lostecho/ripgrep
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

libripgrep: initial commit introducing libripgrep

Browse Source 
libripgrep is not any one library, but rather, a collection of libraries
that roughly separate the following key distinct phases in a grep
implementation:

  1. Pattern matching (e.g., by a regex engine).
  2. Searching a file using a pattern matcher.
  3. Printing results.

Ultimately, both (1) and (3) are defined by de-coupled interfaces, of
which there may be multiple implementations. Namely, (1) is satisfied by
the `Matcher` trait in the `grep-matcher` crate and (3) is satisfied by
the `Sink` trait in the `grep2` crate. The searcher (2) ties everything
together and finds results using a matcher and reports those results
using a `Sink` implementation.

Closes #162
This commit is contained in:
Andrew Gallant
2018-04-29 09:29:52 -04:00
parent 0958837ee1
commit d9ca529356
68 changed files with 18010 additions and 20 deletions
Download Patch File Download Diff File Expand all files Collapse all files

135
grep-searcher/src/lib.rs Normal file
View File

@@ -0,0 +1,135 @@
/*!
This crate provides an implementation of line oriented search, with optional
support for multi-line search.
# Brief overview
The principle type in this crate is a
[`Searcher`](struct.Searcher.html),
which can be configured and built by a
[`SearcherBuilder`](struct.SearcherBuilder.html).
A `Searcher` is responsible for reading bytes from a source (e.g., a file),
executing a search of those bytes using a `Matcher` (e.g., a regex) and then
reporting the results of that search to a
[`Sink`](trait.Sink.html)
(e.g., stdout). The `Searcher` itself is principally responsible for managing
the consumption of bytes from a source and applying a `Matcher` over those
bytes in an efficient way. The `Searcher` is also responsible for inverting
a search, counting lines, reporting contextual lines, detecting binary data
and even deciding whether or not to use memory maps.
A `Matcher` (which is defined in the
[`grep-matcher`](https://crates.io/crates/grep-matcher)
crate) is a trait for describing the lowest levels of pattern search in a
generic way. The interface itself is very similar to the interface of a regular
expression. For example, the
[`grep-regex`](https://crates.io/crates/grep-regex)
crate provides an implementation of the `Matcher` trait using Rust's
[`regex`](https://crates.io/crates/regex)
crate.
Finally, a `Sink` describes how callers receive search results producer by a
`Searcher`. This includes routines that are called at the beginning and end of
a search, in addition to routines that are called when matching or contextual
lines are found by the `Searcher`. Implementations of `Sink` can be trivially
simple, or extraordinarily complex, such as the
`Standard` printer found in the
[`grep-printer`](https://crates.io/crates/grep-printer)
crate, which effectively implements grep-like output.
This crate also provides convenience `Sink` implementations in the
[`sinks`](sinks/index.html)
sub-module for easy searching with closures.
# Example
This example shows how to execute the searcher and read the search results
using the
[`UTF8`](sinks/struct.UTF8.html)
implementation of `Sink`.
```
extern crate grep_matcher;
extern crate grep_regex;
extern crate grep_searcher;
use std::error::Error;
use grep_matcher::Matcher;
use grep_regex::RegexMatcher;
use grep_searcher::Searcher;
use grep_searcher::sinks::UTF8;
const SHERLOCK: &'static [u8] = b"\
For the Doctor Watsons of this world, as opposed to the Sherlock
Holmeses, success in the province of detective work must always
be, to a very large extent, the result of luck. Sherlock Holmes
can extract a clew from a wisp of straw or a flake of cigar ash;
but Doctor Watson has to have it taken out for him and dusted,
and exhibited clearly, with a label attached.
";
# fn main() { example().unwrap() }
fn example() -> Result<(), Box<Error>> {
let matcher = RegexMatcher::new(r"Doctor \w+")?;
let mut matches: Vec<(u64, String)> = vec![];
Searcher::new().search_slice(&matcher, SHERLOCK, UTF8(|lnum, line| {
// We are guaranteed to find a match, so the unwrap is OK.
eprintln!("LINE: {:?}", line);
let mymatch = matcher.find(line.as_bytes())?.unwrap();
matches.push((lnum, line[mymatch].to_string()));
Ok(true)
}))?;
eprintln!("MATCHES: {:?}", matches);
assert_eq!(matches.len(), 2);
assert_eq!(
matches[0],
(1, "Doctor Watsons".to_string())
);
assert_eq!(
matches[1],
(5, "Doctor Watson".to_string())
);
Ok(())
}
```
See also `examples/search-stdin.rs` from the root of this crate's directory
to see a similar example that accepts a pattern on the command line and
searches stdin.
*/
#![deny(missing_docs)]
extern crate bytecount;
extern crate encoding_rs;
extern crate encoding_rs_io;
extern crate grep_matcher;
#[macro_use]
extern crate log;
extern crate memchr;
extern crate memmap;
#[cfg(test)]
extern crate regex;
pub use lines::{LineIter, LineStep};
pub use searcher::{
BinaryDetection, ConfigError, Encoding, MmapChoice,
Searcher, SearcherBuilder,
};
pub use sink::{
Sink, SinkError,
SinkContext, SinkContextKind, SinkFinish, SinkMatch,
};
pub use sink::sinks;
#[macro_use]
mod macros;
mod line_buffer;
mod lines;
mod searcher;
mod sink;
#[cfg(test)]
mod testutil;

