]>
Commit | Line | Data |
---|---|---|
cdc7bbd5 XL |
1 | //! Utils for extracting, inspecting or transforming source code |
2 | ||
3 | #![allow(clippy::module_name_repetitions)] | |
4 | ||
5 | use crate::line_span; | |
6 | use rustc_errors::Applicability; | |
7 | use rustc_hir::{Expr, ExprKind}; | |
8 | use rustc_lint::{LateContext, LintContext}; | |
9 | use rustc_span::hygiene; | |
04454e1e | 10 | use rustc_span::source_map::SourceMap; |
923072b8 | 11 | use rustc_span::{BytePos, Pos, Span, SpanData, SyntaxContext}; |
cdc7bbd5 XL |
12 | use std::borrow::Cow; |
13 | ||
04454e1e FG |
14 | /// Checks if the span starts with the given text. This will return false if the span crosses |
15 | /// multiple files or if source is not available. | |
16 | /// | |
17 | /// This is used to check for proc macros giving unhelpful spans to things. | |
18 | pub fn span_starts_with<T: LintContext>(cx: &T, span: Span, text: &str) -> bool { | |
19 | fn helper(sm: &SourceMap, span: Span, text: &str) -> bool { | |
20 | let pos = sm.lookup_byte_offset(span.lo()); | |
21 | let Some(ref src) = pos.sf.src else { | |
22 | return false; | |
23 | }; | |
24 | let end = span.hi() - pos.sf.start_pos; | |
25 | src.get(pos.pos.0 as usize..end.0 as usize) | |
26 | // Expression spans can include wrapping parenthesis. Remove them first. | |
27 | .map_or(false, |s| s.trim_start_matches('(').starts_with(text)) | |
28 | } | |
29 | helper(cx.sess().source_map(), span, text) | |
30 | } | |
31 | ||
cdc7bbd5 XL |
32 | /// Like `snippet_block`, but add braces if the expr is not an `ExprKind::Block`. |
33 | /// Also takes an `Option<String>` which can be put inside the braces. | |
34 | pub fn expr_block<'a, T: LintContext>( | |
35 | cx: &T, | |
36 | expr: &Expr<'_>, | |
37 | option: Option<String>, | |
38 | default: &'a str, | |
39 | indent_relative_to: Option<Span>, | |
40 | ) -> Cow<'a, str> { | |
41 | let code = snippet_block(cx, expr.span, default, indent_relative_to); | |
42 | let string = option.unwrap_or_default(); | |
43 | if expr.span.from_expansion() { | |
44 | Cow::Owned(format!("{{ {} }}", snippet_with_macro_callsite(cx, expr.span, default))) | |
45 | } else if let ExprKind::Block(_, _) = expr.kind { | |
46 | Cow::Owned(format!("{}{}", code, string)) | |
47 | } else if string.is_empty() { | |
48 | Cow::Owned(format!("{{ {} }}", code)) | |
49 | } else { | |
50 | Cow::Owned(format!("{{\n{};\n{}\n}}", code, string)) | |
51 | } | |
52 | } | |
53 | ||
54 | /// Returns a new Span that extends the original Span to the first non-whitespace char of the first | |
55 | /// line. | |
56 | /// | |
57 | /// ```rust,ignore | |
58 | /// let x = (); | |
59 | /// // ^^ | |
60 | /// // will be converted to | |
61 | /// let x = (); | |
62 | /// // ^^^^^^^^^^ | |
63 | /// ``` | |
64 | pub fn first_line_of_span<T: LintContext>(cx: &T, span: Span) -> Span { | |
65 | first_char_in_first_line(cx, span).map_or(span, |first_char_pos| span.with_lo(first_char_pos)) | |
66 | } | |
67 | ||
68 | fn first_char_in_first_line<T: LintContext>(cx: &T, span: Span) -> Option<BytePos> { | |
69 | let line_span = line_span(cx, span); | |
70 | snippet_opt(cx, line_span).and_then(|snip| { | |
71 | snip.find(|c: char| !c.is_whitespace()) | |
72 | .map(|pos| line_span.lo() + BytePos::from_usize(pos)) | |
73 | }) | |
74 | } | |
75 | ||
