diffctx 1.14.0

Selects the minimum code an LLM needs to review a git diff: walks the dependency graph outward from changed lines and stops when extra context stops paying for itself
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
use std::path::Path;
use std::sync::Arc;

use once_cell::sync::Lazy;
use regex::Regex;
use rustc_hash::{FxHashMap, FxHashSet};
use serde::Serialize;

use crate::config::render::RENDER;
use crate::types::{Fragment, FragmentId, FragmentKind};

/// Orientation header for the diff-context output: tells the reader *what*
/// changed before they read the fragments. Empty for working-tree / no-commit
/// diffs.
#[derive(Default)]
pub struct ChangeSummary {
    pub commit_message: Option<String>,
    pub changed_files: Vec<String>,
    pub deleted_files: Vec<String>,
    pub renamed_files: Vec<(String, String)>,
    pub lockfile_changes: Vec<String>,
    pub ignored_changes: Vec<String>,
    pub policy_excluded_count: usize,
}

pub fn is_zero(n: &usize) -> bool {
    *n == 0
}

fn serialize_renames<S>(renames: &[(String, String)], serializer: S) -> Result<S::Ok, S::Error>
where
    S: serde::Serializer,
{
    use serde::ser::SerializeSeq;
    let mut seq = serializer.serialize_seq(Some(renames.len()))?;
    for (from, to) in renames {
        let mut m = std::collections::BTreeMap::new();
        m.insert("from", from);
        m.insert("to", to);
        seq.serialize_element(&m)?;
    }
    seq.end()
}

#[derive(Serialize)]
pub struct DiffContextOutput {
    pub name: String,
    #[serde(rename = "type")]
    pub output_type: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub commit_message: Option<String>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub changed_files: Vec<String>,
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub deleted_files: Vec<String>,
    #[serde(
        skip_serializing_if = "Vec::is_empty",
        serialize_with = "serialize_renames"
    )]
    pub renamed_files: Vec<(String, String)>,
    /// Lock files touched by the range. Paths only — the raw hunks are
    /// thousands of tokens of checksum churn for one line of signal (#112).
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub lockfile_changes: Vec<String>,
    /// Changed files withheld by ignore rules. Silent exclusion misreads as
    /// "the diff did not touch this" (#188: a reviewer filed "no tests"
    /// against a change whose tests the tool had filtered).
    #[serde(skip_serializing_if = "Vec::is_empty")]
    pub ignored_changes: Vec<String>,
    /// Files excluded by `.diffctx/ignore` or secret-path policy — count
    /// only, see `ScoredState::policy_excluded_count`.
    #[serde(skip_serializing_if = "is_zero")]
    pub policy_excluded_count: usize,
    pub fragment_count: usize,
    pub fragments: Vec<FragmentEntry>,
    #[serde(skip)]
    pub latency: Option<LatencyBreakdown>,
}

