holys3 0.5.1

Indexed regex search for local files and private S3 buckets
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

//! Pattern composition: rg-style -e/-F/-w/-i/-S transforms applied BEFORE
//! the gram planner and the verifier, so both see the identical final regex.

/// THE pattern entry point: escapes (`-F`), resolves case (`-i`/`-S`),
/// composes (`-w`, multi-pattern OR, `(?m)`), and sanitizes line terminators
/// — in that order, so the mandatory sanitize step cannot be skipped at a
/// call site. Smart case analyzes the escaped forms, so `-F -S 'foo('`
/// stays insensitive like rg.
pub(crate) fn build_pattern(
    patterns: &[String],
    fixed_strings: bool,
    word_regexp: bool,
    ignore_case: bool,
    smart_case: bool,
) -> anyhow::Result<String> {
    let patterns: Vec<String> = if fixed_strings {
        patterns.iter().map(|p| regex::escape(p)).collect()
    } else {
        patterns.to_vec()
    };
    let insensitive = ignore_case || (smart_case && smart_case_insensitive(&patterns));
    sanitize_line_terminators(&compose_pattern(&patterns, word_regexp, insensitive))
}

/// `-w` wraps in word boundaries, multiple patterns OR, `-i` prepends
/// `(?i)`. The result may be an invalid regex — the sanitize parse reports
/// that with the real error.
fn compose_pattern(patterns: &[String], word_regexp: bool, insensitive: bool) -> String {
    let transformed: Vec<String> = patterns
        .iter()
        .map(|p| {
            if word_regexp {
                // rg uses HALF word boundaries: full \b anchors incorrectly when the
                // pattern's first/last char is a non-word char (`foo(`, `->`)
                format!(r"\b{{start-half}}(?:{p})\b{{end-half}}")
            } else {
                p.clone()
            }
        })
        .collect();
    let joined = if transformed.len() == 1 {
        transformed.into_iter().next().expect("one pattern")
    } else {
        transformed
            .iter()
            .map(|p| format!("(?:{p})"))
            .collect::<Vec<_>>()
            .join("|")
    };
    // (?m): rg matches line-by-line, so ^ and $ anchor at EVERY line
    // boundary, not just the start and end of the object
    if insensitive {
        format!("(?mi){joined}")
    } else {
        format!("(?m){joined}")
    }
}

/// rg's line-terminator discipline, applied to the COMPOSED pattern: strip
/// \n out of every character class that matches it (this is what makes
/// `[^x]` and `(?s).` line-oriented — a class spanning \n would swallow
/// following lines into one phantom match), then reject any literal \n
/// left (a match can never span the line-oriented unit of output).
fn sanitize_line_terminators(pattern: &str) -> anyhow::Result<String> {
    use regex_syntax::hir::{Class, Hir, HirKind, Literal};
    fn strip(hir: &Hir) -> anyhow::Result<Hir> {
        Ok(match hir.kind() {
            HirKind::Literal(Literal(bytes)) => {
                anyhow::ensure!(
                    !bytes.contains(&b'\n'),
                    "the literal '\\n' is not allowed in a regex"
                );
                hir.clone()
            }
            HirKind::Class(Class::Bytes(class)) => {
                let mut class = class.clone();
                let mut newline =
                    regex_syntax::hir::ClassBytes::new([regex_syntax::hir::ClassBytesRange::new(
                        b'\n', b'\n',
                    )]);
                newline.negate();
                class.intersect(&newline);
                Hir::class(Class::Bytes(class))
            }
            HirKind::Class(Class::Unicode(class)) => {
                let mut class = class.clone();
                let mut newline = regex_syntax::hir::ClassUnicode::new([
                    regex_syntax::hir::ClassUnicodeRange::new('\n', '\n'),
                ]);
                newline.negate();
                class.intersect(&newline);
                Hir::class(Class::Unicode(class))
            }
            HirKind::Repetition(rep) => {
                let mut rep = rep.clone();
                rep.sub = Box::new(strip(&rep.sub)?);
                Hir::repetition(rep)
            }
            HirKind::Capture(cap) => {
                let mut cap = cap.clone();
                cap.sub = Box::new(strip(&cap.sub)?);
                Hir::capture(cap)
            }
            HirKind::Concat(subs) => {
                Hir::concat(subs.iter().map(strip).collect::<anyhow::Result<_>>()?)
            }
            HirKind::Alternation(subs) => {
                Hir::alternation(subs.iter().map(strip).collect::<anyhow::Result<_>>()?)
            }
            HirKind::Empty | HirKind::Look(_) => hir.clone(),
        })
    }
    let hir = regex_syntax::ParserBuilder::new()
        .utf8(false)
        .build()
        .parse(pattern)?;
    Ok(strip(&hir)?.to_string())
}

