syntext 1.1.1

Hybrid code search index for agent workflows
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

//! Property-based tests for the forced boundary coverage invariant.
//!
//! The core property: for any document D and any token-aligned substring Q
//! (bounded by forced boundary characters), every gram in build_covering(Q)
//! must appear in build_all(D). This guarantees zero false negatives for
//! the gram index when queries are token-aligned.

use std::collections::HashSet;

use proptest::prelude::*;
use syntext::tokenizer::{build_all, build_covering, MIN_GRAM_LEN};

// ---------------------------------------------------------------------------
// Curated invariant test (deterministic, exhaustive on small inputs)
// ---------------------------------------------------------------------------

/// For curated documents, verify the covering invariant holds for every
/// token-aligned substring. Token-aligned means the substring starts and
/// ends at positions where forced boundary characters appear.
#[test]
fn covering_subset_of_all_in_context() {
    let documents: &[&[u8]] = &[
        b"fn parse_query(args: &str) -> Query {",
        b"def process_batch(items, config):",
        b"import { HashMap } from 'collections';",
        b"let result = self.name.to_string();",
        b"TODO: fix this before release",
        b"user@example.com sent 192.168.1.1",
        b"camelCaseIdentifier = getValue()",
        b"__init__(self, parse_query_engine)",
        b"PARSE_QUERY_MAX_LEN = 4096",
        b"  for item in items:\n    process_batch(item)",
    ];

    for doc in documents {
        let all: HashSet<u64> = build_all(doc).into_iter().collect();

        // Extract forced-boundary tokens from the document and verify each
        // one's covering grams are in build_all(doc).
        let lower: Vec<u8> = doc.iter().map(|b| b.to_ascii_lowercase()).collect();
        let boundaries = forced_boundary_positions(&lower);

        for i in 0..boundaries.len() {
            for j in (i + 1)..boundaries.len() {
                let start = boundaries[i];
                let end = boundaries[j];
                if end - start < MIN_GRAM_LEN {
                    continue;
                }

                let substr = &lower[start..end];
                if let Some(covering) = build_covering(substr) {
                    for h in &covering {
                        assert!(
                            all.contains(h),
                            "VIOLATION: query={:?} in doc={:?}, gram {:016x} not found\n\
                             boundaries: {:?}\nquery span: [{}..{}]",
                            String::from_utf8_lossy(substr),
                            String::from_utf8_lossy(doc),
                            h,
                            &boundaries[i..=j],
                            start,
                            end
                        );
                    }
                }
            }
        }
    }
}

// ---------------------------------------------------------------------------
// Property-based: token-aligned invariant (must have ZERO failures)
// ---------------------------------------------------------------------------

/// Generate strings that look like code: identifiers separated by
/// punctuation and whitespace.
fn code_like_string() -> impl Strategy<Value = Vec<u8>> {
    let token = "[a-z][a-z0-9]{2,12}";
    let separator = prop_oneof![
        Just(" ".to_string()),
        Just("(".to_string()),
        Just(")".to_string()),
        Just(".".to_string()),
        Just("_".to_string()),
        Just(", ".to_string()),
        Just(": ".to_string()),
        Just("\n".to_string()),
        Just(" = ".to_string()),
        Just(";".to_string()),
    ];

    prop::collection::vec((token, separator), 3..15).prop_map(|pairs| {
        let mut s = String::new();
        for (tok, sep) in pairs {
            s.push_str(&tok);
            s.push_str(&sep);
        }
        s.into_bytes()
    })
}

proptest! {
    #![proptest_config(ProptestConfig::with_cases(5_000))]

    /// For any generated code-like document, the token-aligned covering
    /// invariant must hold: every gram in build_covering(token) must
    /// appear in build_all(document).
    #[test]
    fn token_aligned_invariant(doc in code_like_string()) {
        let lower: Vec<u8> = doc.iter().map(|b| b.to_ascii_lowercase()).collect();
        let all: HashSet<u64> = build_all(&doc).into_iter().collect();
        let boundaries = forced_boundary_positions(&lower);

        for i in 0..boundaries.len() {
            for j in (i + 1)..boundaries.len() {
                let start = boundaries[i];
                let end = boundaries[j];
                if end - start < MIN_GRAM_LEN {
                    continue;
                }

                let substr = &lower[start..end];
                if let Some(covering) = build_covering(substr) {
                    for h in &covering {
                        prop_assert!(
                            all.contains(h),
                            "VIOLATION: query={:?} in doc context, gram not found",
                            String::from_utf8_lossy(substr)
                        );
                    }
                }
            }
        }
    }
}