968
grep-searcher/src/line_buffer.rs Normal file
View File

@@ -0,0 +1,968 @@
use std::cmp;
use std::io;
use std::ptr;
use memchr::{memchr, memrchr};
/// The default buffer capacity that we use for the line buffer.
pub(crate) const DEFAULT_BUFFER_CAPACITY: usize = 8 * (1<<10); // 8 KB
/// The behavior of a searcher in the face of long lines and big contexts.
///
/// When searching data incrementally using a fixed size buffer, this controls
/// the amount of *additional* memory to allocate beyond the size of the buffer
/// to accommodate lines (which may include the lines in a context window, when
/// enabled) that do not fit in the buffer.
///
/// The default is to eagerly allocate without a limit.
#[derive(Clone, Copy, Debug)]
pub enum BufferAllocation {
/// Attempt to expand the size of the buffer until either at least the next
/// line fits into memory or until all available memory is exhausted.
///
/// This is the default.
Eager,
/// Limit the amount of additional memory allocated to the given size. If
/// a line is found that requires more memory than is allowed here, then
/// stop reading and return an error.
Error(usize),
}
impl Default for BufferAllocation {
fn default() -> BufferAllocation {
BufferAllocation::Eager
}
}
/// Create a new error to be used when a configured allocation limit has been
/// reached.
pub fn alloc_error(limit: usize) -> io::Error {
let msg = format!("configured allocation limit ({}) exceeded", limit);
io::Error::new(io::ErrorKind::Other, msg)
}
/// The behavior of binary detection in the line buffer.
///
/// Binary detection is the process of _heuristically_ identifying whether a
/// given chunk of data is binary or not, and then taking an action based on
/// the result of that heuristic. The motivation behind detecting binary data
/// is that binary data often indicates data that is undesirable to search
/// using textual patterns. Of course, there are many cases in which this isn't
/// true, which is why binary detection is disabled by default.
#[derive(Clone, Copy, Debug)]
pub enum BinaryDetection {
/// No binary detection is performed. Data reported by the line buffer may
/// contain arbitrary bytes.
None,
/// The given byte is searched in all contents read by the line buffer. If
/// it occurs, then the data is considered binary and the line buffer acts
/// as if it reached EOF. The line buffer guarantees that this byte will
/// never be observable by callers.
Quit(u8),
/// The given byte is searched in all contents read by the line buffer. If
/// it occurs, then it is replaced by the line terminator. The line buffer
/// guarantees that this byte will never be observable by callers.
Convert(u8),
}
impl Default for BinaryDetection {
fn default() -> BinaryDetection {
BinaryDetection::None
}
}
impl BinaryDetection {
/// Returns true if and only if the detection heuristic demands that
/// the line buffer stop read data once binary data is observed.
fn is_quit(&self) -> bool {
match *self {
BinaryDetection::Quit(_) => true,
_ => false,
}
}
}
/// The configuration of a buffer. This contains options that are fixed once
/// a buffer has been constructed.
#[derive(Clone, Copy, Debug)]
struct Config {
/// The number of bytes to attempt to read at a time.
capacity: usize,
/// The line terminator.
lineterm: u8,
/// The behavior for handling long lines.
buffer_alloc: BufferAllocation,
/// When set, the presence of the given byte indicates binary content.
binary: BinaryDetection,
}
impl Default for Config {
fn default() -> Config {
Config {
capacity: DEFAULT_BUFFER_CAPACITY,
lineterm: b'\n',
buffer_alloc: BufferAllocation::default(),
binary: BinaryDetection::default(),
}
}
}
/// A builder for constructing line buffers.
#[derive(Clone, Debug, Default)]
pub struct LineBufferBuilder {
config: Config,
}
impl LineBufferBuilder {
/// Create a new builder for a buffer.
pub fn new() -> LineBufferBuilder {
LineBufferBuilder { config: Config::default() }
}
/// Create a new line buffer from this builder's configuration.
pub fn build(&self) -> LineBuffer {
LineBuffer {
config: self.config,
buf: vec![0; self.config.capacity],
pos: 0,
last_lineterm: 0,
end: 0,
absolute_byte_offset: 0,
binary_byte_offset: None,
}
}
/// Set the default capacity to use for a buffer.
///
/// In general, the capacity of a buffer corresponds to the amount of data
/// to hold in memory, and the size of the reads to make to the underlying
/// reader.
///
/// This is set to a reasonable default and probably shouldn't be changed
/// unless there's a specific reason to do so.
pub fn capacity(&mut self, capacity: usize) -> &mut LineBufferBuilder {
self.config.capacity = capacity;
self
}
/// Set the line terminator for the buffer.
///
/// Every buffer has a line terminator, and this line terminator is used
/// to determine how to roll the buffer forward. For example, when a read
/// to the buffer's underlying reader occurs, the end of the data that is
/// read is likely to correspond to an incomplete line. As a line buffer,
/// callers should not access this data since it is incomplete. The line
/// terminator is how the line buffer determines the part of the read that
/// is incomplete.
///
/// By default, this is set to `b'\n'`.
pub fn line_terminator(&mut self, lineterm: u8) -> &mut LineBufferBuilder {
self.config.lineterm = lineterm;
self
}
/// Set the maximum amount of additional memory to allocate for long lines.
///
/// In order to enable line oriented search, a fundamental requirement is
/// that, at a minimum, each line must be able to fit into memory. This
/// setting controls how big that line is allowed to be. By default, this
/// is set to `BufferAllocation::Eager`, which means a line buffer will
/// attempt to allocate as much memory as possible to fit a line, and will
/// only be limited by available memory.
///
/// Note that this setting only applies to the amount of *additional*
/// memory to allocate, beyond the capacity of the buffer. That means that
/// a value of `0` is sensible, and in particular, will guarantee that a
/// line buffer will never allocate additional memory beyond its initial
/// capacity.
pub fn buffer_alloc(
&mut self,
behavior: BufferAllocation,
) -> &mut LineBufferBuilder {
self.config.buffer_alloc = behavior;
self
}
/// Whether to enable binary detection or not. Depending on the setting,
/// this can either cause the line buffer to report EOF early or it can
/// cause the line buffer to clean the data.
///
/// By default, this is disabled. In general, binary detection should be
/// viewed as an imperfect heuristic.
pub fn binary_detection(
&mut self,
detection: BinaryDetection,
) -> &mut LineBufferBuilder {
self.config.binary = detection;
self
}
}
/// A line buffer reader efficiently reads a line oriented buffer from an
/// arbitrary reader.
#[derive(Debug)]
pub struct LineBufferReader<'b, R> {
rdr: R,
line_buffer: &'b mut LineBuffer,
}
impl<'b, R: io::Read> LineBufferReader<'b, R> {
/// Create a new buffered reader that reads from `rdr` and uses the given
/// `line_buffer` as an intermediate buffer.
///
/// This does not change the binary detection behavior of the given line
/// buffer.
pub fn new(
rdr: R,
line_buffer: &'b mut LineBuffer,
) -> LineBufferReader<'b, R> {
line_buffer.clear();
LineBufferReader { rdr, line_buffer }
}
/// The absolute byte offset which corresponds to the starting offsets
/// of the data returned by `buffer` relative to the beginning of the
/// underlying reader's contents. As such, this offset does not generally
/// correspond to an offset in memory. It is typically used for reporting
/// purposes. It can also be used for counting the number of bytes that
/// have been searched.
pub fn absolute_byte_offset(&self) -> u64 {
self.line_buffer.absolute_byte_offset()
}
/// If binary data was detected, then this returns the absolute byte offset
/// at which binary data was initially found.
pub fn binary_byte_offset(&self) -> Option<u64> {
self.line_buffer.binary_byte_offset()
}
/// Fill the contents of this buffer by discarding the part of the buffer
/// that has been consumed. The free space created by discarding the
/// consumed part of the buffer is then filled with new data from the
/// reader.
///
/// If EOF is reached, then `false` is returned. Otherwise, `true` is
/// returned. (Note that if this line buffer's binary detection is set to
/// `Quit`, then the presence of binary data will cause this buffer to
/// behave as if it had seen EOF at the first occurrence of binary data.)
///
/// This forwards any errors returned by the underlying reader, and will
/// also return an error if the buffer must be expanded past its allocation
/// limit, as governed by the buffer allocation strategy.
pub fn fill(&mut self) -> Result<bool, io::Error> {
self.line_buffer.fill(&mut self.rdr)
}
/// Return the contents of this buffer.
pub fn buffer(&self) -> &[u8] {
self.line_buffer.buffer()
}
/// Consume the number of bytes provided. This must be less than or equal
/// to the number of bytes returned by `buffer`.
pub fn consume(&mut self, amt: usize) {
self.line_buffer.consume(amt);
}
/// Consumes the remainder of the buffer. Subsequent calls to `buffer` are
/// guaranteed to return an empty slice until the buffer is refilled.
///
/// This is a convenience function for `consume(buffer.len())`.
#[cfg(test)]
fn consume_all(&mut self) {
self.line_buffer.consume_all();
}
}
/// A line buffer manages a (typically fixed) buffer for holding lines.
///
/// Callers should create line buffers sparingly and reuse them when possible.
/// Line buffers cannot be used directly, but instead must be used via the
/// LineBufferReader.
#[derive(Clone, Debug)]
pub struct LineBuffer {
/// The configuration of this buffer.
config: Config,
/// The primary buffer with which to hold data.
buf: Vec<u8>,
/// The current position of this buffer. This is always a valid sliceable
/// index into `buf`, and its maximum value is the length of `buf`.
pos: usize,
/// The end position of searchable content in this buffer. This is either
/// set to just after the final line terminator in the buffer, or to just
/// after the end of the last byte emitted by the reader when the reader
/// has been exhausted.
last_lineterm: usize,
/// The end position of the buffer. This is always greater than or equal to
/// last_lineterm. The bytes between last_lineterm and end, if any, always
/// correspond to a partial line.
end: usize,
/// The absolute byte offset corresponding to `pos`. This is most typically
/// not a valid index into addressable memory, but rather, an offset that
/// is relative to all data that passes through a line buffer (since
/// construction or since the last time `clear` was called).
///
/// When the line buffer reaches EOF, this is set to the position just
/// after the last byte read from the underlying reader. That is, it
/// becomes the total count of bytes that have been read.
absolute_byte_offset: u64,
/// If binary data was found, this records the absolute byte offset at
/// which it was first detected.
binary_byte_offset: Option<u64>,
}
impl LineBuffer {
/// Reset this buffer, such that it can be used with a new reader.
fn clear(&mut self) {
self.pos = 0;
self.last_lineterm = 0;
self.end = 0;
self.absolute_byte_offset = 0;
self.binary_byte_offset = None;
}
/// The absolute byte offset which corresponds to the starting offsets
/// of the data returned by `buffer` relative to the beginning of the
/// reader's contents. As such, this offset does not generally correspond
/// to an offset in memory. It is typically used for reporting purposes,
/// particularly in error messages.
///
/// This is reset to `0` when `clear` is called.
fn absolute_byte_offset(&self) -> u64 {
self.absolute_byte_offset
}
/// If binary data was detected, then this returns the absolute byte offset
/// at which binary data was initially found.
fn binary_byte_offset(&self) -> Option<u64> {
self.binary_byte_offset
}
/// Return the contents of this buffer.
fn buffer(&self) -> &[u8] {
&self.buf[self.pos..self.last_lineterm]
}
/// Return the contents of the free space beyond the end of the buffer as
/// a mutable slice.
fn free_buffer(&mut self) -> &mut [u8] {
&mut self.buf[self.end..]
}
/// Consume the number of bytes provided. This must be less than or equal
/// to the number of bytes returned by `buffer`.
fn consume(&mut self, amt: usize) {
assert!(amt <= self.buffer().len());
self.pos += amt;
self.absolute_byte_offset += amt as u64;
}
/// Consumes the remainder of the buffer. Subsequent calls to `buffer` are
/// guaranteed to return an empty slice until the buffer is refilled.
///
/// This is a convenience function for `consume(buffer.len())`.
#[cfg(test)]
fn consume_all(&mut self) {
let amt = self.buffer().len();
self.consume(amt);
}
/// Fill the contents of this buffer by discarding the part of the buffer
/// that has been consumed. The free space created by discarding the
/// consumed part of the buffer is then filled with new data from the given
/// reader.
///
/// Callers should provide the same reader to this line buffer in
