directiva 0.1.0

A tiny, paste-friendly directive mini-language: ACTION:[<KIND>]NAME[@PATH][=NOTE]
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
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
//! Minimal glob matching: `*` (any run), `?` (single byte), `{a,b,c}` (one level of alternation),
//! `[a-z]` / `[!neg]` (shell-style char-classes), and `\x` (one universal backslash escape).
//!
//! Design choices that distinguish this from a general-purpose glob crate:
//!
//! 1. **Linear matching.** The matcher is the classic two-pointer wildcard algorithm with a single
//!    backtrack anchor — `O(n·m)` worst case, never the exponential blow-up of naive recursive
//!    globs (`a*a*a*a*b` against a long run of `a`s). Directives may carry untrusted patterns.
//! 2. **Not path-aware.** `*` matches `/` like any other byte. A NAME glob runs against bare
//!    identifiers; a PATH glob runs against whole paths where one `*` is expected to span `/`
//!    (`*/tests/*`). Path-segment semantics would be wrong here.
//! 3. **Byte-oriented.** `?` and char-classes match a single **byte**, not a Unicode scalar.
//!
//! Compilation order, both stages escape-aware: brace alternation is expanded first into a small
//! set of sub-patterns (`{,s}` allows an empty branch; an unmatched/`\{`-escaped brace is a
//! literal; only one level expands), then each sub-pattern is tokenized into [`Tok`]s.

/// A compiled glob pattern: brace alternation pre-expanded into one or more tokenized
/// sub-patterns. [`Pattern::matches`] succeeds if *any* sub-pattern matches the whole input.
#[derive(Clone, Debug)]
pub struct Pattern {
    /// Original source text, retained for `Display` / round-tripping.
    src: String,
    /// One tokenized program per brace-expanded alternative.
    alts: Vec<Vec<Tok>>,
}

/// One compiled glob token.
#[derive(Clone, Debug)]
enum Tok {
    /// `*` — match any run of bytes (including `/`).
    Star,
    /// `?` — match exactly one byte.
    AnyOne,
    /// A literal byte (possibly the un-escaped form of `\x`).
    Lit(u8),
    /// `[...]` char-class.
    Class {
        negated: bool,
        items: Vec<ClassItem>,
    },
}

#[derive(Clone, Debug)]
enum ClassItem {
    Byte(u8),
    Range(u8, u8),
}

/// A raw char-class element before range resolution: a literal byte or an unescaped `-`.
enum E {
    Ch(u8),
    Dash,
}

impl Pattern {
    /// Compile a glob pattern. Never fails: malformed brace/class syntax degrades to literals
    /// (shell-glob leniency), so any string is a valid pattern.
    #[must_use]
    pub fn compile(pat: &str) -> Self {
        let alts = expand_braces(pat).iter().map(|s| compile_alt(s)).collect();
        Self {
            src: pat.to_owned(),
            alts,
        }
    }

    /// Returns `true` if any expanded alternative matches `s` in full.
    #[must_use]
    pub fn matches(&self, s: &str) -> bool {
        let bytes = s.as_bytes();
        self.alts.iter().any(|toks| match_one(toks, bytes))
    }

    /// The original pattern text this was compiled from.
    #[must_use]
    pub fn as_str(&self) -> &str {
        &self.src
    }

    /// Number of brace-expanded alternatives (1 when the pattern has no alternation).
    #[must_use]
    pub fn alt_count(&self) -> usize {
        self.alts.len()
    }
}

impl std::fmt::Display for Pattern {
    fn fmt(&self, f: &mut std::fmt::Formatter<'_>) -> std::fmt::Result {
        f.write_str(&self.src)
    }
}

/// `Pattern`s compare by their source text (the compiled form is a pure function of it).
impl PartialEq for Pattern {
    fn eq(&self, other: &Self) -> bool {
        self.src == other.src
    }
}
impl Eq for Pattern {}