pub struct LatencyBreakdown {
    /// Pre-heavy-phase work: hunk parse, untracked scan, ignore resolution and
    /// the `git diff` calls. Outside every timer until #183, which is why the
    /// reported phases could not be reconciled with the wall clock.
    pub pre_phase_ms: f64,
    pub parse_changed_ms: f64,
    pub universe_walk_ms: f64,
    pub discovery_ms: f64,
    pub parse_discovered_ms: f64,
    pub tokenization_ms: f64,
    /// Typed dependency graph construction: edge builders + dedup + hub
    /// suppression + per-source cap. Carved out of `scoring_ms` (which
    /// used to absorb it) so the cost distribution is truthful. Zero for
    /// BM25 mode (no graph built).
    pub graph_build_ms: f64,
    /// Combined graph build + scoring + selection time. Kept for
    /// backward compatibility with the existing checkpoint schema; the
    /// split values below are the new diagnostic signal.
    pub scoring_selection_ms: f64,
    pub total_ms: f64,
    /// Heavy-phase rank computation only (PPR/EGO/BM25 + relevance
    /// filtering). Graph construction is reported in `graph_build_ms`;
    /// the selection stage is excluded.
    pub scoring_ms: f64,
    /// Selection stage only (lazy greedy / Boltzmann + post-passes).
    pub selection_ms: f64,
    /// Size of the candidate fragment universe handed to the scoring
    /// strategy (after fragment generation + signature variants but
    /// before per-strategy filtering). Surfaces blowup on large repos —
    /// pathological scoring time is correlated with this number, not
    /// with `fragment_count` (which is the *output* size after
    /// selection).
    pub candidate_count: usize,
    /// Edge count of the typed dependency graph used by PPR/EGO. Zero
    /// for BM25 mode (no graph built).
    pub edge_count: usize,
    /// Number of greedy iterations actually executed (selected non-core
    /// fragments). Bounded by `selected.len() - core.len()`. Pairs with
    /// `selection_ms` to spot lazy-heap blowup vs. genuine large output.
    pub greedy_iters: usize,
    /// Edge count after merge + hub suppression, before per-source cap.
    pub edges_before_cap: usize,
    /// Edges discarded by the per-source top-K cap.
    pub edges_dropped_by_cap: usize,
    /// Source nodes whose outgoing edge list was truncated by the cap.
    pub nodes_capped: usize,
    /// The K value applied for the per-source cap.
    pub max_out_edges_per_node: usize,
    /// PPR push iteration was truncated by `max_pushes_cap` before
    /// convergence. When true, `rel_scores` are biased toward seeds
    /// and absolute file_recall on this instance should be flagged
    /// in post-analysis. Always false for non-PPR scoring modes.
    pub ppr_truncated: bool,
    pub ppr_forward_pushes: usize,
    pub ppr_backward_pushes: usize,
    /// Additive stopping certificate: upper bound
    /// (`tau * peak_density * remaining_budget`) on utility foregone by
    /// adaptive stopping. 0 when the greedy loop ended for another
    /// reason (budget exhausted, no candidates, singleton override).
    pub stopping_certificate: f64,
    /// Lifetime peak physical memory of the process, sampled in-process
    /// at the end of the run. 0 when the platform query fails.
    pub peak_rss_bytes: u64,
    /// Per-category (raw, first-seen deduped) edge emission counts from
    /// pass 1 of the two-pass edge build, sorted by category name. Names
    /// the builder category behind near-dense emission blowups (#116).
    /// Empty for BM25 mode (no graph built).
    pub edge_emissions_by_category: Vec<(&'static str, u64, u64)>,
}

#[derive(Serialize, Clone)]
pub struct FragmentEntry {
    pub path: String,
    pub lines: String,
    /// `Some("changed")` for fragments overlapping the diff hunks; omitted
    /// (treated as supporting context) otherwise. This is the single signal a
    /// reader needs to tell the change apart from its context.
    #[serde(skip_serializing_if = "Option::is_none")]
    pub role: Option<String>,
    pub kind: String,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub symbol: Option<String>,
    #[serde(skip_serializing_if = "Option::is_none")]
    pub content: Option<Arc<str>>,
}

struct SymbolPatterns {
    function: Vec<Regex>,
    class: Vec<Regex>,
    r#struct: Vec<Regex>,
    interface: Vec<Regex>,
    r#enum: Vec<Regex>,
    r#impl: Vec<Regex>,
    r#type: Vec<Regex>,
    module: Vec<Regex>,
    section: Vec<Regex>,
}

