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
//! A map that maps a span to every position in a file. Usually maps a span to some range of positions.
//! Allows bidirectional lookup.
use std::{fmt, hash::Hash};
use stdx::{always, itertools::Itertools};
use crate::{
EditionedFileId, ErasedFileAstId, ROOT_ERASED_FILE_AST_ID, Span, SpanAnchor, SyntaxContext,
TextRange, TextSize,
};
/// Maps absolute text ranges for the corresponding file to the relevant span data.
#[derive(Debug, PartialEq, Eq, Clone, Hash)]
pub struct SpanMap {
/// The offset stored here is the *end* of the node.
spans: Vec<(TextSize, Span)>,
/// Index of the matched macro arm on successful expansion for declarative macros.
// FIXME: Does it make sense to have this here?
pub matched_arm: Option<u32>,
}
impl SpanMap {
/// Creates a new empty [`SpanMap`].
pub fn empty() -> Self {
Self { spans: Vec::new(), matched_arm: None }
}
/// Finalizes the [`SpanMap`], shrinking its backing storage and validating that the offsets are
/// in order.
pub fn finish(&mut self) {
always!(
self.spans.iter().tuple_windows().all(|(a, b)| a.0 < b.0),
"spans are not in order"
);
self.spans.shrink_to_fit();
}
/// Pushes a new span onto the [`SpanMap`].
pub fn push(&mut self, offset: TextSize, span: Span) {
if cfg!(debug_assertions)
&& let Some(&(last_offset, _)) = self.spans.last()
{
assert!(
last_offset < offset,
"last_offset({last_offset:?}) must be smaller than offset({offset:?})"
);
}
self.spans.push((offset, span));
}
/// Returns all [`TextRange`]s that correspond to the given span.
///
/// Note this does a linear search through the entire backing vector.
pub fn ranges_with_span_exact(
&self,
span: Span,
) -> impl Iterator<Item = (TextRange, SyntaxContext)> + '_ {
self.spans.iter().enumerate().filter_map(move |(idx, &(end, s))| {
if !s.eq_ignoring_ctx(span) {
return None;
}
let start = idx.checked_sub(1).map_or(TextSize::new(0), |prev| self.spans[prev].0);
Some((TextRange::new(start, end), s.ctx))
})
}
/// Returns all [`TextRange`]s whose spans contain the given span.
///
/// Note this does a linear search through the entire backing vector.
pub fn ranges_with_span(
&self,
span: Span,
) -> impl Iterator<Item = (TextRange, SyntaxContext)> + '_ {
self.spans.iter().enumerate().filter_map(move |(idx, &(end, s))| {
if s.anchor != span.anchor {
return None;
}
if !s.range.contains_range(span.range) {
return None;
}
let start = idx.checked_sub(1).map_or(TextSize::new(0), |prev| self.spans[prev].0);
Some((TextRange::new(start, end), s.ctx))
})
}
/// Returns the span at the given position.
pub fn span_at(&self, offset: TextSize) -> Span {
let entry = self.spans.partition_point(|&(it, _)| it <= offset);
self.spans[entry].1
}
/// Returns the spans associated with the given range.
/// In other words, this will return all spans that correspond to all offsets within the given range.
pub fn spans_for_range(&self, range: TextRange) -> impl Iterator<Item = Span> + '_ {
let (start, end) = (range.start(), range.end());
let start_entry = self.spans.partition_point(|&(it, _)| it <= start);
let end_entry = self.spans[start_entry..].partition_point(|&(it, _)| it <= end); // FIXME: this might be wrong?
self.spans[start_entry..][..end_entry].iter().map(|&(_, s)| s)
}
pub fn iter(&self) -> impl Iterator<Item = (TextSize, Span)> + '_ {
self.spans.iter().copied()
}
/// Merges this span map with another span map, where `other` is inserted at (and replaces) `other_range`.