// ── matcher ──────────────────────────────────────────────────────────────────

/// Linear wildcard match of one tokenized sub-pattern against `s`.
///
/// Two-pointer with a single `*` backtrack anchor: when the bytes after a `*` later fail to match,
/// the `*` swallows one more byte and we retry, instead of re-exploring every split recursively.
fn match_one(toks: &[Tok], s: &[u8]) -> bool {
    let (mut ti, mut si) = (0usize, 0usize);
    // `star` = index in `toks` of the `*`; `star_s` = where in `s` it currently stops consuming.
    let mut star: Option<usize> = None;
    let mut star_s = 0usize;

    while si < s.len() {
        match toks.get(ti) {
            Some(Tok::Star) => {
                star = Some(ti);
                star_s = si;
                ti += 1;
            }
            Some(t) if tok_matches(t, s[si]) => {
                ti += 1;
                si += 1;
            }
            _ => {
                if let Some(sp) = star {
                    ti = sp + 1;
                    star_s += 1;
                    si = star_s;
                } else {
                    return false;
                }
            }
        }
    }
    // Trailing `*`s match the empty remainder.
    while matches!(toks.get(ti), Some(Tok::Star)) {
        ti += 1;
    }
    ti == toks.len()
}

fn tok_matches(t: &Tok, b: u8) -> bool {
    match t {
        Tok::Star => false, // handled by the caller; never a single-byte match
        Tok::AnyOne => true,
        Tok::Lit(c) => *c == b,
        Tok::Class { negated, items } => {
            let inside = items.iter().any(|it| match it {
                ClassItem::Byte(x) => *x == b,
                ClassItem::Range(lo, hi) => *lo <= b && b <= *hi,
            });
            inside ^ negated
        }
    }
}

// ── compilation ──────────────────────────────────────────────────────────────

/// Tokenize one brace-free sub-pattern. Escape-aware (`\x` → literal `x`); a `[` that opens no
/// valid class is demoted to a literal byte.
fn compile_alt(s: &str) -> Vec<Tok> {
    let b = s.as_bytes();
    let mut toks = Vec::new();
    let mut i = 0;
    while i < b.len() {
        match b[i] {
            b'\\' => {
                if i + 1 < b.len() {
                    toks.push(Tok::Lit(b[i + 1]));
                    i += 2;
                } else {
                    toks.push(Tok::Lit(b'\\')); // trailing backslash is a literal
                    i += 1;
                }
            }
            b'*' => {
                toks.push(Tok::Star);
                i += 1;
            }
            b'?' => {
                toks.push(Tok::AnyOne);
                i += 1;
            }
            b'[' => {
                if let Some((class, next)) = parse_class(b, i) {
                    toks.push(class);
                    i = next;
                } else {
                    toks.push(Tok::Lit(b'[')); // lenient: unmatched '['
                    i += 1;
                }
            }
            c => {
                toks.push(Tok::Lit(c));
                i += 1;
            }
        }
    }
    toks
}

/// Parse `[...]` starting at `b[start] == '['`. Returns `(Tok::Class, index_after_])` or `None` if
/// there's no closing `]` (caller treats the `[` as a literal). Shell-style:
/// leading `!` negates; `]` is literal when first; `-` is literal when first or last.
fn parse_class(b: &[u8], start: usize) -> Option<(Tok, usize)> {
    let mut i = start + 1;
    let mut negated = false;
    if b.get(i) == Some(&b'!') {
        negated = true;
        i += 1;
    }

    // Collect the class body as bytes-or-dashes, resolving escapes; `]` as the first element is a
    // literal, otherwise it closes the class.
    let mut elems: Vec<E> = Vec::new();
    let mut first = true;
    loop {
        let c = *b.get(i)?; // EOF before ']' ⇒ None
        match c {
            b'\\' => {
                if let Some(&n) = b.get(i + 1) {
                    elems.push(E::Ch(n));
                    i += 2;
                } else {
                    elems.push(E::Ch(b'\\'));
                    i += 1;
                }
            }
            b']' if first => {
                elems.push(E::Ch(b']'));
                i += 1;
            }
            b']' => {
                i += 1;
                break;
            }
            b'-' => {
                elems.push(E::Dash);
                i += 1;
            }
            other => {
                elems.push(E::Ch(other));
                i += 1;
            }
        }
        first = false;
    }