static SYMBOL_PATTERNS: Lazy<SymbolPatterns> = Lazy::new(|| {
    SymbolPatterns {
    function: vec![
        Regex::new(r"(?m)^\s*(?:async\s+)?def\s+(\w+)\s*\(").unwrap(),
        Regex::new(r"(?m)^\s*(?:export\s+)?(?:async\s+)?function\s+(\w+)\s*[\(<]").unwrap(),
        Regex::new(r"(?m)^\s*(?:export\s+)?(?:const|let|var)\s+(\w+)\s*=\s*(?:async\s+)?(?:\([^)]*\)|\w)\s*=>").unwrap(),
        Regex::new(r"(?m)^func\s+(?:\([^)]+\)\s+)?(\w+)\s*[\(\[]").unwrap(),
        Regex::new(r"(?m)^\s*(?:pub\s+)?(?:async\s+)?fn\s+(\w+)\s*[\(<]").unwrap(),
        Regex::new(r"(?m)^\s*(?:(?:public|private|protected|static)\s+)*\w[\w<>\[\],]*\s+(\w+)\s*\(").unwrap(),
    ],
    class: vec![
        Regex::new(r"(?m)^\s*class\s+(\w+)\s*[:\({\s]").unwrap(),
        Regex::new(r"(?m)^\s*(?:export\s+)?(?:abstract\s+)?class\s+(\w+)").unwrap(),
    ],
    r#struct: vec![
        Regex::new(r"(?m)^\s*(?:pub\s+)?struct\s+(\w+)").unwrap(),
        Regex::new(r"(?m)^\s*type\s+(\w+)\s+struct\s*\{").unwrap(),
    ],
    interface: vec![
        Regex::new(r"(?m)^\s*(?:export\s+)?interface\s+(\w+)").unwrap(),
        Regex::new(r"(?m)^\s*type\s+(\w+)\s+interface\s*\{").unwrap(),
        Regex::new(r"(?m)^\s*(?:pub\s+)?trait\s+(\w+)").unwrap(),
    ],
    r#enum: vec![
        Regex::new(r"(?m)^\s*(?:pub\s+)?enum\s+(\w+)").unwrap(),
        Regex::new(r"(?m)^\s*class\s+(\w+)\s*\(.*Enum\)").unwrap(),
    ],
    r#impl: vec![
        Regex::new(r"(?m)^\s*impl(?:<[^>]+>)?\s+(\w+)").unwrap(),
    ],
    r#type: vec![
        Regex::new(r"(?m)^\s*(?:export\s+)?type\s+(\w+)").unwrap(),
        Regex::new(r"(?m)^\s*type\s+(\w+)\s").unwrap(),
    ],
    module: vec![
        Regex::new(r"(?m)^\s*(?:pub\s+)?mod\s+(\w+)").unwrap(),
        Regex::new(r"(?m)^\s*package\s+(\w+)").unwrap(),
    ],
    section: vec![
        Regex::new(r"(?m)^#{1,6}\s+(\S[^\n]*)$").unwrap(),
    ],
}
});

fn extract_symbol(frag: &Fragment) -> Option<String> {
    let patterns = match frag.kind {
        FragmentKind::Function | FragmentKind::FunctionSignature => &SYMBOL_PATTERNS.function,
        FragmentKind::Class | FragmentKind::ClassSignature => &SYMBOL_PATTERNS.class,
        FragmentKind::Struct | FragmentKind::StructSignature => &SYMBOL_PATTERNS.r#struct,
        FragmentKind::Interface | FragmentKind::InterfaceSignature => &SYMBOL_PATTERNS.interface,
        FragmentKind::Enum | FragmentKind::EnumSignature => &SYMBOL_PATTERNS.r#enum,
        FragmentKind::Impl => &SYMBOL_PATTERNS.r#impl,
        FragmentKind::Type => &SYMBOL_PATTERNS.r#type,
        FragmentKind::Module => &SYMBOL_PATTERNS.module,
        FragmentKind::Section => &SYMBOL_PATTERNS.section,
        _ => return None,
    };

    for pattern in patterns {
        if let Some(caps) = pattern.captures(&frag.content) {
            if let Some(m) = caps.get(1) {
                let result = m.as_str().trim();
                return Some(if frag.kind == FragmentKind::Section {
                    result
                        .chars()
                        .take(RENDER.section_symbol_max_chars)
                        .collect()
                } else {
                    result.to_string()
                });
            }
        }
    }
    None
}

use crate::paths::to_posix_display as normalize_path_separators;

pub(crate) fn get_relative_path(frag: &Fragment, repo_root: &Path) -> String {
    let frag_path = Path::new(frag.path());
    if !frag_path.is_absolute() {
        return normalize_path_separators(frag_path.to_string_lossy());
    }
    normalize_path_separators(
        frag_path
            .strip_prefix(repo_root)
            .unwrap_or(frag_path)
            .to_string_lossy(),
    )
}

fn create_fragment_entry(frag: &Fragment, path_str: &str) -> FragmentEntry {
    let symbol = frag.symbol_name.clone().or_else(|| extract_symbol(frag));
    let content = if frag.content.is_empty() {
        None
    } else {
        Some(Arc::clone(&frag.content))
    };