// ---------------------------------------------------------------------------
// Property-based: arbitrary substring coverage (informational)
// ---------------------------------------------------------------------------

/// Quantifies how often non-token-aligned substrings cause coverage
/// violations. This test does NOT assert; it logs violation rates.
/// If the violation rate exceeds 5%, overlapping trigrams (Phase B)
/// are needed to cover mid-token substring queries.
///
/// Run with: cargo test --test boundary_fuzz arbitrary_substring -- --nocapture
#[test]
fn arbitrary_substring_coverage() {
    use proptest::test_runner::{Config, TestRunner};

    let config = Config::with_cases(5_000);
    let mut runner = TestRunner::new(config);

    use std::cell::Cell;
    let total_queries = Cell::new(0u64);
    let total_violations = Cell::new(0u64);
    let total_grams_checked = Cell::new(0u64);

    let result = runner.run(
        &(code_like_string(), 0.0f64..1.0, 0.01f64..0.3),
        |(doc, start_frac, len_frac)| {
            if doc.len() < MIN_GRAM_LEN {
                return Ok(());
            }
            let start = (start_frac * doc.len() as f64) as usize;
            let len = ((len_frac * doc.len() as f64) as usize).max(MIN_GRAM_LEN);
            let end = (start + len).min(doc.len());
            let start = start.min(end.saturating_sub(MIN_GRAM_LEN));
            if end - start < MIN_GRAM_LEN {
                return Ok(());
            }

            let lower: Vec<u8> = doc.iter().map(|b| b.to_ascii_lowercase()).collect();
            let all: std::collections::HashSet<u64> = build_all(&doc).into_iter().collect();
            let substr = &lower[start..end];

            if let Some(covering) = build_covering(substr) {
                total_queries.set(total_queries.get() + 1);
                total_grams_checked.set(total_grams_checked.get() + covering.len() as u64);
                let violations = covering.iter().filter(|h| !all.contains(h)).count();
                if violations > 0 {
                    total_violations.set(total_violations.get() + 1);
                }
            }

            Ok(())
        },
    );

    assert!(result.is_ok(), "proptest runner failed: {:?}", result);

    let tq = total_queries.get();
    let tv = total_violations.get();
    let tg = total_grams_checked.get();
    let violation_rate = if tq > 0 {
        tv as f64 / tq as f64 * 100.0
    } else {
        0.0
    };

    eprintln!(
        "\n--- Non-aligned substring coverage report ---\n\
         Total queries with grams: {}\n\
         Queries with violations:  {} ({:.2}%)\n\
         Total grams checked:      {}\n\
         Verdict: {}\n",
        tq,
        tv,
        violation_rate,
        tg,
        if violation_rate < 5.0 {
            "OK -- overlapping trigrams not needed"
        } else {
            "HIGH -- overlapping trigrams recommended"
        }
    );
}

// ---------------------------------------------------------------------------
// Helper: extract forced boundary positions (mirrors tokenizer logic)
// ---------------------------------------------------------------------------

fn is_forced_boundary(byte: u8) -> bool {
    matches!(
        byte,
        b' ' | b'\t' | b'\n' | b'\r' | 0x0B | 0x0C
            | b'(' | b')' | b'{' | b'}' | b'[' | b']' | b'<' | b'>'
            | b'.' | b',' | b':' | b';'
            | b'=' | b'+' | b'-' | b'*' | b'/' | b'%'
            | b'!' | b'&' | b'|' | b'^' | b'~'
            | b'"' | b'\'' | b'`'
            | b'@' | b'#' | b'$' | b'?'
            | b'_'
            | 0x00..=0x08 | 0x0E..=0x1F | 0x7F
    )
}

fn forced_boundary_positions(bytes: &[u8]) -> Vec<usize> {
    let mut positions = vec![0];
    for i in 1..bytes.len() {
        if is_forced_boundary(bytes[i]) || is_forced_boundary(bytes[i - 1]) {
            positions.push(i);
        }
    }
    if !bytes.is_empty() {
        positions.push(bytes.len());
    }
    positions.dedup();
    positions
}