    // Resolve ranges: `Ch lo, Dash, Ch hi` ⇒ Range(lo,hi); a stray Dash (first/last) is literal.
    let mut items: Vec<ClassItem> = Vec::new();
    let mut k = 0;
    while k < elems.len() {
        match elems[k] {
            E::Ch(lo) => {
                if matches!(elems.get(k + 1), Some(E::Dash)) {
                    if let Some(E::Ch(hi)) = elems.get(k + 2) {
                        items.push(ClassItem::Range(lo, *hi));
                        k += 3;
                        continue;
                    }
                }
                items.push(ClassItem::Byte(lo));
                k += 1;
            }
            E::Dash => {
                items.push(ClassItem::Byte(b'-'));
                k += 1;
            }
        }
    }

    Some((Tok::Class { negated, items }, i))
}

/// Expand one level of `{a,b,c}` brace alternation, recursively re-expanding the tail. Escape-aware
/// — `\{`, `\}`, `\,` are skipped (left for the tokenizer to turn into literals). An unmatched
/// brace falls through as a single literal pattern.
fn expand_braces(pat: &str) -> Vec<String> {
    let b = pat.as_bytes();
    let Some(open) = find_unescaped(b, 0, b'{') else {
        return vec![pat.to_owned()];
    };
    let Some(close) = find_unescaped(b, open + 1, b'}') else {
        return vec![pat.to_owned()];
    };
    let prefix = &pat[..open];
    let alts = &pat[open + 1..close];
    let suffix = &pat[close + 1..];

    let mut out = Vec::new();
    for alt in split_unescaped(alts, b',') {
        for tail in expand_braces(suffix) {
            out.push(format!("{prefix}{alt}{tail}"));
        }
    }
    out
}

/// Index of the first unescaped `needle` byte at or after `from`, honoring `\`-escapes.
fn find_unescaped(b: &[u8], from: usize, needle: u8) -> Option<usize> {
    let mut i = from;
    while i < b.len() {
        match b[i] {
            b'\\' => i += 2, // skip the escaped byte
            c if c == needle => return Some(i),
            _ => i += 1,
        }
    }
    None
}

/// Split `s` on unescaped `sep` bytes (escaped separators stay in the piece).
fn split_unescaped(s: &str, sep: u8) -> Vec<&str> {
    let b = s.as_bytes();
    let mut out = Vec::new();
    let (mut start, mut i) = (0usize, 0usize);
    while i < b.len() {
        match b[i] {
            b'\\' => i += 2,
            c if c == sep => {
                out.push(&s[start..i]);
                i += 1;
                start = i;
            }
            _ => i += 1,
        }
    }
    out.push(&s[start..]);
    out
}

#[cfg(test)]
mod tests {
    use super::{expand_braces, Pattern};

    fn m(pat: &str, s: &str) -> bool {
        Pattern::compile(pat).matches(s)
    }

    #[test]
    fn literal_and_anchors() {
        assert!(m("abc", "abc"));
        assert!(!m("abc", "abd"));
        assert!(!m("abc", "abcd"));
        assert!(!m("abc", "ab"));
    }

    #[test]
    fn star_spans_anything_including_slash() {
        assert!(m("*", ""));
        assert!(m("*", "anything/at/all"));
        assert!(m("a*c", "ac"));
        assert!(m("a*c", "abbbc"));
        assert!(m("*/tests/*", "src/pkg/tests/foo.rs"));
        assert!(!m("*/tests/*", "src/pkg/test/foo.rs"));
    }