    FragmentEntry {
        path: path_str.to_string(),
        lines: format!("{}-{}", frag.start_line(), frag.end_line()),
        role: None,
        kind: frag.kind.as_str().to_string(),
        symbol,
        content,
    }
}

/// A fragment carries the `changed` role when it IS a core, or when it is the
/// hunk-window excerpt that was substituted for one. The excerpt's id is not in
/// `core_ids` — it is a synthetic span cut out of the core — so without this the
/// downshift would silently strip the change marker from the output, which is
/// worse than the over-dump it replaces. `locate.rs` already treats `Excerpt`
/// this way; both surfaces must agree.
fn carries_changed_role(frag: &Fragment, core_ids: &FxHashSet<FragmentId>) -> bool {
    core_ids.contains(&frag.id) || frag.kind == FragmentKind::Excerpt
}

/// Collapse a file's fragments (sorted by start line, ties by descending end
/// line) into the rendered entries. Two behaviors:
/// - a same-role fragment fully contained in the running range (`next.end <=
///   end`) is dropped: its content is already covered by the enclosing
///   fragment (e.g. a symbol-level "function" extraction and a hunk-level
///   "chunk" both covering the same edited lines), so keeping it is pure
///   duplication, not additional information.
/// - a same-role fragment that is line-contiguous with the running range
///   (`next.start == end + 1`) is merged into it.
/// Both are lossless on line coverage and remove the per-fragment scaffolding
/// tax that dominates output on one-line/near-duplicate snippets.
fn merge_file_fragments(
    rel_path: &str,
    frags: &[&Fragment],
    core_ids: &FxHashSet<FragmentId>,
) -> Vec<(bool, u32, FragmentEntry)> {
    let mut out: Vec<(bool, u32, FragmentEntry)> = Vec::new();
    let mut i = 0;
    while i < frags.len() {
        let first = frags[i];
        let role_changed = carries_changed_role(first, core_ids);
        let mut end = first.end_line();
        let mut parts: Vec<&str> = vec![first.content.trim_end_matches('\n')];
        let mut uniform_kind = true;
        let mut j = i + 1;
        while j < frags.len() {
            let next = frags[j];
            if carries_changed_role(next, core_ids) != role_changed {
                break;
            }
            if next.end_line() <= end {
                // Fully contained in the range covered so far - redundant.
                j += 1;
            } else if next.start_line() == end + 1 {
                parts.push(next.content.trim_end_matches('\n'));
                uniform_kind &= next.kind == first.kind;
                end = next.end_line();
                j += 1;
            } else {
                break;
            }
        }

        let mut entry = create_fragment_entry(first, rel_path);
        if j > i + 1 {
            entry.lines = format!("{}-{}", first.start_line(), end);
            let merged = parts.join("\n");
            entry.content = if merged.is_empty() {
                None
            } else {
                Some(Arc::from(merged.as_str()))
            };
            // The merged span is no longer what `first` was, and the kind has
            // to stop claiming otherwise. A one-line `function_signature`
            // followed by contiguous body chunks was emitted as a
            // `function_signature` carrying the whole 101-line function — the
            // exact opposite of what a signature means, since it exists as the
            // cheap stand-in when the full fragment misses the budget (#184).
            // `chunk` is the vocabulary's name for a span of lines with no
            // single semantic identity, which is precisely what a mixed run is.
            if !uniform_kind {
                entry.kind = crate::types::FragmentKind::Chunk.as_str().to_string();
            }
        }
        entry.role = role_changed.then(|| "changed".to_string());
        out.push((role_changed, first.start_line(), entry));
        i = j;
    }
    out
}

pub fn build_diff_context_output(
    repo_root: &Path,
    selected: &[Fragment],
    no_content: bool,
    core_ids: &FxHashSet<FragmentId>,
    rel_scores: &FxHashMap<FragmentId, f64>,
    change: ChangeSummary,
) -> DiffContextOutput {
    let mut by_path: FxHashMap<String, Vec<&Fragment>> = FxHashMap::default();
    for frag in selected {
        by_path
            .entry(get_relative_path(frag, repo_root))
            .or_default()
            .push(frag);
    }