/// rg smart case: insensitive iff the patterns contain at least one literal
/// character and none of the literal characters are uppercase. Classes like
/// `\pL` are not literals. An unparsable raw pattern is "not lowercase" —
/// the sanitize parse surfaces the real error.
fn smart_case_insensitive(patterns: &[String]) -> bool {
    let mut literals = Vec::new();
    for pattern in patterns {
        let Ok(hir) = regex_syntax::parse(pattern) else {
            return false; // compile will surface the real error
        };
        collect_literal_chars(&hir, &mut literals);
    }
    !literals.is_empty() && !literals.iter().any(|c| c.is_uppercase())
}

fn collect_literal_chars(hir: &regex_syntax::hir::Hir, out: &mut Vec<char>) {
    use regex_syntax::hir::HirKind;
    match hir.kind() {
        HirKind::Literal(lit) => out.extend(String::from_utf8_lossy(&lit.0).chars()),
        HirKind::Concat(children) | HirKind::Alternation(children) => {
            for child in children {
                collect_literal_chars(child, out);
            }
        }
        HirKind::Capture(capture) => collect_literal_chars(&capture.sub, out),
        HirKind::Repetition(rep) => collect_literal_chars(&rep.sub, out),
        HirKind::Empty | HirKind::Class(_) | HirKind::Look(_) => {}
    }
}

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

    fn one(p: &str) -> Vec<String> {
        vec![p.to_owned()]
    }

    #[test]
    fn compose_matrix() {
        assert_eq!(compose_pattern(&one(r"a\.b"), false, false), r"(?m)a\.b");
        assert_eq!(
            compose_pattern(&one("foo"), true, false),
            r"(?m)\b{start-half}(?:foo)\b{end-half}"
        );
        assert_eq!(
            compose_pattern(&["a".into(), "b".into()], false, false),
            "(?m)(?:a)|(?:b)"
        );
        assert_eq!(compose_pattern(&one("foo"), false, true), "(?mi)foo");
        assert_eq!(
            compose_pattern(&[r"a\.".into(), "b".into()], true, true),
            r"(?mi)(?:\b{start-half}(?:a\.)\b{end-half})|(?:\b{start-half}(?:b)\b{end-half})"
        );
    }

    #[test]
    fn build_pattern_escapes_fixed_strings() {
        let re = |p: &str, fixed: bool| {
            regex::bytes::Regex::new(&build_pattern(&one(p), fixed, false, false, false).unwrap())
                .unwrap()
        };
        assert!(re("a.b", true).is_match(b"a.b"));
        assert!(!re("a.b", true).is_match(b"axb"));
        assert!(re("a.b", false).is_match(b"axb"));
    }

    #[test]
    fn sanitize_strips_newline_from_classes_and_rejects_literals() {
        let re =
            |p: &str| regex::bytes::Regex::new(&sanitize_line_terminators(p).unwrap()).unwrap();
        // [^x] must not swallow the newline
        assert!(!re("[^x]").is_match(b"\n"));
        // (?s). is a class containing \n: stripped back to line-oriented
        assert!(!re("(?s).").is_match(b"\n"));
        assert!(re("(?s)a.b").is_match(b"axb"));
        assert!(!re("(?s)a.b").is_match(b"a\nb"));
        // \D and \W contain \n
        assert!(!re(r"\D").is_match(b"\n"));
        // literal newlines are rejected, in every spelling
        assert!(sanitize_line_terminators("a\nb").is_err());
        assert!(sanitize_line_terminators(r"a\nb").is_err());
        assert!(sanitize_line_terminators(r"a\x0Ab").is_err());
        // plain patterns round-trip
        assert!(re("(?m)^needle$").is_match(b"x\nneedle\ny"));
    }

    #[test]
    fn smart_case_rules() {
        assert!(smart_case_insensitive(&one("foo")));
        assert!(!smart_case_insensitive(&one("Foo")));
        assert!(smart_case_insensitive(&one(r"foo\pL")));
        assert!(!smart_case_insensitive(&one(r"Foo\pL")));
        assert!(!smart_case_insensitive(&one(r"\d+"))); // no literals
        assert!(!smart_case_insensitive(&one("(")));
        // an unparsable raw pattern errors out of build_pattern
        assert!(build_pattern(&one("("), false, false, false, true).is_err());
    }

    #[test]
    fn fixed_strings_smart_case_analyzes_escaped_pattern() {
        // rg parity: `rg -F -S 'foo('` matches `FOO(`
        let re = |p: &str| {
            regex::bytes::Regex::new(&build_pattern(&one(p), true, false, false, true).unwrap())
                .unwrap()
        };
        assert!(re("foo(").is_match(b"FOO("));
        assert!(re("Foo(").is_match(b"Foo("));
        assert!(!re("Foo(").is_match(b"FOO("));
    }
}