76 | /// Returns the indentation of the line of a span | |
77 | /// | |
78 | /// ```rust,ignore | |
79 | /// let x = (); | |
80 | /// // ^^ -- will return 0 | |
81 | /// let x = (); | |
82 | /// // ^^ -- will return 4 | |
83 | /// ``` | |
84 | pub fn indent_of<T: LintContext>(cx: &T, span: Span) -> Option<usize> { | |
85 | snippet_opt(cx, line_span(cx, span)).and_then(|snip| snip.find(|c: char| !c.is_whitespace())) | |
86 | } | |
87 | ||
88 | /// Gets a snippet of the indentation of the line of a span | |
89 | pub fn snippet_indent<T: LintContext>(cx: &T, span: Span) -> Option<String> { | |
90 | snippet_opt(cx, line_span(cx, span)).map(|mut s| { | |
91 | let len = s.len() - s.trim_start().len(); | |
92 | s.truncate(len); | |
93 | s | |
94 | }) | |
95 | } | |
96 | ||
97 | // If the snippet is empty, it's an attribute that was inserted during macro | |
98 | // expansion and we want to ignore those, because they could come from external | |
99 | // sources that the user has no control over. | |
100 | // For some reason these attributes don't have any expansion info on them, so | |
101 | // we have to check it this way until there is a better way. | |
102 | pub fn is_present_in_source<T: LintContext>(cx: &T, span: Span) -> bool { | |
103 | if let Some(snippet) = snippet_opt(cx, span) { | |
104 | if snippet.is_empty() { | |
105 | return false; | |
106 | } | |
107 | } | |
108 | true | |
109 | } | |
110 | ||
04454e1e | 111 | /// Returns the position just before rarrow |
cdc7bbd5 XL |
112 | /// |
113 | /// ```rust,ignore | |
114 | /// fn into(self) -> () {} | |
115 | /// ^ | |
116 | /// // in case of unformatted code | |
117 | /// fn into2(self)-> () {} | |
118 | /// ^ | |
119 | /// fn into3(self) -> () {} | |
120 | /// ^ | |
121 | /// ``` | |
122 | pub fn position_before_rarrow(s: &str) -> Option<usize> { | |
123 | s.rfind("->").map(|rpos| { | |
124 | let mut rpos = rpos; | |
125 | let chars: Vec<char> = s.chars().collect(); | |
126 | while rpos > 1 { | |
127 | if let Some(c) = chars.get(rpos - 1) { | |
128 | if c.is_whitespace() { | |
129 | rpos -= 1; | |
130 | continue; | |
131 | } | |
132 | } | |
133 | break; | |
134 | } | |
135 | rpos | |
136 | }) | |
137 | } | |
138 | ||
139 | /// Reindent a multiline string with possibility of ignoring the first line. | |
923072b8 | 140 | #[expect(clippy::needless_pass_by_value)] |
cdc7bbd5 XL |
141 | pub fn reindent_multiline(s: Cow<'_, str>, ignore_first: bool, indent: Option<usize>) -> Cow<'_, str> { |
142 | let s_space = reindent_multiline_inner(&s, ignore_first, indent, ' '); | |
143 | let s_tab = reindent_multiline_inner(&s_space, ignore_first, indent, '\t'); | |
144 | reindent_multiline_inner(&s_tab, ignore_first, indent, ' ').into() | |
145 | } | |
146 | ||
147 | fn reindent_multiline_inner(s: &str, ignore_first: bool, indent: Option<usize>, ch: char) -> String { | |
148 | let x = s | |
149 | .lines() | |
a2a8927a | 150 | .skip(usize::from(ignore_first)) |
cdc7bbd5 XL |
151 | .filter_map(|l| { |
152 | if l.is_empty() { | |
153 | None | |
154 | } else { | |
155 | // ignore empty lines | |
156 | Some(l.char_indices().find(|&(_, x)| x != ch).unwrap_or((l.len(), ch)).0) | |
157 | } | |
158 | }) | |
159 | .min() | |
160 | .unwrap_or(0); | |
161 | let indent = indent.unwrap_or(0); | |
162 | s.lines() | |
163 | .enumerate() | |
