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
|
use std::ops::Range;
use std::time::Instant;
use imara_diff::intern::InternedInput;
use imara_diff::Algorithm;
use ropey::RopeSlice;
use crate::{ChangeSet, Rope, Tendril, Transaction};
/// A `imara_diff::Sink` that builds a `ChangeSet` for a character diff of a hunk
struct CharChangeSetBuilder<'a> {
res: &'a mut ChangeSet,
hunk: &'a InternedInput<char>,
pos: u32,
}
impl imara_diff::Sink for CharChangeSetBuilder<'_> {
type Out = ();
fn process_change(&mut self, before: Range<u32>, after: Range<u32>) {
self.res.retain((before.start - self.pos) as usize);
self.res.delete(before.len());
self.pos = before.end;
let res = self.hunk.after[after.start as usize..after.end as usize]
.iter()
.map(|&token| self.hunk.interner[token])
.collect();
self.res.insert(res);
}
fn finish(self) -> Self::Out {
self.res.retain(self.hunk.before.len() - self.pos as usize);
}
}
struct LineChangeSetBuilder<'a> {
res: ChangeSet,
after: RopeSlice<'a>,
file: &'a InternedInput<RopeSlice<'a>>,
current_hunk: InternedInput<char>,
pos: u32,
}
impl imara_diff::Sink for LineChangeSetBuilder<'_> {
type Out = ChangeSet;
fn process_change(&mut self, before: Range<u32>, after: Range<u32>) {
let len = self.file.before[self.pos as usize..before.start as usize]
.iter()
.map(|&it| self.file.interner[it].len_chars())
.sum();
self.res.retain(len);
self.pos = before.end;
// do not perform diffs on large hunks
let len_before = before.end - before.start;
let len_after = after.end - after.start;
// Pure insertions/removals do not require a character diff.
// Very large changes are ignored because their character diff is expensive to compute
// TODO adjust heuristic to detect large changes?
if len_before == 0
|| len_after == 0
|| len_after > 5 * len_before
|| 5 * len_after < len_before && len_before > 10
|| len_before + len_after > 200
{
let remove = self.file.before[before.start as usize..before.end as usize]
.iter()
.map(|&it| self.file.interner[it].len_chars())
.sum();
self.res.delete(remove);
let mut fragment = Tendril::new();
if len_after > 500 {
// copying a rope line by line is slower then copying the entire
// rope. Use to_string for very large changes instead..
if self.file.after.len() == after.end as usize {
if after.start == 0 {
fragment = self.after.to_string().into();
} else {
let start = self.after.line_to_char(after.start as usize);
fragment = self.after.slice(start..).to_string().into();
}
} else if after.start == 0 {
let end = self.after.line_to_char(after.end as usize);
fragment = self.after.slice(..end).to_string().into();
} else {
let start = self.after.line_to_char(after.start as usize);
let end = self.after.line_to_char(after.end as usize);
fragment = self.after.slice(start..end).to_string().into();
}
} else {
for &line in &self.file.after[after.start as usize..after.end as usize] {
for chunk in self.file.interner[line].chunks() {
fragment.push_str(chunk)
}
}
};
self.res.insert(fragment);
} else {
// for reasonably small hunks, generating a ChangeSet from char diff can save memory
// TODO use a tokenizer (word diff?) for improved performance
let hunk_before = self.file.before[before.start as usize..before.end as usize]
.iter()
.flat_map(|&it| self.file.interner[it].chars());
let hunk_after = self.file.after[after.start as usize..after.end as usize]
.iter()
.flat_map(|&it| self.file.interner[it].chars());
self.current_hunk.update_before(hunk_before);
self.current_hunk.update_after(hunk_after);
// the histogram heuristic does not work as well
// for characters because the same characters often reoccur
// use myer diff instead
imara_diff::diff(
Algorithm::Myers,
&self.current_hunk,
CharChangeSetBuilder {
res: &mut self.res,
hunk: &self.current_hunk,
pos: 0,
},
);
self.current_hunk.clear();
}
}
fn finish(mut self) -> Self::Out {
let len = self.file.before[self.pos as usize..]
.iter()
.map(|&it| self.file.interner[it].len_chars())
.sum();
self.res.retain(len);
self.res
}
}
struct RopeLines<'a>(RopeSlice<'a>);
impl<'a> imara_diff::intern::TokenSource for RopeLines<'a> {
type Token = RopeSlice<'a>;
type Tokenizer = ropey::iter::Lines<'a>;
fn tokenize(&self) -> Self::Tokenizer {
self.0.lines()
}
fn estimate_tokens(&self) -> u32 {
// we can provide a perfect estimate which is very nice for performance
self.0.len_lines() as u32
}
}
/// Compares `old` and `new` to generate a [`Transaction`] describing
/// the steps required to get from `old` to `new`.
pub fn compare_ropes(before: &Rope, after: &Rope) -> Transaction {
let start = Instant::now();
let res = ChangeSet::with_capacity(32);
let after = after.slice(..);
let file = InternedInput::new(RopeLines(before.slice(..)), RopeLines(after));
let builder = LineChangeSetBuilder {
res,
file: &file,
after,
pos: 0,
current_hunk: InternedInput::default(),
};
let res = imara_diff::diff(Algorithm::Histogram, &file, builder).into();
log::debug!(
"rope diff took {}s",
Instant::now().duration_since(start).as_secs_f64()
);
res
}
#[cfg(test)]
mod tests {
use super::*;
fn test_identity(a: &str, b: &str) {
let mut old = Rope::from(a);
let new = Rope::from(b);
compare_ropes(&old, &new).apply(&mut old);
assert_eq!(old, new);
}
quickcheck::quickcheck! {
fn test_compare_ropes(a: String, b: String) -> bool {
let mut old = Rope::from(a);
let new = Rope::from(b);
compare_ropes(&old, &new).apply(&mut old);
old == new
}
}
#[test]
fn equal_files() {
test_identity("foo", "foo");
}
#[test]
fn trailing_newline() {
test_identity("foo\n", "foo");
test_identity("foo", "foo\n");
}
#[test]
fn new_file() {
test_identity("", "foo");
}
#[test]
fn deleted_file() {
test_identity("foo", "");
}
}
|