/// subsequent calls to fill. A different reader can only be used
/// immediately following a call to `clear`.
///
/// If EOF is reached, then `false` is returned. Otherwise, `true` is
/// returned. (Note that if this line buffer's binary detection is set to
/// `Quit`, then the presence of binary data will cause this buffer to
/// behave as if it had seen EOF.)
///
/// This forwards any errors returned by `rdr`, and will also return an
/// error if the buffer must be expanded past its allocation limit, as
/// governed by the buffer allocation strategy.
fn fill<R: io::Read>(&mut self, mut rdr: R) -> Result<bool, io::Error> {
// If the binary detection heuristic tells us to quit once binary data
// has been observed, then we no longer read new data and reach EOF
// once the current buffer has been consumed.
if self.config.binary.is_quit() && self.binary_byte_offset.is_some() {
return Ok(!self.buffer().is_empty());
}
self.roll();
assert_eq!(self.pos, 0);
loop {
self.ensure_capacity()?;
let readlen = rdr.read(self.free_buffer())?;
if readlen == 0 {
// We're only done reading for good once the caller has
// consumed everything.
self.last_lineterm = self.end;
return Ok(!self.buffer().is_empty());
}
// Get a mutable view into the bytes we've just read. These are
// the bytes that we do binary detection on, and also the bytes we
// search to find the last line terminator. We need a mutable slice
// in the case of binary conversion.
let oldend = self.end;
self.end += readlen;
let newbytes = &mut self.buf[oldend..self.end];
// Binary detection.
match self.config.binary {
BinaryDetection::None => {} // nothing to do
BinaryDetection::Quit(byte) => {
if let Some(i) = memchr(byte, newbytes) {
self.end = oldend + i;
self.last_lineterm = self.end;
self.binary_byte_offset =
Some(self.absolute_byte_offset + self.end as u64);
// If the first byte in our buffer is a binary byte,
// then our buffer is empty and we should report as
// such to the caller.
return Ok(self.pos < self.end);
}
}
BinaryDetection::Convert(byte) => {
if let Some(i) = replace_bytes(
newbytes,
byte,
self.config.lineterm,
) {
// Record only the first binary offset.
if self.binary_byte_offset.is_none() {
self.binary_byte_offset =
Some(self.absolute_byte_offset
+ (oldend + i) as u64);
}
}
}
}
// Update our `last_lineterm` positions if we read one.
if let Some(i) = memrchr(self.config.lineterm, newbytes) {
self.last_lineterm = oldend + i + 1;
return Ok(true);
}
// At this point, if we couldn't find a line terminator, then we
// don't have a complete line. Therefore, we try to read more!
}
}
/// Roll the unconsumed parts of the buffer to the front.
///
/// This operation is idempotent.
///
/// After rolling, `last_lineterm` and `end` point to the same location,
/// and `pos` is always set to `0`.
fn roll(&mut self) {
if self.pos == self.end {
self.pos = 0;
self.last_lineterm = 0;
self.end = 0;
return;
}
assert!(self.pos < self.end && self.end <= self.buf.len());
let roll_len = self.end - self.pos;
unsafe {
// SAFETY: A buffer contains Copy data, so there's no problem
// moving it around. Safety also depends on our indices being
// in bounds, which they should always be, and we enforce with
// an assert above.
//
// It seems like it should be possible to do this in safe code that
// results in the same codegen. I tried the obvious:
//
// for (src, dst) in (self.pos..self.end).zip(0..) {
// self.buf[dst] = self.buf[src];
// }
//
// But the above does not work, and in fact compiles down to a slow
// byte-by-byte loop. I tried a few other minor variations, but
// alas, better minds might prevail.
//
// Overall, this doesn't save us *too* much. It mostly matters when
// the number of bytes we're copying is large, which can happen
// if the searcher is asked to produce a lot of context. We could
// decide this isn't worth it, but it does make an appreciable
// impact at or around the context=30 range on my machine.
//
// We could also use a temporary buffer that compiles down to two
// memcpys and is faster than the byte-at-a-time loop, but it
// complicates our options for limiting memory allocation a bit.
ptr::copy(
self.buf[self.pos..].as_ptr(),
self.buf.as_mut_ptr(),
roll_len,
);
}
self.pos = 0;
self.last_lineterm = roll_len;
self.end = roll_len;
}
/// Ensures that the internal buffer has a non-zero amount of free space
/// in which to read more data. If there is no free space, then more is
/// allocated. If the allocation must exceed the configured limit, then
/// this returns an error.
fn ensure_capacity(&mut self) -> Result<(), io::Error> {
if !self.free_buffer().is_empty() {
return Ok(());
}
// `len` is used for computing the next allocation size. The capacity
// is permitted to start at `0`, so we make sure it's at least `1`.
let len = cmp::max(1, self.buf.len());
let additional = match self.config.buffer_alloc {
BufferAllocation::Eager => len * 2,
BufferAllocation::Error(limit) => {
let used = self.buf.len() - self.config.capacity;
let n = cmp::min(len * 2, limit - used);
if n == 0 {
return Err(alloc_error(self.config.capacity + limit));
}
n
}
};
assert!(additional > 0);
let newlen = self.buf.len() + additional;
self.buf.resize(newlen, 0);
assert!(!self.free_buffer().is_empty());
Ok(())
}
}
/// Replaces `src` with `replacement` in bytes.
fn replace_bytes(bytes: &mut [u8], src: u8, replacement: u8) -> Option<usize> {
if src == replacement {
return None;
}
let mut first_pos = None;
let mut pos = 0;
while let Some(i) = memchr(src, &bytes[pos..]).map(|i| pos + i) {
if first_pos.is_none() {
first_pos = Some(i);
}
bytes[i] = replacement;
pos = i + 1;
while bytes.get(pos) == Some(&src) {
bytes[pos] = replacement;
pos += 1;
}
}
first_pos
}
#[cfg(test)]
mod tests {
use std::str;
use super::*;
const SHERLOCK: &'static str = "\
For the Doctor Watsons of this world, as opposed to the Sherlock
Holmeses, success in the province of detective work must always
be, to a very large extent, the result of luck. Sherlock Holmes
can extract a clew from a wisp of straw or a flake of cigar ash;
but Doctor Watson has to have it taken out for him and dusted,
and exhibited clearly, with a label attached.\
";
fn s(slice: &str) -> String {
slice.to_string()
}
fn btos(slice: &[u8]) -> &str {
str::from_utf8(slice).unwrap()
}
fn replace_str(
slice: &str,
src: u8,
replacement: u8,
) -> (String, Option<usize>) {
let mut dst = slice.to_string().into_bytes();
let result = replace_bytes(&mut dst, src, replacement);
(String::from_utf8(dst).unwrap(), result)
}
#[test]
fn replace() {
assert_eq!(replace_str("abc", b'b', b'z'), (s("azc"), Some(1)));
assert_eq!(replace_str("abb", b'b', b'z'), (s("azz"), Some(1)));
assert_eq!(replace_str("aba", b'a', b'z'), (s("zbz"), Some(0)));
assert_eq!(replace_str("bbb", b'b', b'z'), (s("zzz"), Some(0)));
assert_eq!(replace_str("bac", b'b', b'z'), (s("zac"), Some(0)));
}
#[test]
fn buffer_basics1() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\n");
assert_eq!(rdr.absolute_byte_offset(), 0);
rdr.consume(5);
assert_eq!(rdr.absolute_byte_offset(), 5);
rdr.consume_all();
assert_eq!(rdr.absolute_byte_offset(), 11);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "maggie");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_basics2() {
let bytes = "homer\nlisa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\nmaggie\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_basics3() {
let bytes = "\n";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_basics4() {
let bytes = "\n\n";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "\n\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_empty() {
let bytes = "";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_zero_capacity() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new().capacity(0).build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
while rdr.fill().unwrap() {
rdr.consume_all();
}
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_small_capacity() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new().capacity(1).build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
let mut got = vec![];
while rdr.fill().unwrap() {
got.extend(rdr.buffer());
rdr.consume_all();
}
assert_eq!(bytes, btos(&got));
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_limited_capacity1() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new()
.capacity(1)
.buffer_alloc(BufferAllocation::Error(5))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\n");
rdr.consume_all();
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "lisa\n");
rdr.consume_all();
// This returns an error because while we have just enough room to
// store maggie in the buffer, we *don't* have enough room to read one
// more byte, so we don't know whether we're at EOF or not, and
// therefore must give up.
assert!(rdr.fill().is_err());
// We can mush on though!
assert_eq!(btos(rdr.buffer()), "m");
rdr.consume_all();
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "aggie");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
}
#[test]
fn buffer_limited_capacity2() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new()
.capacity(1)
.buffer_alloc(BufferAllocation::Error(6))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\n");
rdr.consume_all();
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "lisa\n");
rdr.consume_all();
// We have just enough space.
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "maggie");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
}
#[test]
fn buffer_limited_capacity3() {
let bytes = "homer\nlisa\nmaggie";
let mut linebuf = LineBufferBuilder::new()
.capacity(1)
.buffer_alloc(BufferAllocation::Error(0))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.fill().is_err());
assert_eq!(btos(rdr.buffer()), "");
}
#[test]
fn buffer_binary_none() {
let bytes = "homer\nli\x00sa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new().build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nli\x00sa\nmaggie\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), None);
}
#[test]
fn buffer_binary_quit1() {
let bytes = "homer\nli\x00sa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Quit(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nli");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), 8);
assert_eq!(rdr.binary_byte_offset(), Some(8));
}
#[test]
fn buffer_binary_quit2() {
let bytes = "\x00homer\nlisa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Quit(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(!rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "");
assert_eq!(rdr.absolute_byte_offset(), 0);
assert_eq!(rdr.binary_byte_offset(), Some(0));
}
#[test]
fn buffer_binary_quit3() {
let bytes = "homer\nlisa\nmaggie\n\x00";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Quit(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\nmaggie\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64 - 1);
assert_eq!(rdr.binary_byte_offset(), Some(bytes.len() as u64 - 1));
}
#[test]
fn buffer_binary_quit4() {
let bytes = "homer\nlisa\nmaggie\x00\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Quit(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\nmaggie");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64 - 2);
assert_eq!(rdr.binary_byte_offset(), Some(bytes.len() as u64 - 2));
}
#[test]
fn buffer_binary_quit5() {
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Quit(b'u'))
.build();
let mut rdr = LineBufferReader::new(SHERLOCK.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "\
For the Doctor Watsons of this world, as opposed to the Sherlock
Holmeses, s\
");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), 76);
assert_eq!(rdr.binary_byte_offset(), Some(76));
assert_eq!(SHERLOCK.as_bytes()[76], b'u');
}
#[test]
fn buffer_binary_convert1() {
let bytes = "homer\nli\x00sa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Convert(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nli\nsa\nmaggie\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), Some(8));
}
#[test]
fn buffer_binary_convert2() {
let bytes = "\x00homer\nlisa\nmaggie\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Convert(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "\nhomer\nlisa\nmaggie\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), Some(0));
}
#[test]
fn buffer_binary_convert3() {
let bytes = "homer\nlisa\nmaggie\n\x00";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Convert(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\nmaggie\n\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), Some(bytes.len() as u64 - 1));
}
#[test]
fn buffer_binary_convert4() {
let bytes = "homer\nlisa\nmaggie\x00\n";
let mut linebuf = LineBufferBuilder::new()
.binary_detection(BinaryDetection::Convert(b'\x00'))
.build();
let mut rdr = LineBufferReader::new(bytes.as_bytes(), &mut linebuf);
assert!(rdr.buffer().is_empty());
assert!(rdr.fill().unwrap());
assert_eq!(btos(rdr.buffer()), "homer\nlisa\nmaggie\n\n");
rdr.consume_all();
assert!(!rdr.fill().unwrap());
assert_eq!(rdr.absolute_byte_offset(), bytes.len() as u64);
assert_eq!(rdr.binary_byte_offset(), Some(bytes.len() as u64 - 2));
}
}

