole suite's analysis (v0.5.1 fix).
    #[test]
    fn analyze_degrades_when_one_pack_is_broken() {
        let mut files = BTreeMap::new();
        files.insert(
            "packs/good.yaml".to_owned(),
            Arc::from(
                "macros:\n  greet:\n    params: [who]\n    match: \"I greet {who}\"\n    steps:\n      - hurl: |\n          GET http://x\n",
            ),
        );
        // `bogus` is not a recognized root key → deny_unknown_fields → this pack
        // fails to parse and contributes proef::pack::yaml, but must not sink the rest.
        files.insert(
            "packs/broken.yaml".to_owned(),
            Arc::from("macros: {}\nbogus: true\n"),
        );
        files.insert(
            "f.feature".to_owned(),
            Arc::from("Feature: F\n  Scenario: S\n    When I greet Sam\n"),
        );
        let provider = MemProvider {
            features: vec!["f.feature".to_owned()],
            packs: vec!["packs/good.yaml".to_owned(), "packs/broken.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));

        // The broken pack still reports its own diagnostic.
        let pack_diags = analysis
            .diagnostics
            .get("packs/broken.yaml")
            .expect("broken pack bucket");
        assert!(
            pack_diags.iter().any(|d| d.code == "proef::pack::yaml"),
            "the broken pack must carry its yaml diagnostic: {pack_diags:?}"
        );

        // The good pack still loaded: its macro is in the vocabulary...
        assert!(
            analysis.macros.iter().any(|m| m.name == "greet"),
            "the good pack's macro must survive the broken sibling"
        );
        // ...and the feature bound against it — no cascade, no zeroing.
        assert!(
            analysis
                .bindings
                .iter()
                .any(|b| b.macro_name == "greet" && b.feature == "f.feature"),
            "the feature must still bind to the good-pack macro despite the broken pack"
        );
    }

    #[test]
    fn analyze_records_use_refs_and_match_spans() {
        let mut files = BTreeMap::new();
        files.insert(
            "packs/p.yaml".to_owned(),
            Arc::from(
                "macros:\n  base:\n    match: the base\n    steps:\n      - hurl: |\n          GET http://x\n  wrapper:\n    steps:\n      - use: base\n",
            ),
        );
        let provider = MemProvider {
            features: vec![],
            packs: vec!["packs/p.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));

        // The `use: base` line is indexed, resolved to `base`.
        let u = analysis
            .use_refs
            .iter()
            .find(|u| u.target_macro == "base")
            .expect("use_ref for base");
        assert_eq!(u.pack, "packs/p.yaml");
        let src = provider_text(&provider, "packs/p.yaml");
        assert_eq!(&src[u.span.start..u.span.end], "use: base");

        // `base` carries a match_span; `wrapper` (use-only) does not.
        let base = analysis
            .macros
            .iter()
            .find(|m| m.name == "base")
            .expect("macro base");
        assert!(base.match_span.is_some());
        let wrapper = analysis
            .macros
            .iter()
            .find(|m| m.name == "wrapper")
            .expect("macro wrapper");
        assert!(wrapper.match_span.is_none());
    }

    // Guards the ordinal alignment between parsed `Use` steps and textual `use:`
    // lines: a flow-style step (`- {use: base}`) is valid YAML and parses to a
    // `Use` step, but the line scanner's `use:`-prefix match does not see it (the
    // line reads `- {use: base}`, not a `use:`-prefixed line after the dash
    // strip). Mixed with a block-style `use:` line in the same macro, the parsed
    // count (2) and textual count (1) diverge — per-ordinal pairing would silently
    // attribute the wrong textual span to a step. `index_use_refs` must skip
    // `UseRef` generation for that macro entirely rather than emit a wrong `Some`.
    #[test]
    fn analyze_skips_use_refs_when_flow_and_block_style_counts_diverge() {
        let mut files = BTreeMap::new();
        files.insert(
            "packs/p.yaml".to_owned(),
            Arc::from(
                "macros:\n  base:\n    match: the base\n    steps:\n      - hurl: |\n          GET http://x\n  wrapper:\n    steps:\n      - {use: base}\n      - use: base\n",
            ),
        );
        let provider = MemProvider {
            features: vec![],
            packs: vec!["packs/p.yaml".to_owned()],
            files,
        };
        let empty = BTreeMap::new();
        let analysis = analyze_suite(&ctx_over(&provider, &empty));

        // The pack is valid (flow-style `use:` is legal YAML) — no error
        // diagnostics, so `analyze_suite` did not short-circuit before indexing.
        let errors: usize = analysis
            .diagnostics
            .values()
            .flatten()
            .filter(|d| d.severity == crate::diag::Severity::Error)
            .count();
        assert_eq!(
            errors, 0,
            "the mixed-style pack must be valid, zero errors: {:?}",
            analysis.diagnostics
        );

        // `base` has no `use:` steps of its own, so any `UseRef` in this suite
        // would have to come from `wrapper` — the count mismatch must suppress
        // all of them (no wrong-target/wrong-span `UseRef`, per the "never a
        // wrong `Some`" contract).
        assert!(
            analysis.use_refs.is_empty(),
            "count mismatch must skip UseRef generation for wrapper entirely, \
             not emit a misaligned pairing: {:?}",
            analysis.use_refs
        );
    }

    // Small helper to read a source back for span assertions.
    fn provider_text(p: &MemProvider, name: &str) -> String {
        p.read(name).expect("provider source").to_string()
    }
}