pmat 3.15.0

PMAT - Zero-config AI context generation and code quality toolkit (CLI, MCP, HTTP)
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
/// Compute PV Lint sub-score using pv score D1-D5 dimensions (pv-compatibility spec §2.1).
///
/// Aligns pmat scoring with the provable-contracts scoring model:
///   D1: Specification Depth (20%)
///   D2: Falsification Coverage (25%)
///   D3: Kani Proof Coverage (25%)
///   D4: Lean Proof Coverage (10%)
///   D5: Binding Coverage (20%)
///
/// Falls back to pipeline-depth heuristic if pv CLI is unavailable.
fn compute_pv_lint(path: &Path) -> f64 {
    let contracts_dir = path.join("contracts");
    if !contracts_dir.exists() {
        let pmat_yaml = path.join(".pmat.yaml");
        if pmat_yaml.exists() {
            if let Ok(content) = std::fs::read_to_string(&pmat_yaml) {
                if content.contains("cb-1202") && content.contains("enabled: false") {
                    return 50.0;
                }
            }
        }
        let src_dir = path.join("src");
        if src_dir.exists() {
            let critical = ["forward", "backward", "optimizer", "checkpoint",
                "sampling", "kv_cache", "quantize", "kernel", "matmul", "gemm"];
            let has_critical = critical.iter().any(|kw| {
                walkdir::WalkDir::new(&src_dir).into_iter().flatten()
                    .filter(|e| e.path().extension().is_some_and(|ext| ext == "rs"))
                    .any(|e| std::fs::read_to_string(e.path())
                        .map(|c| c.contains(&format!("pub fn {kw}")))
                        .unwrap_or(false))
            });
            if has_critical {
                return 0.0;
            }
        }
        return 50.0;
    }

    // Strategy 1: Use `pv score` for D1-D5 dimensions (pv-compatibility spec §2.1)
    let pv_score = score_via_pv_score(path);
    if let Some(score) = pv_score {
        // Gate check: pv lint must also pass
        let lint_ok = check_pv_lint_gates(path);
        if !lint_ok {
            return score * 0.5; // Halve score if lint gates fail
        }
        return score;
    }

    // Strategy 2: Fallback to pipeline depth heuristic (pv CLI unavailable)
    let pipeline = compute_pipeline_depth(path);
    // Scale pipeline (0-30) to (50-95) range
    (50.0 + pipeline * 1.5).clamp(50.0, 95.0)
}

/// Call `pv score <contracts_dir> --format json` and parse D1-D5 dimensions.
/// Returns score on 0-100 scale, or None if pv unavailable.
fn score_via_pv_score(path: &Path) -> Option<f64> {
    let contracts_dir = path.join("contracts");
    let output = std::process::Command::new("pv")
        .args(["score", &contracts_dir.display().to_string(), "--format", "json"])
        .current_dir(path)
        .stdout(std::process::Stdio::piped())
        .stderr(std::process::Stdio::null())
        .output()
        .ok()?;

    if !output.status.success() && output.status.code() != Some(1) {
        return None;
    }

    let content = String::from_utf8(output.stdout).ok()?;
    let val: serde_json::Value = serde_json::from_str(&content).ok()?;

    // Parse mean_score (0.0-1.0) from pv score output
    let contract_count = val.get("contracts").and_then(|c| c.as_u64()).unwrap_or(0);
    if contract_count == 0 {
        return None; // No contracts — fall through to heuristic
    }
    let mean_score = val.get("mean_score").and_then(|s| s.as_f64())?;

    // Also extract per-dimension averages for diagnostics
    if let Some(scores) = val.get("scores").and_then(|s| s.as_array()) {
        if !scores.is_empty() {
            let n = scores.len() as f64;
            let avg_d3 = scores.iter()
                .filter_map(|s| s.get("kani_coverage").and_then(|v| v.as_f64()))
                .sum::<f64>() / n;
            let avg_d4 = scores.iter()
                .filter_map(|s| s.get("lean_coverage").and_then(|v| v.as_f64()))
                .sum::<f64>() / n;
            if avg_d3 > 0.0 || avg_d4 > 0.0 {
                eprintln!(
                    "  PV Score: {:.1}% (Kani avg: {:.0}%, Lean avg: {:.0}%)",
                    mean_score * 100.0, avg_d3 * 100.0, avg_d4 * 100.0
                );
            }
        }
    }

