unslop 0.1.5

Deterministic linter for essays, posts, email, reports, and docs that finds AI writing patterns
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

//! density family: integer running sums, population variance via
//! `n*sum_sq - sum^2`, integer cross-multiplication against rational
//! thresholds. Density rules escalate to whole-document review and never
//! substitute for individual hits.

use crate::engine::{CompiledPolicy, Hit};
use crate::extract::Doc;
use crate::input::Prepared;
use crate::Config;

pub const HANDLED: &[&str] = &[
    "SLOP-D001",
    "SLOP-D002",
    "SLOP-D003",
    "SLOP-D004",
    "SLOP-C009",
    "SLOP-L002",
];

fn count_rule_hits(cp: &CompiledPolicy, hits: &[Hit], ids: &[String]) -> u64 {
    hits.iter()
        .filter(|h| ids.iter().any(|id| &cp.pkg.rules[h.rule].id == id))
        .count() as u64
}

fn param_rule_ids(rule: &crate::policy::Rule) -> Vec<String> {
    rule.params
        .as_table()
        .and_then(|t| t.get("count_rules"))
        .and_then(|v| v.as_array())
        .map(|a| {
            a.iter()
                .filter_map(|v| v.as_str().map(|s| s.to_string()))
                .collect()
        })
        .unwrap_or_default()
}

pub fn evaluate(
    cp: &CompiledPolicy,
    prepared: &Prepared,
    doc: &Doc,
    config: &Config,
    hits: &mut Vec<Hit>,
) {
    let whole = crate::first_char_span(&prepared.text);

    for id in ["SLOP-D001", "SLOP-D004"] {
        let Some(idx) = super::active(cp, config, id) else {
            continue;
        };
        let rule = &cp.pkg.rules[idx];
        let ids = param_rule_ids(rule);
        let threshold = super::param_i64(rule, "threshold").unwrap_or(3) as u64;
        let count = count_rule_hits(cp, hits, &ids);
        let per_document = rule
            .params
            .as_table()
            .and_then(|t| t.get("per_document"))
            .and_then(|v| v.as_bool())
            .unwrap_or(false);
        let fires = if per_document {
            count >= threshold
        } else {
            let per_words = super::param_i64(rule, "per_words").unwrap_or(500) as u64;
            let words = doc.stats.word_count;
            // Rate check by cross-multiplication, plus an absolute floor so
            // one hit in a short note does not escalate.
            words > 0 && count >= threshold && count * per_words >= threshold * words
        };
        if fires {
            hits.push(Hit::new(idx, whole.clone()));
        }
    }

    if let Some(idx) = super::active(cp, config, "SLOP-D002") {
        let rule = &cp.pkg.rules[idx];
        let min_bullets = super::param_i64(rule, "min_bullets").unwrap_or(8) as u64;
        let min_pct = super::param_i64(rule, "min_link_bullet_pct").unwrap_or(60) as u64;
        if doc.stats.bullets >= min_bullets
            && doc.stats.bullets_with_link * 100 >= min_pct * doc.stats.bullets
        {
            hits.push(Hit::new(idx, whole.clone()));
        }
    }

    // Rate instruments, not gates: SLOP-C009 over the contrast family and
    // SLOP-L002 over the agentive passive. No threshold param by design, the
    // absence being the deferred per-profile probe. Zero hits emit nothing,
    // since an all-zero line is noise.
    for (id, unit) in [
        ("SLOP-C009", "contrast hits"),
        ("SLOP-L002", "passive hits"),
    ] {
        let Some(idx) = super::active(cp, config, id) else {
            continue;
        };
        let rule = &cp.pkg.rules[idx];
        let ids = param_rule_ids(rule);
        let per_words = super::param_i64(rule, "per_words").unwrap_or(1000) as u64;
        let count = count_rule_hits(cp, hits, &ids);
        let words = doc.stats.word_count;
        if count > 0 && words > 0 && per_words > 0 {
            // One decimal place via integer math, no floats.
            let tenths = count * per_words * 10 / words;
            let mut hit = Hit::new(idx, whole.clone());
            hit.detail = Some(format!(
                "{}.{} per {} words ({} {} / {} words)",
                tenths / 10,
                tenths % 10,
                per_words,
                count,
                unit,
                words
            ));
            hits.push(hit);
        }
    }

    if let Some(idx) = super::active(cp, config, "SLOP-D003") {
        let rule = &cp.pkg.rules[idx];
        let min_paragraphs = super::param_i64(rule, "min_paragraphs").unwrap_or(5) as usize;
        let cv_pct = super::param_i64(rule, "max_length_cv_pct").unwrap_or(20) as u128;
        let words: Vec<u64> = doc
            .stats
            .paragraph_words
            .iter()
            .copied()
            .filter(|&w| w > 0)
            .collect();
        if words.len() >= min_paragraphs {
            let n = words.len() as u128;
            let s: u128 = words.iter().map(|&w| w as u128).sum();
            let sq: u128 = words.iter().map(|&w| (w as u128) * (w as u128)).sum();
            // cv <= t/100  <=>  100^2 * (n*sq - s^2) <= t^2 * s^2
            if s > 0 && 10_000 * (n * sq - s * s) <= cv_pct * cv_pct * s * s {
                hits.push(Hit::new(idx, whole.clone()));
            }
        }
    }
}