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

impl DataScienceAnalyzer {
    /// Cluster findings using K-means
    ///
    /// Features used for clustering:
    /// - Severity (ordinal encoded)
    /// - Category (one-hot encoded conceptually, but simplified here)
    /// - File path similarity (simplified to directory grouping)
    /// - Line proximity (normalized)
    #[provable_contracts_macros::contract("pmat-core.yaml", equation = "check_compliance")]
    pub fn cluster_findings(&self, findings: &mut [Finding]) -> Vec<FindingCluster> {
        if findings.is_empty() || findings.len() < self.k_clusters {
            // Not enough findings to cluster meaningfully
            if !findings.is_empty() {
                for finding in findings.iter_mut() {
                    finding.cluster_id = Some(0);
                }
                return vec![FindingCluster {
                    id: 0,
                    size: findings.len(),
                    primary_category: findings
                        .first()
                        .map(|f| f.category.clone())
                        .unwrap_or_default(),
                    cohesion: 1.0,
                    description: "All findings".to_string(),
                    finding_ids: findings.iter().map(|f| f.id.clone()).collect(),
                }];
            }
            return Vec::new();
        }

        // Build feature vectors for each finding
        let vectors: Vec<Vec<f32>> = findings
            .iter()
            .map(|f| self.finding_to_features(f))
            .collect();

        // Convert to aprender Matrix
        let rows = vectors.len();
        let cols = vectors[0].len();
        let data: Vec<f32> = vectors.iter().flat_map(|v| v.iter().copied()).collect();

        let matrix = match Matrix::from_vec(rows, cols, data) {
            Ok(m) => m,
            Err(_) => return Vec::new(),
        };

        // Run K-means clustering
        let mut kmeans = KMeans::new(self.k_clusters).with_max_iter(100);

        if kmeans.fit(&matrix).is_err() {
            return Vec::new();
        }

        let labels = kmeans.predict(&matrix);

        // Assign cluster IDs to findings
        for (finding, &label) in findings.iter_mut().zip(labels.iter()) {
            finding.cluster_id = Some(label);
        }

        // Build cluster summaries
        self.build_cluster_summaries(findings, &labels)
    }

    /// Convert a finding to a feature vector
    #[allow(clippy::cast_possible_truncation)]
    fn finding_to_features(&self, finding: &Finding) -> Vec<f32> {
        let mut features = vec![0.0f32; 6];

        // Severity (0-3)
        features[0] = match finding.severity {
            super::types::Severity::Low => 0.0,
            super::types::Severity::Medium => 1.0,
            super::types::Severity::High => 2.0,
            super::types::Severity::Critical => 3.0,
        };

        // Confidence
        features[1] = finding.confidence;

        // Category hash (simplified)
        features[2] = (finding.category.len() % 10) as f32;

        // File path hash (simplified)
        features[3] = (finding.location.file.to_string_lossy().len() % 20) as f32;

        // Line number (normalized to 0-1 range, assuming max 10000 lines)
        features[4] = (finding.location.line as f32 / 10000.0).min(1.0);

        // Has fix suggestion
        features[5] = if finding.fix_suggestion.is_some() {
            1.0
        } else {
            0.0
        };

        features
    }

    /// Build cluster summaries from labels
    fn build_cluster_summaries(
        &self,
        findings: &[Finding],
        labels: &[usize],
    ) -> Vec<FindingCluster> {
        let mut cluster_findings: HashMap<usize, Vec<&Finding>> = HashMap::new();

        for (finding, &label) in findings.iter().zip(labels.iter()) {
            cluster_findings.entry(label).or_default().push(finding);
        }

        cluster_findings
            .into_iter()
            .map(|(id, cluster_items)| {
                // Find dominant category
                let mut category_counts: HashMap<&str, usize> = HashMap::new();
                for finding in &cluster_items {
                    *category_counts.entry(&finding.category).or_insert(0) += 1;
                }
                let primary_category = category_counts
                    .into_iter()
                    .max_by_key(|(_, count)| *count)
                    .map(|(cat, _)| cat.to_string())
                    .unwrap_or_default();

                // Calculate cohesion (simplified: inverse of category diversity)
                let unique_categories: std::collections::HashSet<_> =
                    cluster_items.iter().map(|f| &f.category).collect();
                let cohesion = 1.0 / (unique_categories.len() as f64).max(1.0);

                FindingCluster {
                    id,
                    size: cluster_items.len(),
                    primary_category: primary_category.clone(),
                    cohesion,
                    description: format!("{} Issues", primary_category),
                    finding_ids: cluster_items.iter().map(|f| f.id.clone()).collect(),
                }
            })
            .collect()
    }
}