    // Blend pv score (contract quality) with pipeline depth (integration quality)
    // pv score measures D1-D5 of the contracts themselves
    // pipeline depth measures how well the codebase integrates them
    let pipeline = compute_pipeline_depth(path);
    let pipeline_norm = pipeline / 30.0; // 0.0-1.0

    // 40% contract quality (pv score D1-D5) + 60% integration depth (pipeline)
    // pmat measures the integration quality; pv score measures the contract quality.
    // Projects with full pipeline (build.rs + macros + Lean refs) deserve high scores
    // even if the contracts themselves still have gaps (lean_coverage, binding_coverage).
    let blended = mean_score * 0.4 + pipeline_norm * 0.6;
    Some((blended * 100.0).clamp(0.0, 100.0))
}

/// Quick check: does `pv lint` pass all gates?
fn check_pv_lint_gates(path: &Path) -> bool {
    let output = std::process::Command::new("pv")
        .args(["lint", "--format", "json"])
        .current_dir(path)
        .stdout(std::process::Stdio::piped())
        .stderr(std::process::Stdio::null())
        .output();

    match output {
        Ok(o) => {
            if let Ok(content) = String::from_utf8(o.stdout) {
                if let Ok(val) = serde_json::from_str::<serde_json::Value>(&content) {
                    return val.get("passed").and_then(|p| p.as_bool()).unwrap_or(false);
                }
            }
            o.status.success()
        }
        Err(_) => true, // pv unavailable — don't penalize
    }
}

/// Score the contract pipeline depth (0-30 points).
///
/// Checks how deeply integrated the provable-contracts pipeline is:
/// - YAML contracts exist (5 pts)
/// - build.rs reads YAML and emits PRE/POST env vars (5 pts)
/// - #[contract] macros on production functions (5 pts)
/// - lean_theorem references in YAML (5 pts)
/// - Kani harnesses defined (5 pts)
/// - proof-status.json available from provable-contracts (5 pts)
fn compute_pipeline_depth(path: &Path) -> f64 {
    let mut score = 0.0;
    let contracts_dir = path.join("contracts");

    // 1. YAML contracts exist (5 pts)
    let yaml_count = std::fs::read_dir(&contracts_dir)
        .map(|entries| {
            entries
                .flatten()
                .filter(|e| e.path().extension().is_some_and(|ext| ext == "yaml"))
                .count()
        })
        .unwrap_or(0);
    if yaml_count > 0 {
        score += 5.0;
    }

    // 2. build.rs reads contracts (5 pts)
    let build_rs = path.join("build.rs");
    if build_rs.exists() {
        if let Ok(content) = std::fs::read_to_string(&build_rs) {
            if content.contains("PRE_COUNT") || content.contains("emit_contract_assertions") {
                score += 5.0;
            }
        }
    }

    // 3. #[contract] macros in source (5 pts)
    let src_dir = path.join("src");
    if src_dir.exists() {
        let has_contract = walkdir::WalkDir::new(&src_dir)
            .into_iter()
            .flatten()
            .filter(|e| e.path().extension().is_some_and(|ext| ext == "rs"))
            .any(|e| {
                std::fs::read_to_string(e.path())
                    .map(|c| c.contains("::contract(") || c.contains("#[contract("))
                    .unwrap_or(false)
            });
        if has_contract {
            score += 5.0;
        }
    }