    // Changed code first (the answer to "what changed"), then supporting
    // context ordered by descending per-file relevance so the reader's primacy
    // attention lands on the most relevant material, not on alphabetical noise.
    let mut changed: Vec<(String, u32, FragmentEntry)> = Vec::new();
    let mut context: Vec<(f64, String, u32, FragmentEntry)> = Vec::new();
    for (rel_path, frags) in &by_path {
        let mut sorted: Vec<&Fragment> = frags.clone();
        // Tie-break by descending end line so, among same-start fragments, the
        // widest range sorts first and containment-absorption below (which scans
        // forward from the first entry of a run) sees the enclosing range before
        // any of its nested sub-fragments.
        sorted.sort_by_key(|f| (f.start_line(), std::cmp::Reverse(f.end_line())));
        let file_rel = sorted
            .iter()
            .map(|f| rel_scores.get(&f.id).copied().unwrap_or(0.0))
            .fold(0.0_f64, f64::max);
        for (role_changed, start, mut entry) in merge_file_fragments(rel_path, &sorted, core_ids) {
            if no_content {
                entry.content = None;
            }
            if role_changed {
                changed.push((rel_path.clone(), start, entry));
            } else {
                context.push((file_rel, rel_path.clone(), start, entry));
            }
        }
    }

    changed.sort_by(|a, b| a.0.cmp(&b.0).then(a.1.cmp(&b.1)));
    context.sort_by(|a, b| {
        b.0.partial_cmp(&a.0)
            .unwrap_or(std::cmp::Ordering::Equal)
            .then(a.1.cmp(&b.1))
            .then(a.2.cmp(&b.2))
    });

    let mut fragments_out: Vec<FragmentEntry> = Vec::with_capacity(changed.len() + context.len());
    fragments_out.extend(changed.into_iter().map(|(_, _, e)| e));
    fragments_out.extend(context.into_iter().map(|(_, _, _, e)| e));

    let resolved = repo_root
        .canonicalize()
        .unwrap_or_else(|_| repo_root.to_path_buf());
    let name = resolved
        .file_name()
        .map(|n| n.to_string_lossy().to_string())
        .unwrap_or_else(|| resolved.to_string_lossy().to_string());

    DiffContextOutput {
        name,
        output_type: "diff_context".to_string(),
        commit_message: change.commit_message,
        changed_files: change.changed_files,
        deleted_files: change.deleted_files,
        renamed_files: change.renamed_files,
        lockfile_changes: change.lockfile_changes,
        ignored_changes: change.ignored_changes,
        policy_excluded_count: change.policy_excluded_count,
        fragment_count: fragments_out.len(),
        fragments: fragments_out,
        latency: None,
    }
}

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

    fn empty_output(renamed_files: Vec<(String, String)>) -> DiffContextOutput {
        DiffContextOutput {
            name: "repo".to_string(),
            output_type: "diff_context".to_string(),
            commit_message: None,
            changed_files: Vec::new(),
            deleted_files: Vec::new(),
            renamed_files,
            lockfile_changes: Vec::new(),
            ignored_changes: Vec::new(),
            policy_excluded_count: 0,
            fragment_count: 0,
            fragments: Vec::new(),
            latency: None,
        }
    }

    #[test]
    fn renamed_files_serialize_as_labelled_from_to_in_yaml() {
        let out = empty_output(vec![("old.py".to_string(), "new.py".to_string())]);
        let yaml = serde_yaml::to_string(&out).unwrap();
        assert!(
            yaml.contains("from: old.py"),
            "expected labelled `from:` entry, got:\n{yaml}"
        );
        assert!(
            yaml.contains("to: new.py"),
            "expected labelled `to:` entry, got:\n{yaml}"
        );
        // Guards against serde's default tuple-as-two-element-sequence shape
        // (`- - old.py\n  - new.py`), which drops the from/to labels.
        assert!(!yaml.contains("- - old.py"));
    }

    #[test]
    fn renamed_files_serialize_as_labelled_from_to_in_json() {
        let out = empty_output(vec![("old.py".to_string(), "new.py".to_string())]);
        let json = serde_json::to_value(&out).unwrap();
        let renamed = json["renamed_files"]
            .as_array()
            .expect("renamed_files must serialize as an array");
        assert_eq!(renamed.len(), 1);
        assert_eq!(renamed[0]["from"], "old.py");
        assert_eq!(renamed[0]["to"], "new.py");
        assert!(
            renamed[0].is_object(),
            "must not serialize as a positional [old, new] tuple: {renamed:?}"
        );
    }