    // byte-oriented `?`
    #[test]
    fn question_matches_single_byte() {
        assert!(m("a?c", "abc"));
        assert!(!m("a?c", "ac"));
        assert!(!m("a?c", "abbc"));
        // a 2-byte UTF-8 scalar ('é' = 0xC3 0xA9) needs two `?`
        assert!(m("??", "é"));
        assert!(!m("?", "é"));
    }

    // no exponential blow-up
    #[test]
    fn collapsed_stars_do_not_blow_up() {
        let pat = "a*a*a*a*a*a*a*a*b";
        let s = "a".repeat(64);
        assert!(!m(pat, &s));
        assert!(m("***", "anything"));
    }

    // shell-style char-classes
    #[test]
    fn char_classes() {
        assert!(m("[a-z]bc", "xbc"));
        assert!(!m("[a-z]bc", "Xbc"));
        assert!(m("test_[0-9]*", "test_42x"));
        // negation
        assert!(m("[!0-9]", "a"));
        assert!(!m("[!0-9]", "7"));
        // ']' literal when first; '-' literal when first or last
        assert!(m("[]a]", "]"));
        assert!(m("[]a]", "a"));
        assert!(m("[a-]", "a"));
        assert!(m("[a-]", "-"));
        assert!(m("[-a]", "-"));
    }

    // escaping reaches the glob layer
    #[test]
    fn escaping_literals() {
        assert!(m(r"\*", "*"));
        assert!(!m(r"\*", "anything"));
        assert!(m(r"\[lit\]", "[lit]"));
        assert!(m(r"a\?b", "a?b"));
        assert!(!m(r"a\?b", "axb"));
        // escaped brace is NOT alternation
        assert_eq!(Pattern::compile(r"\{a,b\}").alt_count(), 1);
        assert!(m(r"\{a,b\}", "{a,b}"));
    }

    #[test]
    fn brace_expansion_covers_all_conventions() {
        let e = expand_braces("*/{test,tests,__tests__}/*");
        assert_eq!(e.len(), 3);
        assert!(e.contains(&"*/test/*".to_owned()));
        assert!(e.contains(&"*/tests/*".to_owned()));
        assert!(e.contains(&"*/__tests__/*".to_owned()));
    }

    #[test]
    fn brace_alternation_matches() {
        let p = Pattern::compile("*/{test,tests,__tests__}/*");
        assert!(p.matches("packages/compiler-cli/test/compliance/foo.ts"));
        assert!(p.matches("packages/svelte/tests/css/test.ts"));
        assert!(p.matches("packages/next/src/__tests__/foo.test.ts"));
        assert!(!p.matches("src/components/Button.ts"));
    }

    #[test]
    fn file_extension_alternation() {
        let p = Pattern::compile("*.{test,spec}.*");
        assert!(p.matches("src/foo.test.ts"));
        assert!(p.matches("src/foo.spec.ts"));
        assert!(!p.matches("src/foo.ts"));
    }

    #[test]
    fn empty_alternation_branch_is_legal() {
        let e = expand_braces("*/{,s}post*");
        assert!(e.contains(&"*/post*".to_owned()));
        assert!(e.contains(&"*/spost*".to_owned()));
    }

    #[test]
    fn two_groups_compose() {
        let e = expand_braces("{a,b}{c,d}");
        assert_eq!(e.len(), 4);
        for want in ["ac", "ad", "bc", "bd"] {
            assert!(e.contains(&want.to_owned()), "missing {want}");
        }
    }

    #[test]
    fn no_braces_is_one_alternative() {
        assert_eq!(Pattern::compile("*tests/*").alt_count(), 1);
        assert_eq!(expand_braces("*tests/*"), vec!["*tests/*".to_owned()]);
    }

    #[test]
    fn unmatched_brace_is_literal() {
        assert!(m("a{b", "a{b"));
        assert_eq!(Pattern::compile("a{b").alt_count(), 1);
    }
}