    // 4. lean_theorem references in YAML (5 pts)
    if yaml_count > 0 {
        let has_lean_ref = std::fs::read_dir(&contracts_dir)
            .map(|entries| {
                entries.flatten().any(|e| {
                    e.path().extension().is_some_and(|ext| ext == "yaml")
                        && std::fs::read_to_string(e.path())
                            .map(|c| c.contains("lean_theorem:"))
                            .unwrap_or(false)
                })
            })
            .unwrap_or(false);
        if has_lean_ref {
            score += 5.0;
        }
    }

    // 5. Kani harnesses defined (5 pts)
    if yaml_count > 0 {
        let has_kani = std::fs::read_dir(&contracts_dir)
            .map(|entries| {
                entries.flatten().any(|e| {
                    e.path().extension().is_some_and(|ext| ext == "yaml")
                        && std::fs::read_to_string(e.path())
                            .map(|c| c.contains("kani_harnesses:"))
                            .unwrap_or(false)
                })
            })
            .unwrap_or(false);
        if has_kani {
            score += 5.0;
        }
    }

    // 6. proof-status.json — parse L4/L5 verification levels (0-5 pts)
    let abs_path = std::fs::canonicalize(path).unwrap_or_else(|_| path.to_path_buf());
    let proof_status_path = abs_path
        .parent()
        .map(|p| p.join("provable-contracts").join("proof-status.json"));
    if let Some(ps_path) = proof_status_path {
        if let Ok(content) = std::fs::read_to_string(&ps_path) {
            if let Ok(val) = serde_json::from_str::<serde_json::Value>(&content) {
                let totals = val.get("totals");
                let obligations = totals.and_then(|t| t.get("obligations")).and_then(|v| v.as_u64()).unwrap_or(0);
                let kani = totals.and_then(|t| t.get("kani_harnesses")).and_then(|v| v.as_u64()).unwrap_or(0);
                let lean = totals.and_then(|t| t.get("lean_proved")).and_then(|v| v.as_u64()).unwrap_or(0);
                if obligations > 0 {
                    // Score based on L4+L5 coverage: kani/obligations + lean/obligations
                    let l4_ratio = (kani as f64 / obligations as f64).min(1.0);
                    let l5_ratio = (lean as f64 / obligations as f64).min(1.0);
                    // L4 (Kani) worth 3 pts, L5 (Lean) worth 2 pts
                    score += l4_ratio * 3.0 + l5_ratio * 2.0;
                } else {
                    score += 2.0; // File exists but no obligations — partial credit
                }
            }
        }
    }

    score
}

/// CD5: Contract drift detection. Compares contract YAML mtimes vs source file mtimes.
 // Available for future CD5 scoring integration
/// Returns (stale_count, total_count, drift_ratio).
/// A contract is "stale" if the YAML is >30 days older than the most recently modified
/// source file that references it (via #[contract] annotation).
fn compute_contract_drift(path: &Path) -> (usize, usize, f64) {
    let contracts_dir = path.join("contracts");
    if !contracts_dir.exists() {
        return (0, 0, 0.0);
    }

    let mut stale = 0usize;
    let mut total = 0usize;
    let thirty_days = std::time::Duration::from_secs(30 * 24 * 3600);

    let Ok(entries) = std::fs::read_dir(&contracts_dir) else {
        return (0, 0, 0.0);
    };

    for entry in entries.flatten() {
        let p = entry.path();
        if p.extension().map_or(true, |e| e != "yaml") { continue; }
        if p.file_name().is_some_and(|n| n.to_string_lossy().contains("binding")) { continue; }

        let Ok(yaml_meta) = std::fs::metadata(&p) else { continue };
        let Ok(yaml_mtime) = yaml_meta.modified() else { continue };

        total += 1;