    #[test]
    fn renamed_files_empty_is_omitted_from_output() {
        let out = empty_output(Vec::new());
        let json = serde_json::to_value(&out).unwrap();
        assert!(json.get("renamed_files").is_none());
    }

    fn frag_at(path: &str) -> Fragment {
        Fragment {
            id: FragmentId::new(Arc::from(path), 1, 5),
            kind: FragmentKind::Function,
            content: Arc::from(""),
            identifiers: FxHashSet::default(),
            token_count: 1,
            symbol_name: None,
        }
    }

    #[cfg(unix)]
    #[test]
    fn get_relative_path_posix_backslash_in_filename_round_trips_unchanged() {
        // On POSIX `\` is a legal filename character: `src\utils.py` is one
        // file, not `src/utils.py` in a subdirectory. Rewriting the
        // separator here would report a path that does not exist.
        let frag = frag_at("src\\utils.py");
        let root = Path::new("/repo");
        let rel = get_relative_path(&frag, root);
        assert_eq!(rel, "src\\utils.py");
    }

    #[cfg(unix)]
    #[test]
    fn get_relative_path_strips_repo_root_on_posix() {
        let frag = frag_at("/repo/src/lib.rs");
        let root = Path::new("/repo");
        let rel = get_relative_path(&frag, root);
        assert_eq!(rel, "src/lib.rs");
    }
}

#[cfg(test)]
mod merge_kind_tests {
    use super::*;
    use crate::types::{FragmentId, FragmentKind};

    fn frag(start: u32, end: u32, kind: FragmentKind, body: &str) -> Fragment {
        Fragment {
            id: FragmentId::new(Arc::from("a.py"), start, end),
            kind,
            content: Arc::from(body),
            identifiers: FxHashSet::default(),
            token_count: 10,
            symbol_name: None,
        }
    }

    /// The defect this guards (#184): a one-line signature followed by
    /// contiguous body chunks was emitted as a `function_signature` carrying the
    /// whole function. A signature exists to be the cheap stand-in when the full
    /// fragment misses the budget, so one that holds the body is the opposite of
    /// its own contract — and `drop_redundant_signatures` decides using that
    /// contract.
    #[test]
    fn a_merged_run_of_mixed_kinds_does_not_claim_the_first_kind() {
        let frags = vec![
            frag(1, 1, FragmentKind::FunctionSignature, "def big(a):"),
            frag(2, 3, FragmentKind::Chunk, "    x = 1\n    y = 2"),
        ];
        let refs: Vec<&Fragment> = frags.iter().collect();
        let out = merge_file_fragments("a.py", &refs, &FxHashSet::default());

        assert_eq!(out.len(), 1, "contiguous fragments should merge into one");
        assert_eq!(out[0].2.kind, "chunk");
        assert_eq!(out[0].2.lines, "1-3");
    }

    /// A run that is genuinely all one kind keeps it — the rule is about the
    /// label becoming false, not about merging itself.
    #[test]
    fn a_merged_run_of_one_kind_keeps_it() {
        let frags = vec![
            frag(1, 2, FragmentKind::Chunk, "a\nb"),
            frag(3, 4, FragmentKind::Chunk, "c\nd"),
        ];
        let refs: Vec<&Fragment> = frags.iter().collect();
        let out = merge_file_fragments("a.py", &refs, &FxHashSet::default());

        assert_eq!(out.len(), 1);
        assert_eq!(out[0].2.kind, "chunk");
    }

    /// An unmerged fragment is untouched: nothing about it became untrue.
    #[test]
    fn a_lone_fragment_keeps_its_kind() {
        let frags = vec![frag(1, 1, FragmentKind::FunctionSignature, "def big(a):")];
        let refs: Vec<&Fragment> = frags.iter().collect();
        let out = merge_file_fragments("a.py", &refs, &FxHashSet::default());

        assert_eq!(out[0].2.kind, "function_signature");
    }
}