164 | .map(|(i, l)| { | |
165 | if (ignore_first && i == 0) || l.is_empty() { | |
166 | l.to_owned() | |
167 | } else if x > indent { | |
168 | l.split_at(x - indent).1.to_owned() | |
169 | } else { | |
170 | " ".repeat(indent - x) + l | |
171 | } | |
172 | }) | |
173 | .collect::<Vec<String>>() | |
174 | .join("\n") | |
175 | } | |
176 | ||
a2a8927a | 177 | /// Converts a span to a code snippet if available, otherwise returns the default. |
cdc7bbd5 XL |
178 | /// |
179 | /// This is useful if you want to provide suggestions for your lint or more generally, if you want | |
a2a8927a XL |
180 | /// to convert a given `Span` to a `str`. To create suggestions consider using |
181 | /// [`snippet_with_applicability`] to ensure that the applicability stays correct. | |
cdc7bbd5 XL |
182 | /// |
183 | /// # Example | |
184 | /// ```rust,ignore | |
a2a8927a XL |
185 | /// // Given two spans one for `value` and one for the `init` expression. |
186 | /// let value = Vec::new(); | |
187 | /// // ^^^^^ ^^^^^^^^^^ | |
188 | /// // span1 span2 | |
189 | /// | |
190 | /// // The snipped call would return the corresponding code snippet | |
191 | /// snippet(cx, span1, "..") // -> "value" | |
192 | /// snippet(cx, span2, "..") // -> "Vec::new()" | |
cdc7bbd5 XL |
193 | /// ``` |
194 | pub fn snippet<'a, T: LintContext>(cx: &T, span: Span, default: &'a str) -> Cow<'a, str> { | |
195 | snippet_opt(cx, span).map_or_else(|| Cow::Borrowed(default), From::from) | |
196 | } | |
197 | ||
94222f64 | 198 | /// Same as [`snippet`], but it adapts the applicability level by following rules: |
cdc7bbd5 XL |
199 | /// |
200 | /// - Applicability level `Unspecified` will never be changed. | |
201 | /// - If the span is inside a macro, change the applicability level to `MaybeIncorrect`. | |
202 | /// - If the default value is used and the applicability level is `MachineApplicable`, change it to | |
203 | /// `HasPlaceholders` | |
204 | pub fn snippet_with_applicability<'a, T: LintContext>( | |
205 | cx: &T, | |
206 | span: Span, | |
207 | default: &'a str, | |
208 | applicability: &mut Applicability, | |
209 | ) -> Cow<'a, str> { | |
210 | if *applicability != Applicability::Unspecified && span.from_expansion() { | |
211 | *applicability = Applicability::MaybeIncorrect; | |
212 | } | |
213 | snippet_opt(cx, span).map_or_else( | |
214 | || { | |
215 | if *applicability == Applicability::MachineApplicable { | |
216 | *applicability = Applicability::HasPlaceholders; | |
217 | } | |
218 | Cow::Borrowed(default) | |
219 | }, | |
220 | From::from, | |
221 | ) | |
222 | } | |
223 | ||
224 | /// Same as `snippet`, but should only be used when it's clear that the input span is | |
225 | /// not a macro argument. | |
226 | pub fn snippet_with_macro_callsite<'a, T: LintContext>(cx: &T, span: Span, default: &'a str) -> Cow<'a, str> { | |
227 | snippet(cx, span.source_callsite(), default) | |
228 | } | |
229 | ||
230 | /// Converts a span to a code snippet. Returns `None` if not available. | |
231 | pub fn snippet_opt<T: LintContext>(cx: &T, span: Span) -> Option<String> { | |
232 | cx.sess().source_map().span_to_snippet(span).ok() | |
233 | } | |
234 | ||
235 | /// Converts a span (from a block) to a code snippet if available, otherwise use default. | |
236 | /// | |