///
/// The length of the replacement node needs to be `other_size`.
pub fn merge(&mut self, other_range: TextRange, other_size: TextSize, other: &SpanMap) {
// I find the following diagram helpful to illustrate the bounds and why we use `<` or `<=`:
// --------------------------------------------------------------------
// 1 3 5 6 7 10 11 <-- offsets we store
// 0-1 1-3 3-5 5-6 6-7 7-10 10-11 <-- ranges these offsets refer to
// 3 .. 7 <-- other_range
// 3-5 5-6 6-7 <-- ranges we replace (len = 7-3 = 4)
// ^^^^^^^^^^^ ^^^^^^^^^^
// remove shift
// 2 3 5 9 <-- offsets we insert
// 0-2 2-3 3-5 5-9 <-- ranges we insert (other_size = 9-0 = 9)
// ------------------------------------
// 1 3
// 0-1 1-3 <-- these remain intact
// 5 6 8 12
// 3-5 5-6 6-8 8-12 <-- we shift these by other_range.start() and insert them
// 15 16
// 12-15 15-16 <-- we shift these by other_size-other_range.len() = 9-4 = 5
// ------------------------------------
// 1 3 5 6 8 12 15 16 <-- final offsets we store
// 0-1 1-3 3-5 5-6 6-8 8-12 12-15 15-16 <-- final ranges
self.spans.retain_mut(|(offset, _)| {
if other_range.start() < *offset && *offset <= other_range.end() {
false
} else {
if *offset > other_range.end() {
*offset += other_size;
*offset -= other_range.len();
}
true
}
});
self.spans
.extend(other.spans.iter().map(|&(offset, span)| (offset + other_range.start(), span)));
self.spans.sort_unstable_by_key(|&(offset, _)| offset);
// Matched arm info is no longer correct once we have multiple macros.
self.matched_arm = None;
}
}
#[cfg(not(no_salsa_async_drops))]
impl Drop for SpanMap {
fn drop(&mut self) {
let spans = std::mem::take(&mut self.spans);
static SPAN_MAP_DROP_THREAD: std::sync::OnceLock<
std::sync::mpsc::Sender<Vec<(TextSize, Span)>>,
> = std::sync::OnceLock::new();
SPAN_MAP_DROP_THREAD
.get_or_init(|| {
let (sender, receiver) = std::sync::mpsc::channel::<Vec<(TextSize, Span)>>();
std::thread::Builder::new()
.name("SpanMapDropper".to_owned())
.spawn(move || {
loop {
// block on a receive
_ = receiver.recv();
// then drain the entire channel
while receiver.try_recv().is_ok() {}
// and sleep for a bit
std::thread::sleep(std::time::Duration::from_millis(100));
}
// why do this over just a `receiver.iter().for_each(drop)`? To reduce contention on the channel lock.
// otherwise this thread will constantly wake up and sleep again.
})
.unwrap();
sender
})
.send(spans)
.unwrap();
}
}
#[derive(PartialEq, Eq, Hash, Debug)]
pub struct RealSpanMap {
file_id: EditionedFileId,
/// Invariant: Sorted vec over TextSize
// FIXME: SortedVec<(TextSize, ErasedFileAstId)>?
pairs: Box<[(TextSize, ErasedFileAstId)]>,
end: TextSize,
}
impl fmt::Display for RealSpanMap {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
writeln!(f, "RealSpanMap({:?}):", self.file_id)?;
for span in self.pairs.iter() {
writeln!(f, "{}: {:#?}", u32::from(span.0), span.1)?;
}
Ok(())
}
}
impl RealSpanMap {
/// Creates a real file span map that returns absolute ranges (relative ranges to the root ast id).
pub fn absolute(file_id: EditionedFileId) -> Self {
RealSpanMap {
file_id,
pairs: Box::from([(TextSize::new(0), ROOT_ERASED_FILE_AST_ID)]),
end: TextSize::new(!0),
}
}
pub fn from_file(
file_id: EditionedFileId,
pairs: Box<[(TextSize, ErasedFileAstId)]>,
end: TextSize,
) -> Self {
Self { file_id, pairs, end }
}
pub fn span_for_range(&self, range: TextRange) -> Span {
assert!(
range.end() <= self.end,
"range {range:?} goes beyond the end of the file {:?}",
self.end
);
let start = range.start();
let idx = self
.pairs
.binary_search_by(|&(it, _)| it.cmp(&start).then(std::cmp::Ordering::Less))
.unwrap_err();
let (offset, ast_id) = self.pairs[idx - 1];
Span {
range: range - offset,
anchor: SpanAnchor { file_id: self.file_id, ast_id },
ctx: SyntaxContext::root(self.file_id.edition()),
}
}
}