462
grep-searcher/src/lines.rs Normal file
View File

@@ -0,0 +1,462 @@
/*!
A collection of routines for performing operations on lines.
*/
use bytecount;
use memchr::{memchr, memrchr};
use grep_matcher::{LineTerminator, Match};
/// An iterator over lines in a particular slice of bytes.
///
/// Line terminators are considered part of the line they terminate. All lines
/// yielded by the iterator are guaranteed to be non-empty.
///
/// `'b` refers to the lifetime of the underlying bytes.
#[derive(Debug)]
pub struct LineIter<'b> {
bytes: &'b [u8],
stepper: LineStep,
}
impl<'b> LineIter<'b> {
/// Create a new line iterator that yields lines in the given bytes that
/// are terminated by `line_term`.
pub fn new(line_term: u8, bytes: &'b [u8]) -> LineIter<'b> {
LineIter {
bytes: bytes,
stepper: LineStep::new(line_term, 0, bytes.len()),
}
}
}
impl<'b> Iterator for LineIter<'b> {
type Item = &'b [u8];
fn next(&mut self) -> Option<&'b [u8]> {
self.stepper.next_match(self.bytes).map(|m| &self.bytes[m])
}
}
/// An explicit iterator over lines in a particular slice of bytes.
///
/// This iterator avoids borrowing the bytes themselves, and instead requires
/// callers to explicitly provide the bytes when moving through the iterator.
/// While not idiomatic, this provides a simple way of iterating over lines
/// that doesn't require borrowing the slice itself, which can be convenient.
///
/// Line terminators are considered part of the line they terminate. All lines
/// yielded by the iterator are guaranteed to be non-empty.
#[derive(Debug)]
pub struct LineStep {
line_term: u8,
pos: usize,
end: usize,
}
impl LineStep {
/// Create a new line iterator over the given range of bytes using the
/// given line terminator.
///
/// Callers should provide the actual bytes for each call to `next`. The
/// same slice must be provided to each call.
///
/// This panics if `start` is not less than or equal to `end`.
pub fn new(line_term: u8, start: usize, end: usize) -> LineStep {
LineStep { line_term, pos: start, end: end }
}
/// Return the start and end position of the next line in the given bytes.
///
/// The caller must past exactly the same slice of bytes for each call to
/// `next`.
///
/// The range returned includes the line terminator. Ranges are always
/// non-empty.
pub fn next(&mut self, bytes: &[u8]) -> Option<(usize, usize)> {
self.next_impl(bytes)
}
/// Like next, but returns a `Match` instead of a tuple.
#[inline(always)]
pub(crate) fn next_match(&mut self, bytes: &[u8]) -> Option<Match> {
self.next_impl(bytes).map(|(s, e)| Match::new(s, e))
}
#[inline(always)]
fn next_impl(&mut self, mut bytes: &[u8]) -> Option<(usize, usize)> {
bytes = &bytes[..self.end];
match memchr(self.line_term, &bytes[self.pos..]) {
None => {
if self.pos < bytes.len() {
let m = (self.pos, bytes.len());
assert!(m.0 <= m.1);
self.pos = m.1;
Some(m)
} else {
None
}
}
Some(line_end) => {
let m = (self.pos, self.pos + line_end + 1);
assert!(m.0 <= m.1);
self.pos = m.1;
Some(m)
}
}
}
}
/// Count the number of occurrences of `line_term` in `bytes`.
pub fn count(bytes: &[u8], line_term: u8) -> u64 {
bytecount::count(bytes, line_term) as u64
}
/// Given a line that possibly ends with a terminator, return that line without
/// the terminator.
#[inline(always)]
pub fn without_terminator(bytes: &[u8], line_term: LineTerminator) -> &[u8] {
let line_term = line_term.as_bytes();
let start = bytes.len().saturating_sub(line_term.len());
if bytes.get(start..) == Some(line_term) {
return &bytes[..bytes.len() - line_term.len()];
}
bytes
}
/// Return the start and end offsets of the lines containing the given range
/// of bytes.
///
/// Line terminators are considered part of the line they terminate.
#[inline(always)]
pub fn locate(
bytes: &[u8],
line_term: u8,
range: Match,
) -> Match {
let line_start = memrchr(line_term, &bytes[0..range.start()])
.map_or(0, |i| i + 1);
let line_end =
if range.end() > line_start && bytes[range.end() - 1] == line_term {
range.end()
} else {
memchr(line_term, &bytes[range.end()..])
.map_or(bytes.len(), |i| range.end() + i + 1)
};
Match::new(line_start, line_end)
}
/// Returns the minimal starting offset of the line that occurs `count` lines
/// before the last line in `bytes`.
///
/// Lines are terminated by `line_term`. If `count` is zero, then this returns
/// the starting offset of the last line in `bytes`.
///
/// If `bytes` ends with a line terminator, then the terminator itself is
/// considered part of the last line.
pub fn preceding(bytes: &[u8], line_term: u8, count: usize) -> usize {
preceding_by_pos(bytes, bytes.len(), line_term, count)
}
/// Returns the minimal starting offset of the line that occurs `count` lines
/// before the line containing `pos`. Lines are terminated by `line_term`.
/// If `count` is zero, then this returns the starting offset of the line
/// containing `pos`.
///
/// If `pos` points just past a line terminator, then it is considered part of
/// the line that it terminates. For example, given `bytes = b"abc\nxyz\n"`
/// and `pos = 7`, `preceding(bytes, pos, b'\n', 0)` returns `4` (as does `pos
/// = 8`) and `preceding(bytes, pos, `b'\n', 1)` returns `0`.
fn preceding_by_pos(
bytes: &[u8],
mut pos: usize,
line_term: u8,
mut count: usize,
) -> usize {
if pos == 0 {
return 0;
} else if bytes[pos - 1] == line_term {
pos -= 1;
}
loop {
match memrchr(line_term, &bytes[..pos]) {
None => {
return 0;
}
Some(i) => {
if count == 0 {
return i + 1;
} else if i == 0 {
return 0;
}
count -= 1;
pos = i;
}
}
}
}
#[cfg(test)]
mod tests {
use std::ops::Range;
use std::str;
use grep_matcher::Match;
use super::*;
const SHERLOCK: &'static str = "\
For the Doctor Watsons of this world, as opposed to the Sherlock
Holmeses, success in the province of detective work must always
be, to a very large extent, the result of luck. Sherlock Holmes
can extract a clew from a wisp of straw or a flake of cigar ash;
but Doctor Watson has to have it taken out for him and dusted,
and exhibited clearly, with a label attached.\
";
fn m(start: usize, end: usize) -> Match {
Match::new(start, end)
}
fn lines(text: &str) -> Vec<&str> {
let mut results = vec![];
let mut it = LineStep::new(b'\n', 0, text.len());
while let Some(m) = it.next_match(text.as_bytes()) {
results.push(&text[m]);
}
results
}
fn line_ranges(text: &str) -> Vec<Range<usize>> {
let mut results = vec![];
let mut it = LineStep::new(b'\n', 0, text.len());
while let Some(m) = it.next_match(text.as_bytes()) {
results.push(m.start()..m.end());
}
results
}
fn prev(text: &str, pos: usize, count: usize) -> usize {
preceding_by_pos(text.as_bytes(), pos, b'\n', count)
}
fn loc(text: &str, start: usize, end: usize) -> Match {
locate(text.as_bytes(), b'\n', Match::new(start, end))
}
#[test]
fn line_count() {
assert_eq!(0, count(b"", b'\n'));
assert_eq!(1, count(b"\n", b'\n'));
assert_eq!(2, count(b"\n\n", b'\n'));
assert_eq!(2, count(b"a\nb\nc", b'\n'));
}
#[test]
fn line_locate() {
let t = SHERLOCK;
let lines = line_ranges(t);