237 | /// This trims the code of indentation, except for the first line. Use it for blocks or block-like | |
238 | /// things which need to be printed as such. | |
239 | /// | |
240 | /// The `indent_relative_to` arg can be used, to provide a span, where the indentation of the | |
241 | /// resulting snippet of the given span. | |
242 | /// | |
243 | /// # Example | |
244 | /// | |
245 | /// ```rust,ignore | |
246 | /// snippet_block(cx, block.span, "..", None) | |
247 | /// // where, `block` is the block of the if expr | |
248 | /// if x { | |
249 | /// y; | |
250 | /// } | |
251 | /// // will return the snippet | |
252 | /// { | |
253 | /// y; | |
254 | /// } | |
255 | /// ``` | |
256 | /// | |
257 | /// ```rust,ignore | |
258 | /// snippet_block(cx, block.span, "..", Some(if_expr.span)) | |
259 | /// // where, `block` is the block of the if expr | |
260 | /// if x { | |
261 | /// y; | |
262 | /// } | |
263 | /// // will return the snippet | |
264 | /// { | |
265 | /// y; | |
266 | /// } // aligned with `if` | |
267 | /// ``` | |
268 | /// Note that the first line of the snippet always has 0 indentation. | |
269 | pub fn snippet_block<'a, T: LintContext>( | |
270 | cx: &T, | |
271 | span: Span, | |
272 | default: &'a str, | |
273 | indent_relative_to: Option<Span>, | |
274 | ) -> Cow<'a, str> { | |
275 | let snip = snippet(cx, span, default); | |
276 | let indent = indent_relative_to.and_then(|s| indent_of(cx, s)); | |
277 | reindent_multiline(snip, true, indent) | |
278 | } | |
279 | ||
280 | /// Same as `snippet_block`, but adapts the applicability level by the rules of | |
281 | /// `snippet_with_applicability`. | |
282 | pub fn snippet_block_with_applicability<'a, T: LintContext>( | |
283 | cx: &T, | |
284 | span: Span, | |
285 | default: &'a str, | |
286 | indent_relative_to: Option<Span>, | |
287 | applicability: &mut Applicability, | |
288 | ) -> Cow<'a, str> { | |
289 | let snip = snippet_with_applicability(cx, span, default, applicability); | |
290 | let indent = indent_relative_to.and_then(|s| indent_of(cx, s)); | |
291 | reindent_multiline(snip, true, indent) | |
292 | } | |
293 | ||
294 | /// Same as `snippet_with_applicability`, but first walks the span up to the given context. This | |
295 | /// will result in the macro call, rather then the expansion, if the span is from a child context. | |
296 | /// If the span is not from a child context, it will be used directly instead. | |
297 | /// | |
298 | /// e.g. Given the expression `&vec![]`, getting a snippet from the span for `vec![]` as a HIR node | |
299 | /// would result in `box []`. If given the context of the address of expression, this function will | |
300 | /// correctly get a snippet of `vec![]`. | |
301 | /// | |
302 | /// This will also return whether or not the snippet is a macro call. | |
5099ac24 | 303 | pub fn snippet_with_context<'a>( |
cdc7bbd5 XL |
304 | cx: &LateContext<'_>, |
305 | span: Span, | |
306 | outer: SyntaxContext, | |
307 | default: &'a str, | |
308 | applicability: &mut Applicability, | |
309 | ) -> (Cow<'a, str>, bool) { | |
17df50a5 XL |
310 | let (span, is_macro_call) = walk_span_to_context(span, outer).map_or_else( |
311 | || { | |
312 | // The span is from a macro argument, and the outer context is the macro using the argument | |
313 | if *applicability != Applicability::Unspecified { | |