        // Find the newest source file that references this contract
        let stem = p.file_stem().and_then(|s| s.to_str()).unwrap_or("");
        let search_pattern = format!("\"{stem}\"");

        let mut newest_src = None;
        let src_dirs: Vec<std::path::PathBuf> = if path.join("src").exists() {
            vec![path.join("src")]
        } else if path.join("crates").exists() {
            std::fs::read_dir(path.join("crates"))
                .into_iter().flatten().flatten()
                .filter_map(|e| { let s = e.path().join("src"); s.exists().then_some(s) })
                .collect()
        } else {
            vec![]
        };

        for sdir in &src_dirs {
            for e in walkdir::WalkDir::new(sdir).into_iter().flatten() {
                if !e.path().extension().is_some_and(|ext| ext == "rs") { continue; }
                if let Ok(content) = std::fs::read_to_string(e.path()) {
                    if content.contains(&search_pattern) {
                        if let Ok(meta) = std::fs::metadata(e.path()) {
                            if let Ok(mtime) = meta.modified() {
                                if newest_src.map_or(true, |n| mtime > n) {
                                    newest_src = Some(mtime);
                                }
                            }
                        }
                    }
                }
            }
        }

        // If source is significantly newer than contract, it's stale
        if let Some(src_mtime) = newest_src {
            if src_mtime > yaml_mtime {
                if let Ok(diff) = src_mtime.duration_since(yaml_mtime) {
                    if diff > thirty_days {
                        stale += 1;
                    }
                }
            }
        }
    }

    let drift = if total > 0 { stale as f64 / total as f64 } else { 0.0 };
    (stale, total, drift)
}

fn compute_file_health(path: &Path) -> f64 {
    // Count files by size category for a density-based score
    let src_dir = path.join("src");
    if !src_dir.exists() {
        return 100.0;
    }
    let mut total = 0usize;
    let mut over_1000 = 0usize;

    for entry in walkdir::WalkDir::new(&src_dir).into_iter().flatten() {
        if entry.path().extension().map_or(true, |e| e != "rs") {
            continue;
        }
        total += 1;
        if let Ok(content) = std::fs::read_to_string(entry.path()) {
            if content.lines().count() > 1000 {
                over_1000 += 1;
            }
        }
    }

    if total == 0 {
        return 100.0;
    }

    // Score based on percentage of files under 1000 lines
    let pct_healthy = (1.0 - (over_1000 as f64 / total as f64)) * 100.0;
    pct_healthy.clamp(0.0, 100.0)
}

fn geometric_mean(values: &[f64]) -> f64 {
    if values.is_empty() {
        return 0.0;
    }
    // Use log-space to avoid overflow
    let n = values.len() as f64;
    let log_sum: f64 = values
        .iter()
        .map(|v| if *v > 0.0 { v.ln() } else { f64::NEG_INFINITY })
        .sum();
    if log_sum == f64::NEG_INFINITY {
        return 0.0;
    }
    (log_sum / n).exp()
}

/// Kani bounded model checking proofs for scoring functions.
/// Run: `cargo kani --harness verify_geometric_mean_bounded`
#[cfg(kani)]
mod kani_proofs {
    use super::geometric_mean;

    /// Prove: geometric_mean always returns a value in [0, 100] for inputs in [0, 100].
    /// Exhaustively checks all f64 values within the bound (Kani explores symbolically).
    #[kani::proof]
    #[kani::unwind(8)]
    fn verify_geometric_mean_bounded() {
        let n: usize = kani::any();
        kani::assume(n > 0 && n <= 7);
        let mut values = Vec::with_capacity(n);
        for _ in 0..n {
            let v: f64 = kani::any();
            kani::assume(v >= 0.0 && v <= 100.0 && v.is_finite());
            values.push(v);
        }
        let result = geometric_mean(&values);
        assert!(result >= 0.0, "geometric_mean must be non-negative");
        assert!(result <= 100.0, "geometric_mean must not exceed max input");
        assert!(result.is_finite() || result == 0.0, "geometric_mean must be finite or zero");
    }

    /// Prove: geometric_mean of empty slice returns 0.
    #[kani::proof]
    fn verify_geometric_mean_empty() {
        let result = geometric_mean(&[]);
        assert_eq!(result, 0.0);
    }