assert_eq!(
loc(t, lines[0].start, lines[0].end),
m(lines[0].start, lines[0].end));
assert_eq!(
loc(t, lines[0].start + 1, lines[0].end),
m(lines[0].start, lines[0].end));
assert_eq!(
loc(t, lines[0].end - 1, lines[0].end),
m(lines[0].start, lines[0].end));
assert_eq!(
loc(t, lines[0].end, lines[0].end),
m(lines[1].start, lines[1].end));
assert_eq!(
loc(t, lines[5].start, lines[5].end),
m(lines[5].start, lines[5].end));
assert_eq!(
loc(t, lines[5].start + 1, lines[5].end),
m(lines[5].start, lines[5].end));
assert_eq!(
loc(t, lines[5].end - 1, lines[5].end),
m(lines[5].start, lines[5].end));
assert_eq!(
loc(t, lines[5].end, lines[5].end),
m(lines[5].start, lines[5].end));
}
#[test]
fn line_locate_weird() {
assert_eq!(loc("", 0, 0), m(0, 0));
assert_eq!(loc("\n", 0, 1), m(0, 1));
assert_eq!(loc("\n", 1, 1), m(1, 1));
assert_eq!(loc("\n\n", 0, 0), m(0, 1));
assert_eq!(loc("\n\n", 0, 1), m(0, 1));
assert_eq!(loc("\n\n", 1, 1), m(1, 2));
assert_eq!(loc("\n\n", 1, 2), m(1, 2));
assert_eq!(loc("\n\n", 2, 2), m(2, 2));
assert_eq!(loc("a\nb\nc", 0, 1), m(0, 2));
assert_eq!(loc("a\nb\nc", 1, 2), m(0, 2));
assert_eq!(loc("a\nb\nc", 2, 3), m(2, 4));
assert_eq!(loc("a\nb\nc", 3, 4), m(2, 4));
assert_eq!(loc("a\nb\nc", 4, 5), m(4, 5));
assert_eq!(loc("a\nb\nc", 5, 5), m(4, 5));
}
#[test]
fn line_iter() {
assert_eq!(lines("abc"), vec!["abc"]);
assert_eq!(lines("abc\n"), vec!["abc\n"]);
assert_eq!(lines("abc\nxyz"), vec!["abc\n", "xyz"]);
assert_eq!(lines("abc\nxyz\n"), vec!["abc\n", "xyz\n"]);
assert_eq!(lines("abc\n\n"), vec!["abc\n", "\n"]);
assert_eq!(lines("abc\n\n\n"), vec!["abc\n", "\n", "\n"]);
assert_eq!(lines("abc\n\nxyz"), vec!["abc\n", "\n", "xyz"]);
assert_eq!(lines("abc\n\nxyz\n"), vec!["abc\n", "\n", "xyz\n"]);
assert_eq!(lines("abc\nxyz\n\n"), vec!["abc\n", "xyz\n", "\n"]);
assert_eq!(lines("\n"), vec!["\n"]);
assert_eq!(lines(""), Vec::<&str>::new());
}
#[test]
fn line_iter_empty() {
let mut it = LineStep::new(b'\n', 0, 0);
assert_eq!(it.next(b"abc"), None);
}
#[test]
fn preceding_lines_doc() {
// These are the examples mentions in the documentation of `preceding`.
let bytes = b"abc\nxyz\n";
assert_eq!(4, preceding_by_pos(bytes, 7, b'\n', 0));
assert_eq!(4, preceding_by_pos(bytes, 8, b'\n', 0));
assert_eq!(0, preceding_by_pos(bytes, 7, b'\n', 1));
assert_eq!(0, preceding_by_pos(bytes, 8, b'\n', 1));
}
#[test]
fn preceding_lines_sherlock() {
let t = SHERLOCK;
let lines = line_ranges(t);
// The following tests check the count == 0 case, i.e., finding the
// beginning of the line containing the given position.
assert_eq!(0, prev(t, 0, 0));
assert_eq!(0, prev(t, 1, 0));
// The line terminator is addressed by `end-1` and terminates the line
// it is part of.
assert_eq!(0, prev(t, lines[0].end - 1, 0));
assert_eq!(lines[0].start, prev(t, lines[0].end, 0));
// The end position of line addresses the byte immediately following a
// line terminator, which puts it on the following line.
assert_eq!(lines[1].start, prev(t, lines[0].end + 1, 0));
// Now tests for count > 0.
assert_eq!(0, prev(t, 0, 1));
assert_eq!(0, prev(t, 0, 2));
assert_eq!(0, prev(t, 1, 1));
assert_eq!(0, prev(t, 1, 2));
assert_eq!(0, prev(t, lines[0].end - 1, 1));
assert_eq!(0, prev(t, lines[0].end - 1, 2));
assert_eq!(0, prev(t, lines[0].end, 1));
assert_eq!(0, prev(t, lines[0].end, 2));
assert_eq!(lines[3].start, prev(t, lines[4].end - 1, 1));
assert_eq!(lines[3].start, prev(t, lines[4].end, 1));
assert_eq!(lines[4].start, prev(t, lines[4].end + 1, 1));
// The last line has no line terminator.
assert_eq!(lines[5].start, prev(t, lines[5].end, 0));
assert_eq!(lines[5].start, prev(t, lines[5].end - 1, 0));
assert_eq!(lines[4].start, prev(t, lines[5].end, 1));
assert_eq!(lines[0].start, prev(t, lines[5].end, 5));
}
#[test]
fn preceding_lines_short() {
let t = "a\nb\nc\nd\ne\nf\n";
let lines = line_ranges(t);
assert_eq!(12, t.len());
assert_eq!(lines[5].start, prev(t, lines[5].end, 0));
assert_eq!(lines[4].start, prev(t, lines[5].end, 1));
assert_eq!(lines[3].start, prev(t, lines[5].end, 2));
assert_eq!(lines[2].start, prev(t, lines[5].end, 3));
assert_eq!(lines[1].start, prev(t, lines[5].end, 4));
assert_eq!(lines[0].start, prev(t, lines[5].end, 5));
assert_eq!(lines[0].start, prev(t, lines[5].end, 6));
assert_eq!(lines[5].start, prev(t, lines[5].end - 1, 0));
assert_eq!(lines[4].start, prev(t, lines[5].end - 1, 1));
assert_eq!(lines[3].start, prev(t, lines[5].end - 1, 2));
assert_eq!(lines[2].start, prev(t, lines[5].end - 1, 3));
assert_eq!(lines[1].start, prev(t, lines[5].end - 1, 4));
assert_eq!(lines[0].start, prev(t, lines[5].end - 1, 5));
assert_eq!(lines[0].start, prev(t, lines[5].end - 1, 6));
assert_eq!(lines[4].start, prev(t, lines[5].start, 0));
assert_eq!(lines[3].start, prev(t, lines[5].start, 1));
assert_eq!(lines[2].start, prev(t, lines[5].start, 2));
assert_eq!(lines[1].start, prev(t, lines[5].start, 3));
assert_eq!(lines[0].start, prev(t, lines[5].start, 4));
assert_eq!(lines[0].start, prev(t, lines[5].start, 5));
assert_eq!(lines[3].start, prev(t, lines[4].end - 1, 1));
assert_eq!(lines[2].start, prev(t, lines[4].start, 1));
assert_eq!(lines[2].start, prev(t, lines[3].end - 1, 1));
assert_eq!(lines[1].start, prev(t, lines[3].start, 1));
assert_eq!(lines[1].start, prev(t, lines[2].end - 1, 1));
assert_eq!(lines[0].start, prev(t, lines[2].start, 1));
assert_eq!(lines[0].start, prev(t, lines[1].end - 1, 1));
assert_eq!(lines[0].start, prev(t, lines[1].start, 1));
assert_eq!(lines[0].start, prev(t, lines[0].end - 1, 1));
assert_eq!(lines[0].start, prev(t, lines[0].start, 1));
}
#[test]
fn preceding_lines_empty1() {
let t = "\n\n\nd\ne\nf\n";
let lines = line_ranges(t);
assert_eq!(9, t.len());
assert_eq!(lines[0].start, prev(t, lines[0].end, 0));
assert_eq!(lines[0].start, prev(t, lines[0].end, 1));
assert_eq!(lines[1].start, prev(t, lines[1].end, 0));
assert_eq!(lines[0].start, prev(t, lines[1].end, 1));
assert_eq!(lines[5].start, prev(t, lines[5].end, 0));
assert_eq!(lines[4].start, prev(t, lines[5].end, 1));
assert_eq!(lines[3].start, prev(t, lines[5].end, 2));
assert_eq!(lines[2].start, prev(t, lines[5].end, 3));
assert_eq!(lines[1].start, prev(t, lines[5].end, 4));
assert_eq!(lines[0].start, prev(t, lines[5].end, 5));
assert_eq!(lines[0].start, prev(t, lines[5].end, 6));
}
#[test]
fn preceding_lines_empty2() {
let t = "a\n\n\nd\ne\nf\n";
let lines = line_ranges(t);
assert_eq!(10, t.len());
assert_eq!(lines[0].start, prev(t, lines[0].end, 0));
assert_eq!(lines[0].start, prev(t, lines[0].end, 1));
assert_eq!(lines[1].start, prev(t, lines[1].end, 0));
assert_eq!(lines[0].start, prev(t, lines[1].end, 1));
assert_eq!(lines[5].start, prev(t, lines[5].end, 0));
assert_eq!(lines[4].start, prev(t, lines[5].end, 1));
assert_eq!(lines[3].start, prev(t, lines[5].end, 2));
assert_eq!(lines[2].start, prev(t, lines[5].end, 3));
assert_eq!(lines[1].start, prev(t, lines[5].end, 4));
assert_eq!(lines[0].start, prev(t, lines[5].end, 5));
assert_eq!(lines[0].start, prev(t, lines[5].end, 6));
}
}