314 | *applicability = Applicability::MaybeIncorrect; | |
315 | } | |
316 | // TODO: get the argument span. | |
317 | (span, false) | |
318 | }, | |
319 | |outer_span| (outer_span, span.ctxt() != outer), | |
320 | ); | |
cdc7bbd5 XL |
321 | |
322 | ( | |
323 | snippet_with_applicability(cx, span, default, applicability), | |
324 | is_macro_call, | |
325 | ) | |
326 | } | |
327 | ||
17df50a5 XL |
328 | /// Walks the span up to the target context, thereby returning the macro call site if the span is |
329 | /// inside a macro expansion, or the original span if it is not. Note this will return `None` in the | |
330 | /// case of the span being in a macro expansion, but the target context is from expanding a macro | |
331 | /// argument. | |
332 | /// | |
333 | /// Given the following | |
334 | /// | |
335 | /// ```rust,ignore | |
336 | /// macro_rules! m { ($e:expr) => { f($e) }; } | |
337 | /// g(m!(0)) | |
338 | /// ``` | |
339 | /// | |
340 | /// If called with a span of the call to `f` and a context of the call to `g` this will return a | |
341 | /// span containing `m!(0)`. However, if called with a span of the literal `0` this will give a span | |
342 | /// containing `0` as the context is the same as the outer context. | |
343 | /// | |
344 | /// This will traverse through multiple macro calls. Given the following: | |
345 | /// | |
346 | /// ```rust,ignore | |
347 | /// macro_rules! m { ($e:expr) => { n!($e, 0) }; } | |
348 | /// macro_rules! n { ($e:expr, $f:expr) => { f($e, $f) }; } | |
349 | /// g(m!(0)) | |
350 | /// ``` | |
351 | /// | |
352 | /// If called with a span of the call to `f` and a context of the call to `g` this will return a | |
353 | /// span containing `m!(0)`. | |
354 | pub fn walk_span_to_context(span: Span, outer: SyntaxContext) -> Option<Span> { | |
355 | let outer_span = hygiene::walk_chain(span, outer); | |
356 | (outer_span.ctxt() == outer).then(|| outer_span) | |
357 | } | |
358 | ||
cdc7bbd5 XL |
359 | /// Removes block comments from the given `Vec` of lines. |
360 | /// | |
361 | /// # Examples | |
362 | /// | |
363 | /// ```rust,ignore | |
364 | /// without_block_comments(vec!["/*", "foo", "*/"]); | |
365 | /// // => vec![] | |
366 | /// | |
367 | /// without_block_comments(vec!["bar", "/*", "foo", "*/"]); | |
368 | /// // => vec!["bar"] | |
369 | /// ``` | |
370 | pub fn without_block_comments(lines: Vec<&str>) -> Vec<&str> { | |
371 | let mut without = vec![]; | |
372 | ||
373 | let mut nest_level = 0; | |
374 | ||
375 | for line in lines { | |
376 | if line.contains("/*") { | |
377 | nest_level += 1; | |
378 | continue; | |
379 | } else if line.contains("*/") { | |
380 | nest_level -= 1; | |
381 | continue; | |
382 | } | |
383 | ||
384 | if nest_level == 0 { | |
385 | without.push(line); | |
386 | } | |
387 | } | |
388 | ||
389 | without | |
390 | } | |
391 | ||
923072b8 FG |
392 | /// Trims the whitespace from the start and the end of the span. |
393 | pub fn trim_span(sm: &SourceMap, span: Span) -> Span { | |
394 | let data = span.data(); | |
395 | let sf: &_ = &sm.lookup_source_file(data.lo); | |
396 | let Some(src) = sf.src.as_deref() else { | |
397 | return span; | |
398 | }; | |
399 | let Some(snip) = &src.get((data.lo - sf.start_pos).to_usize()..(data.hi - sf.start_pos).to_usize()) else { | |
400 | return span; | |
401 | }; | |