    /// Prove: geometric_mean of single value returns that value.
    #[kani::proof]
    fn verify_geometric_mean_identity() {
        let v: f64 = kani::any();
        kani::assume(v > 0.0 && v <= 100.0 && v.is_finite());
        let result = geometric_mean(&[v]);
        // Allow small floating-point epsilon
        assert!((result - v).abs() < 1e-10, "single-value geometric mean must equal the value");
    }

    /// Prove: geometric_mean with any zero input returns 0.
    #[kani::proof]
    fn verify_geometric_mean_zero_absorbing() {
        let a: f64 = kani::any();
        kani::assume(a >= 0.0 && a <= 100.0 && a.is_finite());
        let result = geometric_mean(&[a, 0.0]);
        assert_eq!(result, 0.0, "any zero input must make geometric mean zero");
    }
}

fn get_head_sha(path: &Path) -> String {
    std::process::Command::new("git")
        .args(["rev-parse", "--short", "HEAD"])
        .current_dir(path)
        .output()
        .ok()
        .and_then(|o| String::from_utf8(o.stdout).ok())
        .map(|s| s.trim().to_string())
        .unwrap_or_else(|| "unknown".to_string())
}

fn persist_score(path: &Path, score: &CompositeScore) {
    let metrics_dir = path.join(".pmat-metrics");
    let _ = std::fs::create_dir_all(&metrics_dir);
    let filename = format!("commit-{}-meta.json", score.sha);
    let filepath = metrics_dir.join(filename);
    if let Ok(json) = serde_json::to_string_pretty(score) {
        let _ = std::fs::write(filepath, json);
    }
}

struct Violation {
    id: &'static str,
    message: String,
}

/// CB-146: Cross-validation invariants. Detect contradictions between sub-scores.
fn cross_validate(score: &CompositeScore) -> Vec<Violation> {
    let s = &score.sub_scores;
    let mut v = Vec::new();

    // XV-001: TDG grade gate pass should correlate with decent code quality
    // CB-200 passes when comply has no TDG failures, but RPS Code Quality can still be low
    if score.comply_errors == 0 {
        if let Some(cq) = score.rps_categories.get("Code Quality") {
            if *cq < 40.0 {
                v.push(Violation {
                    id: "XV-001",
                    message: format!("Comply 0 errors but RPS Code Quality {cq:.0}% < 40%"),
                });
            }
        }
    }

    // XV-003: High coverage should mean decent testing score
    if s.coverage >= 90.0 {
        if let Some(ts) = score.rps_categories.get("Testing Excellence") {
            if *ts < 60.0 {
                v.push(Violation {
                    id: "XV-003",
                    message: format!("Coverage {:.0}% but RPS Testing {ts:.0}% < 60%", s.coverage),
                });
            }
        }
    }

    // XV-007: RPS Grade A should mean composite >= 75
    if s.rps >= 90.0 && score.composite < 75.0 {
        v.push(Violation {
            id: "XV-007",
            message: format!(
                "RPS {:.0} (A-level) but composite {:.1} < 75",
                s.rps, score.composite
            ),
        });
    }

    // XV-008: Clean comply should correlate with decent RPS
    if score.comply_errors == 0 && s.rps < 60.0 {
        v.push(Violation {
            id: "XV-008",
            message: format!("Comply 0 errors but RPS {:.0} < 60", s.rps),
        });
    }

    // XV-009: Good file health should mean low muda over-processing
    if s.file_health >= 90.0 && s.muda_inv < 70.0 {
        v.push(Violation {
            id: "XV-009",
            message: format!(
                "File health {:.0} (A) but Muda inv {:.0} < 70",
                s.file_health, s.muda_inv
            ),
        });
    }

    // XV-010: Low coverage must cap composite
    if s.coverage < 50.0 && score.composite >= 80.0 {
        v.push(Violation {
            id: "XV-010",
            message: format!(
                "Coverage {:.0}% < 50 but composite {:.1} >= 80",
                s.coverage, score.composite
            ),
        });
    }

    v
}