24
grep-searcher/src/macros.rs Normal file
View File

@@ -0,0 +1,24 @@
#[cfg(test)]
#[macro_export]
macro_rules! assert_eq_printed {
($expected:expr, $got:expr, $($tt:tt)*) => {
let expected = &*$expected;
let got = &*$got;
let label = format!($($tt)*);
if expected != got {
panic!("
printed outputs differ! (label: {})
expected:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
{}
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
got:
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
{}
~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
", label, expected, got);
}
}
}

592
grep-searcher/src/searcher/core.rs Normal file
View File

@@ -0,0 +1,592 @@
use std::cmp;
use memchr::memchr;
use grep_matcher::{LineMatchKind, Matcher};
use lines::{self, LineStep};
use line_buffer::BinaryDetection;
use searcher::{Config, Range, Searcher};
use sink::{
Sink, SinkError,
SinkFinish, SinkContext, SinkContextKind, SinkMatch,
};
#[derive(Debug)]
pub struct Core<'s, M: 's, S> {
config: &'s Config,
matcher: M,
searcher: &'s Searcher,
sink: S,
binary: bool,
pos: usize,
absolute_byte_offset: u64,
binary_byte_offset: Option<usize>,
line_number: Option<u64>,
last_line_counted: usize,
last_line_visited: usize,
after_context_left: usize,
has_sunk: bool,
}
impl<'s, M: Matcher, S: Sink> Core<'s, M, S> {
pub fn new(
searcher: &'s Searcher,
matcher: M,
sink: S,
binary: bool,
) -> Core<'s, M, S> {
let line_number =
if searcher.config.line_number {
Some(1)
} else {
None
};
let core = Core {
config: &searcher.config,
matcher: matcher,
searcher: searcher,
sink: sink,
binary: binary,
pos: 0,
absolute_byte_offset: 0,
binary_byte_offset: None,
line_number: line_number,
last_line_counted: 0,
last_line_visited: 0,
after_context_left: 0,
has_sunk: false,
};
if !core.searcher.multi_line_with_matcher(&core.matcher) {
if core.is_line_by_line_fast() {
trace!("searcher core: will use fast line searcher");
} else {
trace!("searcher core: will use slow line searcher");
}
}
core
}
pub fn pos(&self) -> usize {
self.pos
}
pub fn set_pos(&mut self, pos: usize) {
self.pos = pos;
}
pub fn binary_byte_offset(&self) -> Option<u64> {
self.binary_byte_offset.map(|offset| offset as u64)
}
pub fn matcher(&self) -> &M {
&self.matcher
}
pub fn matched(
&mut self,
buf: &[u8],
range: &Range,
) -> Result<bool, S::Error> {
self.sink_matched(buf, range)
}
pub fn begin(&mut self) -> Result<bool, S::Error> {
self.sink.begin(&self.searcher)
}
pub fn finish(
&mut self,
byte_count: u64,
binary_byte_offset: Option<u64>,
) -> Result<(), S::Error> {
self.sink.finish(
&self.searcher,
&SinkFinish {
byte_count,
binary_byte_offset,
})
}
pub fn match_by_line(&mut self, buf: &[u8]) -> Result<bool, S::Error> {
if self.is_line_by_line_fast() {
self.match_by_line_fast(buf)
} else {
self.match_by_line_slow(buf)
}
}
pub fn roll(&mut self, buf: &[u8]) -> usize {
let consumed =
if self.config.max_context() == 0 {
buf.len()
} else {
// It might seem like all we need to care about here is just
// the "before context," but in order to sink the context
// separator (when before_context==0 and after_context>0), we
// need to know something about the position of the previous
// line visited, even if we're at the beginning of the buffer.
let context_start = lines::preceding(
buf,
self.config.line_term.as_byte(),
self.config.max_context(),
);
let consumed = cmp::max(context_start, self.last_line_visited);
consumed
};
self.count_lines(buf, consumed);
self.absolute_byte_offset += consumed as u64;
self.last_line_counted = 0;
self.last_line_visited = 0;
self.set_pos(buf.len() - consumed);
consumed
}
pub fn detect_binary(&mut self, buf: &[u8], range: &Range) -> bool {
if self.binary_byte_offset.is_some() {
return true;
}
let binary_byte = match self.config.binary.0 {
BinaryDetection::Quit(b) => b,
_ => return false,
};
if let Some(i) = memchr(binary_byte, &buf[*range]) {
self.binary_byte_offset = Some(range.start() + i);
true
} else {
false
}
}
pub fn before_context_by_line(
&mut self,
buf: &[u8],
upto: usize,
) -> Result<bool, S::Error> {
if self.config.before_context == 0 {
return Ok(true);
}
let range = Range::new(self.last_line_visited, upto);
if range.is_empty() {
return Ok(true);
}
let before_context_start = range.start() + lines::preceding(
&buf[range],
self.config.line_term.as_byte(),
self.config.before_context - 1,
);
let range = Range::new(before_context_start, range.end());
let mut stepper = LineStep::new(
self.config.line_term.as_byte(),
range.start(),
range.end(),
);
while let Some(line) = stepper.next_match(buf) {
if !self.sink_break_context(line.start())? {
return Ok(false);
}
if !self.sink_before_context(buf, &line)? {
return Ok(false);
}
}
Ok(true)
}
pub fn after_context_by_line(
&mut self,
buf: &[u8],
upto: usize,
) -> Result<bool, S::Error> {
if self.after_context_left == 0 {
return Ok(true);
}
let range = Range::new(self.last_line_visited, upto);
let mut stepper = LineStep::new(
self.config.line_term.as_byte(),
range.start(),
range.end(),
);
while let Some(line) = stepper.next_match(buf) {
if !self.sink_after_context(buf, &line)? {
return Ok(false);
}
if self.after_context_left == 0 {
break;
}
}
Ok(true)
}
pub fn other_context_by_line(
&mut self,
buf: &[u8],
upto: usize,
) -> Result<bool, S::Error> {
let range = Range::new(self.last_line_visited, upto);
let mut stepper = LineStep::new(
self.config.line_term.as_byte(),
range.start(),
range.end(),
);
while let Some(line) = stepper.next_match(buf) {
if !self.sink_other_context(buf, &line)? {
return Ok(false);
}
}
Ok(true)
}
fn match_by_line_slow(&mut self, buf: &[u8]) -> Result<bool, S::Error> {
debug_assert!(!self.searcher.multi_line_with_matcher(&self.matcher));
let range = Range::new(self.pos(), buf.len());
let mut stepper = LineStep::new(
self.config.line_term.as_byte(),
range.start(),
range.end(),
);
while let Some(line) = stepper.next_match(buf) {
let matched = {
// Stripping the line terminator is necessary to prevent some
// classes of regexes from matching the empty position *after*
// the end of the line. For example, `(?m)^$` will match at
// position (2, 2) in the string `a\n`.
let slice = lines::without_terminator(
&buf[line],
self.config.line_term,
);
match self.matcher.shortest_match(slice) {
Err(err) => return Err(S::Error::error_message(err)),
Ok(result) => result.is_some(),
}
};
self.set_pos(line.end());
if matched != self.config.invert_match {
if !self.before_context_by_line(buf, line.start())? {
return Ok(false);
}
if !self.sink_matched(buf, &line)? {
return Ok(false);
}
} else if self.after_context_left >= 1 {
if !self.sink_after_context(buf, &line)? {
return Ok(false);
}
} else if self.config.passthru {
if !self.sink_other_context(buf, &line)? {
return Ok(false);
}
}
}
Ok(true)
}
fn match_by_line_fast(&mut self, buf: &[u8]) -> Result<bool, S::Error> {
debug_assert!(!self.config.passthru);
while !buf[self.pos()..].is_empty() {
if self.config.invert_match {
if !self.match_by_line_fast_invert(buf)? {
return Ok(false);
}
} else if let Some(line) = self.find_by_line_fast(buf)? {
if self.config.max_context() > 0 {
if !self.after_context_by_line(buf, line.start())? {
return Ok(false);
}
if !self.before_context_by_line(buf, line.start())? {
return Ok(false);
}
}
self.set_pos(line.end());
if !self.sink_matched(buf, &line)? {
return Ok(false);
}
} else {
break;
}
}
if !self.after_context_by_line(buf, buf.len())? {
return Ok(false);
}
self.set_pos(buf.len());
Ok(true)
}
#[inline(always)]
fn match_by_line_fast_invert(
&mut self,
buf: &[u8],
) -> Result<bool, S::Error> {
assert!(self.config.invert_match);
let invert_match = match self.find_by_line_fast(buf)? {
None => {
let range = Range::new(self.pos(), buf.len());
self.set_pos(range.end());
range
}
Some(line) => {
let range = Range::new(self.pos(), line.start());
self.set_pos(line.end());
range
}
};
if invert_match.is_empty() {
return Ok(true);
}
if !self.after_context_by_line(buf, invert_match.start())? {
return Ok(false);
}
if !self.before_context_by_line(buf, invert_match.start())? {
return Ok(false);
}
let mut stepper = LineStep::new(
self.config.line_term.as_byte(),
invert_match.start(),
invert_match.end(),
);
while let Some(line) = stepper.next_match(buf) {
if !self.sink_matched(buf, &line)? {
return Ok(false);
}
}
Ok(true)
}
#[inline(always)]
fn find_by_line_fast(
&self,
buf: &[u8],
) -> Result<Option<Range>, S::Error> {
debug_assert!(!self.searcher.multi_line_with_matcher(&self.matcher));
debug_assert!(self.is_line_by_line_fast());
let mut pos = self.pos();
while !buf[pos..].is_empty() {
match self.matcher.find_candidate_line(&buf[pos..]) {
Err(err) => return Err(S::Error::error_message(err)),
Ok(None) => return Ok(None),
Ok(Some(LineMatchKind::Confirmed(i))) => {
let line = lines::locate(
buf,
self.config.line_term.as_byte(),
Range::zero(i).offset(pos),
);
// If we matched beyond the end of the buffer, then we
// don't report this as a match.
if line.start() == buf.len() {
pos = buf.len();
continue;
}
return Ok(Some(line));
}
Ok(Some(LineMatchKind::Candidate(i))) => {
let line = lines::locate(
buf,
self.config.line_term.as_byte(),
Range::zero(i).offset(pos),
);
// We need to strip the line terminator here to match the
// semantics of line-by-line searching. Namely, regexes
// like `(?m)^$` can match at the final position beyond a
// line terminator, which is non-sensical in line oriented
// matching.
let slice = lines::without_terminator(
&buf[line],
self.config.line_term,
);
match self.matcher.is_match(slice) {
Err(err) => return Err(S::Error::error_message(err)),
Ok(true) => return Ok(Some(line)),
Ok(false) => {
pos = line.end();
continue;
}
}
}
}
}
Ok(None)
}
#[inline(always)]
fn sink_matched(
&mut self,
buf: &[u8],
range: &Range,
) -> Result<bool, S::Error> {
if self.binary && self.detect_binary(buf, range) {
return Ok(false);
}
if !self.sink_break_context(range.start())? {
return Ok(false);
}
self.count_lines(buf, range.start());
let offset = self.absolute_byte_offset + range.start() as u64;
let linebuf =
if self.config.line_term.is_crlf() {
// Normally, a line terminator is never part of a match, but
// if the line terminator is CRLF, then it's possible for `\r`
// to end up in the match, which we generally don't want. So
// we strip it here.
lines::without_terminator(&buf[*range], self.config.line_term)
} else {
&buf[*range]
};
let keepgoing = self.sink.matched(
&self.searcher,
&SinkMatch {
line_term: self.config.line_term,
bytes: linebuf,
absolute_byte_offset: offset,
line_number: self.line_number,
},
)?;
if !keepgoing {
return Ok(false);
}
self.last_line_visited = range.end();
self.after_context_left = self.config.after_context;
self.has_sunk = true;
Ok(true)
}
fn sink_before_context(
&mut self,
buf: &[u8],
range: &Range,
) -> Result<bool, S::Error> {
if self.binary && self.detect_binary(buf, range) {
return Ok(false);
}
self.count_lines(buf, range.start());
let offset = self.absolute_byte_offset + range.start() as u64;
let keepgoing = self.sink.context(
&self.searcher,
&SinkContext {
line_term: self.config.line_term,
bytes: &buf[*range],
kind: SinkContextKind::Before,
absolute_byte_offset: offset,
line_number: self.line_number,
},
)?;
if !keepgoing {
return Ok(false);
}
self.last_line_visited = range.end();
self.has_sunk = true;
Ok(true)
}
fn sink_after_context(
&mut self,
buf: &[u8],
range: &Range,
) -> Result<bool, S::Error> {
assert!(self.after_context_left >= 1);
if self.binary && self.detect_binary(buf, range) {
return Ok(false);
}
self.count_lines(buf, range.start());
let offset = self.absolute_byte_offset + range.start() as u64;
let keepgoing = self.sink.context(
&self.searcher,
&SinkContext {
line_term: self.config.line_term,
bytes: &buf[*range],
kind: SinkContextKind::After,
absolute_byte_offset: offset,
line_number: self.line_number,
},
)?;
if !keepgoing {
return Ok(false);
}
self.last_line_visited = range.end();
self.after_context_left -= 1;
self.has_sunk = true;
Ok(true)
}
fn sink_other_context(
&mut self,
buf: &[u8],
range: &Range,
) -> Result<bool, S::Error> {
if self.binary && self.detect_binary(buf, range) {
return Ok(false);
}
self.count_lines(buf, range.start());
let offset = self.absolute_byte_offset + range.start() as u64;
let keepgoing = self.sink.context(
&self.searcher,
&SinkContext {
line_term: self.config.line_term,
bytes: &buf[*range],
kind: SinkContextKind::Other,
absolute_byte_offset: offset,
line_number: self.line_number,
},
)?;
if !keepgoing {
return Ok(false);
}
self.last_line_visited = range.end();
self.has_sunk = true;
Ok(true)
}
fn sink_break_context(
&mut self,
start_of_line: usize,
) -> Result<bool, S::Error> {
let is_gap = self.last_line_visited < start_of_line;
let any_context =
self.config.before_context > 0
|| self.config.after_context > 0;
if !any_context || !self.has_sunk || !is_gap {
Ok(true)
} else {
self.sink.context_break(&self.searcher)
}
}
fn count_lines(&mut self, buf: &[u8], upto: usize) {
if let Some(ref mut line_number) = self.line_number {
if self.last_line_counted >= upto {
return;
}
let slice = &buf[self.last_line_counted..upto];
let count = lines::count(slice, self.config.line_term.as_byte());
*line_number += count;
self.last_line_counted = upto;
}
}
fn is_line_by_line_fast(&self) -> bool {
debug_assert!(!self.searcher.multi_line_with_matcher(&self.matcher));
if self.config.passthru {
return false;
}
if let Some(line_term) = self.matcher.line_terminator() {
if line_term == self.config.line_term {
return true;
}
}
if let Some(non_matching) = self.matcher.non_matching_bytes() {
// If the line terminator is CRLF, we don't actually need to care
// whether the regex can match `\r` or not. Namely, a `\r` is
// neither necessary nor sufficient to terminate a line. A `\n` is
// always required.
if non_matching.contains(self.config.line_term.as_byte()) {
return true;
}
}
false
}
}