402 | let trim_start = snip.len() - snip.trim_start().len(); | |
403 | let trim_end = snip.len() - snip.trim_end().len(); | |
404 | SpanData { | |
405 | lo: data.lo + BytePos::from_usize(trim_start), | |
406 | hi: data.hi - BytePos::from_usize(trim_end), | |
407 | ctxt: data.ctxt, | |
408 | parent: data.parent, | |
409 | } | |
410 | .span() | |
411 | } | |
412 | ||
cdc7bbd5 XL |
413 | #[cfg(test)] |
414 | mod test { | |
415 | use super::{reindent_multiline, without_block_comments}; | |
416 | ||
417 | #[test] | |
418 | fn test_reindent_multiline_single_line() { | |
419 | assert_eq!("", reindent_multiline("".into(), false, None)); | |
420 | assert_eq!("...", reindent_multiline("...".into(), false, None)); | |
421 | assert_eq!("...", reindent_multiline(" ...".into(), false, None)); | |
422 | assert_eq!("...", reindent_multiline("\t...".into(), false, None)); | |
423 | assert_eq!("...", reindent_multiline("\t\t...".into(), false, None)); | |
424 | } | |
425 | ||
426 | #[test] | |
427 | #[rustfmt::skip] | |
428 | fn test_reindent_multiline_block() { | |
429 | assert_eq!("\ | |
430 | if x { | |
431 | y | |
432 | } else { | |
433 | z | |
434 | }", reindent_multiline(" if x { | |
435 | y | |
436 | } else { | |
437 | z | |
438 | }".into(), false, None)); | |
439 | assert_eq!("\ | |
440 | if x { | |
441 | \ty | |
442 | } else { | |
443 | \tz | |
444 | }", reindent_multiline(" if x { | |
445 | \ty | |
446 | } else { | |
447 | \tz | |
448 | }".into(), false, None)); | |
449 | } | |
450 | ||
451 | #[test] | |
452 | #[rustfmt::skip] | |
453 | fn test_reindent_multiline_empty_line() { | |
454 | assert_eq!("\ | |
455 | if x { | |
456 | y | |
457 | ||
458 | } else { | |
459 | z | |
460 | }", reindent_multiline(" if x { | |
461 | y | |
462 | ||
463 | } else { | |
464 | z | |
465 | }".into(), false, None)); | |
466 | } | |
467 | ||
468 | #[test] | |
469 | #[rustfmt::skip] | |
470 | fn test_reindent_multiline_lines_deeper() { | |
471 | assert_eq!("\ | |
472 | if x { | |
473 | y | |
474 | } else { | |
475 | z | |
476 | }", reindent_multiline("\ | |
477 | if x { | |
478 | y | |
479 | } else { | |
480 | z | |
481 | }".into(), true, Some(8))); | |
482 | } | |
483 | ||
484 | #[test] | |
485 | fn test_without_block_comments_lines_without_block_comments() { | |
486 | let result = without_block_comments(vec!["/*", "", "*/"]); | |
487 | println!("result: {:?}", result); | |
488 | assert!(result.is_empty()); | |
489 | ||
490 | let result = without_block_comments(vec!["", "/*", "", "*/", "#[crate_type = \"lib\"]", "/*", "", "*/", ""]); | |
491 | assert_eq!(result, vec!["", "#[crate_type = \"lib\"]", ""]); | |
492 | ||
493 | let result = without_block_comments(vec!["/* rust", "", "*/"]); | |
494 | assert!(result.is_empty()); | |
495 | ||
496 | let result = without_block_comments(vec!["/* one-line comment */"]); | |
497 | assert!(result.is_empty()); | |
498 | ||
499 | let result = without_block_comments(vec!["/* nested", "/* multi-line", "comment", "*/", "test", "*/"]); | |
500 | assert!(result.is_empty()); | |
501 | ||
502 | let result = without_block_comments(vec!["/* nested /* inline /* comment */ test */ */"]); | |
503 | assert!(result.is_empty()); | |
504 | ||
505 | let result = without_block_comments(vec!["foo", "bar", "baz"]); | |
506 | assert_eq!(result, vec!["foo", "bar", "baz"]); | |
507 | } | |
508 | } |