rustc_span/
caching_source_map_view.rs

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
use std::ops::Range;

use rustc_data_structures::sync::Lrc;

use crate::source_map::SourceMap;
use crate::{BytePos, Pos, RelativeBytePos, SourceFile, SpanData};

#[derive(Clone)]
struct CacheEntry {
    time_stamp: usize,
    line_number: usize,
    // The line's byte position range in the `SourceMap`. This range will fail to contain a valid
    // position in certain edge cases. Spans often start/end one past something, and when that
    // something is the last character of a file (this can happen when a file doesn't end in a
    // newline, for example), we'd still like for the position to be considered within the last
    // line. However, it isn't according to the exclusive upper bound of this range. We cannot
    // change the upper bound to be inclusive, because for most lines, the upper bound is the same
    // as the lower bound of the next line, so there would be an ambiguity.
    //
    // Since the containment aspect of this range is only used to see whether or not the cache
    // entry contains a position, the only ramification of the above is that we will get cache
    // misses for these rare positions. A line lookup for the position via `SourceMap::lookup_line`
    // after a cache miss will produce the last line number, as desired.
    line: Range<BytePos>,
    file: Lrc<SourceFile>,
    file_index: usize,
}

impl CacheEntry {
    #[inline]
    fn update(
        &mut self,
        new_file_and_idx: Option<(Lrc<SourceFile>, usize)>,
        pos: BytePos,
        time_stamp: usize,
    ) {
        if let Some((file, file_idx)) = new_file_and_idx {
            self.file = file;
            self.file_index = file_idx;
        }

        let pos = self.file.relative_position(pos);
        let line_index = self.file.lookup_line(pos).unwrap();
        let line_bounds = self.file.line_bounds(line_index);
        self.line_number = line_index + 1;
        self.line = line_bounds;
        self.touch(time_stamp);
    }

    #[inline]
    fn touch(&mut self, time_stamp: usize) {
        self.time_stamp = time_stamp;
    }
}

#[derive(Clone)]
pub struct CachingSourceMapView<'sm> {
    source_map: &'sm SourceMap,
    line_cache: [CacheEntry; 3],
    time_stamp: usize,
}

impl<'sm> CachingSourceMapView<'sm> {
    pub fn new(source_map: &'sm SourceMap) -> CachingSourceMapView<'sm> {
        let files = source_map.files();
        let first_file = Lrc::clone(&files[0]);
        let entry = CacheEntry {
            time_stamp: 0,
            line_number: 0,
            line: BytePos(0)..BytePos(0),
            file: first_file,
            file_index: 0,
        };

        CachingSourceMapView {
            source_map,
            line_cache: [entry.clone(), entry.clone(), entry],
            time_stamp: 0,
        }
    }

    pub fn byte_pos_to_line_and_col(
        &mut self,
        pos: BytePos,
    ) -> Option<(Lrc<SourceFile>, usize, RelativeBytePos)> {
        self.time_stamp += 1;

        // Check if the position is in one of the cached lines
        let cache_idx = self.cache_entry_index(pos);
        if cache_idx != -1 {
            let cache_entry = &mut self.line_cache[cache_idx as usize];
            cache_entry.touch(self.time_stamp);

            let col = RelativeBytePos(pos.to_u32() - cache_entry.line.start.to_u32());
            return Some((Lrc::clone(&cache_entry.file), cache_entry.line_number, col));
        }

        // No cache hit ...
        let oldest = self.oldest_cache_entry_index();

        // If the entry doesn't point to the correct file, get the new file and index.
        let new_file_and_idx = if !file_contains(&self.line_cache[oldest].file, pos) {
            Some(self.file_for_position(pos)?)
        } else {
            None
        };

        let cache_entry = &mut self.line_cache[oldest];
        cache_entry.update(new_file_and_idx, pos, self.time_stamp);

        let col = RelativeBytePos(pos.to_u32() - cache_entry.line.start.to_u32());
        Some((Lrc::clone(&cache_entry.file), cache_entry.line_number, col))
    }

    pub fn span_data_to_lines_and_cols(
        &mut self,
        span_data: &SpanData,
    ) -> Option<(Lrc<SourceFile>, usize, BytePos, usize, BytePos)> {
        self.time_stamp += 1;

        // Check if lo and hi are in the cached lines.
        let lo_cache_idx: isize = self.cache_entry_index(span_data.lo);
        let hi_cache_idx = self.cache_entry_index(span_data.hi);

        if lo_cache_idx != -1 && hi_cache_idx != -1 {
            // Cache hit for span lo and hi. Check if they belong to the same file.
            let result = {
                let lo = &self.line_cache[lo_cache_idx as usize];
                let hi = &self.line_cache[hi_cache_idx as usize];

                if lo.file_index != hi.file_index {
                    return None;
                }

                (
                    Lrc::clone(&lo.file),
                    lo.line_number,
                    span_data.lo - lo.line.start,
                    hi.line_number,
                    span_data.hi - hi.line.start,
                )
            };

            self.line_cache[lo_cache_idx as usize].touch(self.time_stamp);
            self.line_cache[hi_cache_idx as usize].touch(self.time_stamp);

            return Some(result);
        }

        // No cache hit or cache hit for only one of span lo and hi.
        let oldest = if lo_cache_idx != -1 || hi_cache_idx != -1 {
            let avoid_idx = if lo_cache_idx != -1 { lo_cache_idx } else { hi_cache_idx };
            self.oldest_cache_entry_index_avoid(avoid_idx as usize)
        } else {
            self.oldest_cache_entry_index()
        };

