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
//! The rolling receive buffer and the pattern matcher that scans it.
use memchr::memmem;
// At 115200 baud a console fills slowly, so a few hundred KB keeps enough
// scrollback that expect() can still find a line that scrolled past a moment
// ago. The buffer can grow to twice this before it is trimmed.
pub const DEFAULT_RING_CAP: usize = 512 * 1024;
/// A rolling window of recently received bytes, addressed by an absolute offset
/// so a client can page through with a cursor even after old bytes are dropped.
pub struct Ring {
buf: Vec<u8>,
base: u64, // absolute offset of buf[0]
cap: usize,
}
impl Ring {
pub fn new(cap: usize) -> Self {
Self {
buf: Vec::new(),
base: 0,
cap: cap.max(1),
}
}
pub fn total(&self) -> u64 {
self.base + self.buf.len() as u64
}
pub fn append(&mut self, bytes: &[u8]) {
self.buf.extend_from_slice(bytes);
// Trim in large chunks. Draining on every append would shift the whole
// buffer for each received chunk once full. Letting it grow to twice the
// cap first makes the shift amortized O(1) per byte.
if self.buf.len() > self.cap * 2 {
let drop = self.buf.len() - self.cap;
self.buf.drain(..drop);
self.base += drop as u64;
}
}
// Bytes from `cursor` to the end, plus the absolute offset the slice starts
// at. `cursor` is clamped into the retained window, so a cursor pointing at
// bytes already dropped simply starts at the oldest retained byte.
pub fn slice_from(&self, cursor: u64) -> (u64, &[u8]) {
let start = cursor.clamp(self.base, self.total());
let idx = (start - self.base) as usize;
(start, &self.buf[idx..])
}
pub fn base(&self) -> u64 {
self.base
}
pub fn tail_lines(&self, lines: usize) -> String {
if lines == 0 {
return String::new();
}
// Scan backwards for the newline where the requested tail starts, so the
// cost tracks the answer size instead of the whole retained buffer. A
// trailing newline only terminates the last line, it does not start a
// new one.
let buf = &self.buf;
let end = buf.len() - usize::from(buf.last() == Some(&b'\n'));
let mut count = 0;
let mut start = 0;
for i in (0..end).rev() {
if buf[i] == b'\n' {
count += 1;
if count == lines {
start = i + 1;
break;
}
}
}
let tail = String::from_utf8_lossy(&buf[start..]);
let all: Vec<&str> = tail.lines().collect();
all.join("\n")
}
}
/// Either a plain substring or a compiled regex, matched against raw bytes so
/// the offsets it returns line up with the ring.
pub enum Matcher {
Substr(Vec<u8>),
Regex(regex::bytes::Regex),
}
impl Matcher {
pub fn build(pattern: &str, regex: bool) -> Result<Self, String> {
if regex {
regex::bytes::Regex::new(pattern)
.map(Matcher::Regex)
.map_err(|e| e.to_string())
} else {
Ok(Matcher::Substr(pattern.as_bytes().to_vec()))
}
}
// Offset just past the first match in `hay`, or None.
pub fn find_end(&self, hay: &[u8]) -> Option<usize> {
match self {
Matcher::Substr(needle) => memmem::find(hay, needle).map(|i| i + needle.len()),
Matcher::Regex(re) => re.find(hay).map(|m| m.end()),
}
}
// Where the next scan can resume after a miss that ended at absolute offset
// `end`. A substring hit can straddle the scanned region's edge by at most
// needle length minus one, so scanning restarts just before it. A regex can
// match a span of any length, so it rescans from `start` every time.
pub fn resume_from(&self, start: u64, end: u64) -> u64 {
match self {
Matcher::Substr(needle) => {
let overlap = needle.len().saturating_sub(1) as u64;
end.saturating_sub(overlap).max(start)
}
Matcher::Regex(_) => start,
}
}
}
#[cfg(test)]
mod tests {
use super::*;
#[test]
fn ring_drops_oldest_and_tracks_offset() {
let mut ring = Ring::new(DEFAULT_RING_CAP);
ring.append(b"hello ");
ring.append(b"world");
assert_eq!(ring.total(), 11);
let (abs, slice) = ring.slice_from(6);
assert_eq!(abs, 6);
assert_eq!(slice, b"world");
}
#[test]
fn ring_cursor_before_window_clamps_to_base() {
let mut ring = Ring::new(DEFAULT_RING_CAP);
ring.append(b"abc");
let (abs, slice) = ring.slice_from(0);
assert_eq!(abs, 0);
assert_eq!(slice, b"abc");
}
// A small cap makes the trim path cheap to test. The window keeps the cap
// after a trim, and the base moves by exactly what was dropped.
#[test]
fn ring_trims_to_cap_and_moves_base() {
let mut ring = Ring::new(4);
ring.append(&[b'x'; 9]);
assert_eq!(ring.total(), 9);
assert_eq!(ring.base(), 5);
let (abs, slice) = ring.slice_from(0);
assert_eq!(abs, 5);
assert_eq!(slice.len(), 4);
}
#[test]
fn tail_lines_returns_last_lines() {
let mut ring = Ring::new(DEFAULT_RING_CAP);
ring.append(b"one\r\ntwo\r\nthree\r\npartial");
assert_eq!(ring.tail_lines(2), "three\npartial");
assert_eq!(ring.tail_lines(10), "one\ntwo\nthree\npartial");
assert_eq!(ring.tail_lines(0), "");
}
#[test]
fn tail_lines_ignores_trailing_newline() {
let mut ring = Ring::new(DEFAULT_RING_CAP);
ring.append(b"a\nb\n");
assert_eq!(ring.tail_lines(1), "b");
assert_eq!(ring.tail_lines(2), "a\nb");
}
#[test]
fn substr_match_returns_offset_past_match() {
let m = Matcher::build("ready> ", false).unwrap();
// "ready> " sits at bytes 11..18, so the offset just past it is 18.
assert_eq!(m.find_end(b"value = 1\r\nready> "), Some(18));
assert_eq!(m.find_end(b"still running"), None);
}
#[test]
fn regex_match_finds_tagged_line() {
let m = Matcher::build(r"\(T\d\)", true).unwrap();
// "(T2)" sits at bytes 8..12, so the offset just past it is 12.
assert_eq!(m.find_end(b"timeout (T2) 61 sec"), Some(12));
assert_eq!(m.find_end(b"timeout pending"), None);
}
#[test]
fn substr_scan_resumes_with_overlap_and_regex_rescans() {
let sub = Matcher::build("abc", false).unwrap();
assert_eq!(sub.resume_from(0, 100), 98);
assert_eq!(sub.resume_from(99, 100), 99); // never before the start cursor
let re = Matcher::build("a.*b", true).unwrap();
assert_eq!(re.resume_from(5, 100), 5);
}
}