1506
grep-searcher/src/searcher/glue.rs Normal file
View File

File diff suppressed because it is too large Load Diff

106
grep-searcher/src/searcher/mmap.rs Normal file
View File

@@ -0,0 +1,106 @@
use std::fs::File;
use std::path::Path;
use memmap::Mmap;
/// Controls the strategy used for determining when to use memory maps.
///
/// If a searcher is called in circumstances where it is possible to use memory
/// maps, and memory maps are enabled, then it will attempt to do so if it
/// believes it will make the search faster.
///
/// By default, memory maps are disabled.
#[derive(Clone, Debug)]
pub struct MmapChoice(MmapChoiceImpl);
#[derive(Clone, Debug)]
enum MmapChoiceImpl {
Auto,
Never,
}
impl Default for MmapChoice {
fn default() -> MmapChoice {
MmapChoice(MmapChoiceImpl::Never)
}
}
impl MmapChoice {
/// Use memory maps when they are believed to be advantageous.
///
/// The heuristics used to determine whether to use a memory map or not
/// may depend on many things, including but not limited to, file size
/// and platform.
///
/// If memory maps are unavailable or cannot be used for a specific input,
/// then normal OS read calls are used instead.
///
/// # Safety
///
/// This constructor is not safe because there is no obvious way to
/// encapsulate the safety of file backed memory maps on all platforms
/// without simultaneously negating some or all of their benefits.
///
/// The specific contract the caller is required to uphold isn't precise,
/// but it basically amounts to something like, "the caller guarantees that
/// the underlying file won't be mutated." This, of course, isn't feasible
/// in many environments. However, command line tools may still decide to
/// take the risk of, say, a `SIGBUS` occurring while attempting to read a
/// memory map.
pub unsafe fn auto() -> MmapChoice {
MmapChoice(MmapChoiceImpl::Auto)
}
/// Never use memory maps, no matter what. This is the default.
pub fn never() -> MmapChoice {
MmapChoice(MmapChoiceImpl::Never)
}
/// Return a memory map if memory maps are enabled and if creating a
/// memory from the given file succeeded and if memory maps are believed
/// to be advantageous for performance.
///
/// If this does attempt to open a memory map and it fails, then `None`
/// is returned and the corresponding error (along with the file path, if
/// present) is logged at the debug level.
pub(crate) fn open(
&self,
file: &File,
path: Option<&Path>,
) -> Option<Mmap> {
if !self.is_enabled() {
return None;
}
if cfg!(target_os = "macos") {
// I guess memory maps on macOS aren't great. Should re-evaluate.
return None;
}
// SAFETY: This is acceptable because the only way `MmapChoiceImpl` can
// be `Auto` is if the caller invoked the `auto` constructor, which
// is itself not safe. Thus, this is a propagation of the caller's
// assertion that using memory maps is safe.
match unsafe { Mmap::map(file) } {
Ok(mmap) => Some(mmap),
Err(err) => {
if let Some(path) = path {
debug!(
"{}: failed to open memory map: {}",
path.display(),
err
);
} else {
debug!("failed to open memory map: {}", err);
}
None
}
}
}
/// Whether this strategy may employ memory maps or not.
pub(crate) fn is_enabled(&self) -> bool {
match self.0 {
MmapChoiceImpl::Auto => true,
MmapChoiceImpl::Never => false,
}
}
}