        // If the entry doesn't point to the correct file, get the new file and index.
        // Return early if the file containing beginning of span doesn't contain end of span.
        let new_file_and_idx = if !file_contains(&self.line_cache[oldest].file, span_data.lo) {
            let new_file_and_idx = self.file_for_position(span_data.lo)?;
            if !file_contains(&new_file_and_idx.0, span_data.hi) {
                return None;
            }

            Some(new_file_and_idx)
        } else {
            let file = &self.line_cache[oldest].file;
            if !file_contains(file, span_data.hi) {
                return None;
            }

            None
        };

        // Update the cache entries.
        let (lo_idx, hi_idx) = match (lo_cache_idx, hi_cache_idx) {
            // Oldest cache entry is for span_data.lo line.
            (-1, -1) => {
                let lo = &mut self.line_cache[oldest];
                lo.update(new_file_and_idx, span_data.lo, self.time_stamp);

                if !lo.line.contains(&span_data.hi) {
                    let new_file_and_idx = Some((Lrc::clone(&lo.file), lo.file_index));
                    let next_oldest = self.oldest_cache_entry_index_avoid(oldest);
                    let hi = &mut self.line_cache[next_oldest];
                    hi.update(new_file_and_idx, span_data.hi, self.time_stamp);
                    (oldest, next_oldest)
                } else {
                    (oldest, oldest)
                }
            }
            // Oldest cache entry is for span_data.lo line.
            (-1, _) => {
                let lo = &mut self.line_cache[oldest];
                lo.update(new_file_and_idx, span_data.lo, self.time_stamp);
                let hi = &mut self.line_cache[hi_cache_idx as usize];
                hi.touch(self.time_stamp);
                (oldest, hi_cache_idx as usize)
            }
            // Oldest cache entry is for span_data.hi line.
            (_, -1) => {
                let hi = &mut self.line_cache[oldest];
                hi.update(new_file_and_idx, span_data.hi, self.time_stamp);
                let lo = &mut self.line_cache[lo_cache_idx as usize];
                lo.touch(self.time_stamp);
                (lo_cache_idx as usize, oldest)
            }
            _ => {
                panic!(
                    "the case of neither value being equal to -1 was handled above and the function returns."
                );
            }
        };

        let lo = &self.line_cache[lo_idx];
        let hi = &self.line_cache[hi_idx];

        // Span lo and hi may equal line end when last line doesn't
        // end in newline, hence the inclusive upper bounds below.
        assert!(span_data.lo >= lo.line.start);
        assert!(span_data.lo <= lo.line.end);
        assert!(span_data.hi >= hi.line.start);
        assert!(span_data.hi <= hi.line.end);
        assert!(lo.file.contains(span_data.lo));
        assert!(lo.file.contains(span_data.hi));
        assert_eq!(lo.file_index, hi.file_index);

        Some((
            Lrc::clone(&lo.file),
            lo.line_number,
            span_data.lo - lo.line.start,
            hi.line_number,
            span_data.hi - hi.line.start,
        ))
    }

    fn cache_entry_index(&self, pos: BytePos) -> isize {
        for (idx, cache_entry) in self.line_cache.iter().enumerate() {
            if cache_entry.line.contains(&pos) {
                return idx as isize;
            }
        }

        -1
    }

    fn oldest_cache_entry_index(&self) -> usize {
        let mut oldest = 0;

        for idx in 1..self.line_cache.len() {
            if self.line_cache[idx].time_stamp < self.line_cache[oldest].time_stamp {
                oldest = idx;
            }
        }

        oldest
    }

    fn oldest_cache_entry_index_avoid(&self, avoid_idx: usize) -> usize {
        let mut oldest = if avoid_idx != 0 { 0 } else { 1 };

        for idx in 0..self.line_cache.len() {
            if idx != avoid_idx
                && self.line_cache[idx].time_stamp < self.line_cache[oldest].time_stamp
            {
                oldest = idx;
            }
        }

        oldest
    }

    fn file_for_position(&self, pos: BytePos) -> Option<(Lrc<SourceFile>, usize)> {
        if !self.source_map.files().is_empty() {
            let file_idx = self.source_map.lookup_source_file_idx(pos);
            let file = &self.source_map.files()[file_idx];

            if file_contains(file, pos) {
                return Some((Lrc::clone(file), file_idx));
            }
        }

        None
    }
}

#[inline]
fn file_contains(file: &SourceFile, pos: BytePos) -> bool {
    // `SourceMap::lookup_source_file_idx` and `SourceFile::contains` both consider the position
    // one past the end of a file to belong to it. Normally, that's what we want. But for the
    // purposes of converting a byte position to a line and column number, we can't come up with a
    // line and column number if the file is empty, because an empty file doesn't contain any
    // lines. So for our purposes, we don't consider empty files to contain any byte position.
    file.contains(pos) && !file.is_empty()
}