956
grep-searcher/src/searcher/mod.rs Normal file
View File

@@ -0,0 +1,956 @@
use std::cell::RefCell;
use std::cmp;
use std::fmt;
use std::fs::File;
use std::io::{self, Read};
use std::path::Path;
use encoding_rs;
use encoding_rs_io::DecodeReaderBytesBuilder;
use grep_matcher::{LineTerminator, Match, Matcher};
use line_buffer::{
self, BufferAllocation, LineBuffer, LineBufferBuilder, LineBufferReader,
DEFAULT_BUFFER_CAPACITY, alloc_error,
};
use searcher::glue::{ReadByLine, SliceByLine, MultiLine};
use sink::{Sink, SinkError};
pub use self::mmap::MmapChoice;
mod core;
mod glue;
mod mmap;
/// We use this type alias since we want the ergonomics of a matcher's `Match`
/// type, but in practice, we use it for arbitrary ranges, so give it a more
/// accurate name. This is only used in the searcher's internals.
type Range = Match;
/// The behavior of binary detection while searching.
///
/// Binary detection is the process of _heuristically_ identifying whether a
/// given chunk of data is binary or not, and then taking an action based on
/// the result of that heuristic. The motivation behind detecting binary data
/// is that binary data often indicates data that is undesirable to search
/// using textual patterns. Of course, there are many cases in which this isn't
/// true, which is why binary detection is disabled by default.
///
/// Unfortunately, binary detection works differently depending on the type of
/// search being executed:
///
/// 1. When performing a search using a fixed size buffer, binary detection is
/// applied to the buffer's contents as it is filled. Binary detection must
/// be applied to the buffer directly because binary files may not contain
/// line terminators, which could result in exorbitant memory usage.
/// 2. When performing a search using memory maps or by reading data off the
/// heap, then binary detection is only guaranteed to be applied to the
/// parts corresponding to a match. When `Quit` is enabled, then the first
/// few KB of the data are searched for binary data.
#[derive(Clone, Debug, Default)]
pub struct BinaryDetection(line_buffer::BinaryDetection);
impl BinaryDetection {
/// No binary detection is performed. Data reported by the searcher may
/// contain arbitrary bytes.
///
/// This is the default.
pub fn none() -> BinaryDetection {
BinaryDetection(line_buffer::BinaryDetection::None)
}
/// Binary detection is performed by looking for the given byte.
///
/// When searching is performed using a fixed size buffer, then the
/// contents of that buffer are always searched for the presence of this
/// byte. If it is found, then the underlying data is considered binary
/// and the search stops as if it reached EOF.
///
/// When searching is performed with the entire contents mapped into
/// memory, then binary detection is more conservative. Namely, only a
/// fixed sized region at the beginning of the contents are detected for
/// binary data. As a compromise, any subsequent matching (or context)
/// lines are also searched for binary data. If binary data is detected at
/// any point, then the search stops as if it reached EOF.
pub fn quit(binary_byte: u8) -> BinaryDetection {
BinaryDetection(line_buffer::BinaryDetection::Quit(binary_byte))
}
// TODO(burntsushi): Figure out how to make binary conversion work. This
// permits implementing GNU grep's default behavior, which is to zap NUL
// bytes but still execute a search (if a match is detected, then GNU grep
// stops and reports that a match was found but doesn't print the matching
// line itself).
//
// This behavior is pretty simple to implement using the line buffer (and
// in fact, it is already implemented and tested), since there's a fixed
// size buffer that we can easily write to. The issue arises when searching
// a `&[u8]` (whether on the heap or via a memory map), since this isn't
// something we can easily write to.
/// The given byte is searched in all contents read by the line buffer. If
/// it occurs, then it is replaced by the line terminator. The line buffer
/// guarantees that this byte will never be observable by callers.
#[allow(dead_code)]
fn convert(binary_byte: u8) -> BinaryDetection {
BinaryDetection(line_buffer::BinaryDetection::Convert(binary_byte))
}
}
/// An encoding to use when searching.
///
/// An encoding can be used to configure a
/// [`SearcherBuilder`](struct.SearchBuilder.html)
/// to transcode source data from an encoding to UTF-8 before searching.
///
/// An `Encoding` will always be cheap to clone.
#[derive(Clone, Debug)]
pub struct Encoding(&'static encoding_rs::Encoding);
impl Encoding {
/// Create a new encoding for the specified label.
///
/// The encoding label provided is mapped to an encoding via the set of
/// available choices specified in the
/// [Encoding Standard](https://encoding.spec.whatwg.org/#concept-encoding-get).
/// If the given label does not correspond to a valid encoding, then this
/// returns an error.
pub fn new(label: &str) -> Result<Encoding, ConfigError> {
let label = label.as_bytes();
match encoding_rs::Encoding::for_label_no_replacement(label) {
Some(encoding) => Ok(Encoding(encoding)),
None => {
Err(ConfigError::UnknownEncoding { label: label.to_vec() })
}
}
}
}
/// The internal configuration of a searcher. This is shared among several
/// search related types, but is only ever written to by the SearcherBuilder.
#[derive(Clone, Debug)]
pub struct Config {
/// The line terminator to use.
line_term: LineTerminator,
/// Whether to invert matching.
invert_match: bool,
/// The number of lines after a match to include.
after_context: usize,
/// The number of lines before a match to include.
before_context: usize,
/// Whether to enable unbounded context or not.
passthru: bool,
/// Whether to count line numbers.
line_number: bool,
/// The maximum amount of heap memory to use.
///
/// When not given, no explicit limit is enforced. When set to `0`, then
/// only the memory map search strategy is available.
heap_limit: Option<usize>,
/// The memory map strategy.
mmap: MmapChoice,
/// The binary data detection strategy.
binary: BinaryDetection,
/// Whether to enable matching across multiple lines.
multi_line: bool,
/// An encoding that, when present, causes the searcher to transcode all
/// input from the encoding to UTF-8.
encoding: Option<Encoding>,
}
impl Default for Config {
fn default() -> Config {
Config {
line_term: LineTerminator::default(),
invert_match: false,
after_context: 0,
before_context: 0,
passthru: false,
line_number: true,
heap_limit: None,
mmap: MmapChoice::default(),
binary: BinaryDetection::default(),
multi_line: false,
encoding: None,
}
}
}
impl Config {
/// Return the maximal amount of lines needed to fulfill this
/// configuration's context.
///
/// If this returns `0`, then no context is ever needed.
fn max_context(&self) -> usize {
cmp::max(self.before_context, self.after_context)
}
/// Build a line buffer from this configuration.
fn line_buffer(&self) -> LineBuffer {
let mut builder = LineBufferBuilder::new();
builder
.line_terminator(self.line_term.as_byte())
.binary_detection(self.binary.0);
if let Some(limit) = self.heap_limit {
let (capacity, additional) =
if limit <= DEFAULT_BUFFER_CAPACITY {
(limit, 0)
} else {
(DEFAULT_BUFFER_CAPACITY, limit - DEFAULT_BUFFER_CAPACITY)
};
builder
.capacity(capacity)
.buffer_alloc(BufferAllocation::Error(additional));
}
builder.build()
}
}
/// An error that can occur when building a searcher.
///
/// This error occurs when a non-sensical configuration is present when trying
/// to construct a `Searcher` from a `SearcherBuilder`.
#[derive(Clone, Debug, Eq, PartialEq)]
pub enum ConfigError {
/// Indicates that the heap limit configuration prevents all possible
/// search strategies from being used. For example, if the heap limit is
/// set to 0 and memory map searching is disabled or unavailable.
SearchUnavailable,
/// Occurs when a matcher reports a line terminator that is different than
/// the one configured in the searcher.
MismatchedLineTerminators {
/// The matcher's line terminator.
matcher: LineTerminator,
/// The searcher's line terminator.
searcher: LineTerminator,
},
/// Occurs when no encoding could be found for a particular label.
UnknownEncoding {
/// The provided encoding label that could not be found.
label: Vec<u8>,
},
/// Hints that destructuring should not be exhaustive.
///
/// This enum may grow additional variants, so this makes sure clients
/// don't count on exhaustive matching. (Otherwise, adding a new variant
/// could break existing code.)
#[doc(hidden)]
__Nonexhaustive,
}
impl ::std::error::Error for ConfigError {
fn description(&self) -> &str { "grep-searcher configuration error" }
}
impl fmt::Display for ConfigError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
match *self {
ConfigError::SearchUnavailable => {
write!(f, "grep config error: no available searchers")
}
ConfigError::MismatchedLineTerminators { matcher, searcher } => {
write!(
f,
"grep config error: mismatched line terminators, \
matcher has {:?} but searcher has {:?}",
matcher,
searcher
)
}
ConfigError::UnknownEncoding { ref label } => {
write!(
f,
"grep config error: unknown encoding: {}",
String::from_utf8_lossy(label),
)
}
_ => panic!("BUG: unexpected variant found"),
}
}
}
/// A builder for configuring a searcher.
///
/// A search builder permits specifying the configuration of a searcher,
/// including options like whether to invert the search or to enable multi
/// line search.
///
/// Once a searcher has been built, it is beneficial to reuse that searcher
/// for multiple searches, if possible.
#[derive(Clone, Debug)]
pub struct SearcherBuilder {
config: Config,
}
impl Default for SearcherBuilder {
fn default() -> SearcherBuilder {
SearcherBuilder::new()
}
}
impl SearcherBuilder {
/// Create a new searcher builder with a default configuration.
pub fn new() -> SearcherBuilder {
SearcherBuilder {
config: Config::default(),
}
}
/// Build a searcher with the given matcher.
pub fn build(&self) -> Searcher {
let mut config = self.config.clone();
if config.passthru {
config.before_context = 0;
config.after_context = 0;
}
let mut decode_builder = DecodeReaderBytesBuilder::new();
decode_builder
.encoding(self.config.encoding.as_ref().map(|e| e.0))
.utf8_passthru(true)
.bom_override(true);
Searcher {
config: config,
decode_builder: decode_builder,
decode_buffer: RefCell::new(vec![0; 8 * (1<<10)]),
line_buffer: RefCell::new(self.config.line_buffer()),
multi_line_buffer: RefCell::new(vec![]),
}
}
/// Set the line terminator that is used by the searcher.
///
/// When using a searcher, if the matcher provided has a line terminator
/// set, then it must be the same as this one. If they aren't, building
/// a searcher will return an error.
///
/// By default, this is set to `b'\n'`.
pub fn line_terminator(
&mut self,
line_term: LineTerminator,
) -> &mut SearcherBuilder {
self.config.line_term = line_term;
self
}
/// Whether to invert matching, whereby lines that don't match are reported
/// instead of reporting lines that do match.
///
/// By default, this is disabled.
pub fn invert_match(&mut self, yes: bool) -> &mut SearcherBuilder {
self.config.invert_match = yes;
self
}
/// Whether to count and include line numbers with matching lines.
///
/// This is enabled by default. There is a small performance penalty
/// associated with computing line numbers, so this can be disabled when
/// this isn't desirable.
pub fn line_number(&mut self, yes: bool) -> &mut SearcherBuilder {
self.config.line_number = yes;
self
}
/// Whether to enable multi line search or not.
///
/// When multi line search is enabled, matches *may* match across multiple
/// lines. Conversely, when multi line search is disabled, it is impossible
/// for any match to span more than one line.
///
/// **Warning:** multi line search requires having the entire contents to
/// search mapped in memory at once. When searching files, memory maps
/// will be used if possible and if they are enabled, which avoids using
/// your program's heap. However, if memory maps cannot be used (e.g.,
/// for searching streams like `stdin` or if transcoding is necessary),
/// then the entire contents of the stream are read on to the heap before
/// starting the search.
///
/// This is disabled by default.
pub fn multi_line(&mut self, yes: bool) -> &mut SearcherBuilder {
self.config.multi_line = yes;
self
}
/// Whether to include a fixed number of lines after every match.
///
/// When this is set to a non-zero number, then the searcher will report
/// `line_count` contextual lines after every match.
///
/// This is set to `0` by default.
pub fn after_context(
&mut self,
line_count: usize,
) -> &mut SearcherBuilder {
self.config.after_context = line_count;
self
}
/// Whether to include a fixed number of lines before every match.
///
/// When this is set to a non-zero number, then the searcher will report
/// `line_count` contextual lines before every match.
///
/// This is set to `0` by default.
pub fn before_context(
&mut self,
line_count: usize,
) -> &mut SearcherBuilder {
self.config.before_context = line_count;
self
}
/// Whether to enable the "passthru" feature or not.
///
/// When passthru is enabled, it effectively treats all non-matching lines
/// as contextual lines. In other words, enabling this is akin to
/// requesting an unbounded number of before and after contextual lines.
///
/// When passthru mode is enabled, any `before_context` or `after_context`
/// settings are ignored by setting them to `0`.
///
/// This is disabled by default.
pub fn passthru(&mut self, yes: bool) -> &mut SearcherBuilder {
self.config.passthru = yes;
self
}
/// Set an approximate limit on the amount of heap space used by a
/// searcher.
///
/// The heap limit is enforced in two scenarios:
///
/// * When searching using a fixed size buffer, the heap limit controls
/// how big this buffer is allowed to be. Assuming contexts are disabled,
/// the minimum size of this buffer is the length (in bytes) of the
/// largest single line in the contents being searched. If any line
/// exceeds the heap limit, then an error will be returned.
/// * When performing a multi line search, a fixed size buffer cannot be
/// used. Thus, the only choices are to read the entire contents on to
/// the heap, or use memory maps. In the former case, the heap limit set
/// here is enforced.
///
/// If a heap limit is set to `0`, then no heap space is used. If there are
/// no alternative strategies available for searching without heap space
/// (e.g., memory maps are disabled), then the searcher wil return an error
/// immediately.
///
/// By default, no limit is set.
pub fn heap_limit(
&mut self,
bytes: Option<usize>,
) -> &mut SearcherBuilder {
self.config.heap_limit = bytes;
self
}
/// Set the strategy to employ use of memory maps.
///
/// Currently, there are only two strategies that can be employed:
///
/// * **Automatic** - A searcher will use heuristics, including but not
/// limited to file size and platform, to determine whether to use memory
/// maps or not.
/// * **Never** - Memory maps will never be used. If multi line search is
/// enabled, then the entire contents will be read on to the heap before
/// searching begins.
///
/// The default behavior is **never**. Generally speaking, and perhaps
/// against conventional wisdom, memory maps don't necessarily enable
/// faster searching. For example, depending on the platform, using memory
/// maps while searching a large directory can actually be quite a bit
/// slower than using normal read calls because of the overhead of managing
/// the memory maps.
///
/// Memory maps can be faster in some cases however. On some platforms,
/// when searching a very large file that *is already in memory*, it can
/// be slightly faster to search it as a memory map instead of using
/// normal read calls.
///
/// Finally, memory maps have a somewhat complicated safety story in Rust.
/// If you aren't sure whether enabling memory maps is worth it, then just
/// don't bother with it.
///
/// **WARNING**: If your process is searching a file backed memory map
/// at the same time that file is truncated, then it's possible for the
/// process to terminate with a bus error.
pub fn memory_map(
&mut self,
strategy: MmapChoice,
) -> &mut SearcherBuilder {
self.config.mmap = strategy;
self
}
/// Set the binary detection strategy.
///
/// The binary detection strategy determines not only how the searcher
/// detects binary data, but how it responds to the presence of binary
/// data. See the [`BinaryDetection`](struct.BinaryDetection.html) type
/// for more information.
///
/// By default, binary detection is disabled.
pub fn binary_detection(
&mut self,
detection: BinaryDetection,
) -> &mut SearcherBuilder {
self.config.binary = detection;
self
}
/// Set the encoding used to read the source data before searching.
///
/// When an encoding is provided, then the source data is _unconditionally_
/// transcoded using the encoding, unless a BOM is present. If a BOM is
/// present, then the encoding indicated by the BOM is used instead. If the
/// transcoding process encounters an error, then bytes are replaced with
/// the Unicode replacement codepoint.
///
/// When no encoding is specified (the default), then BOM sniffing is used
/// to determine whether the source data is UTF-8 or UTF-16, and
/// transcoding will be performed automatically. If no BOM could be found,
/// then the source data is searched _as if_ it were UTF-8. However, so
/// long as the source data is at least ASCII compatible, then it is
/// possible for a search to produce useful results.
pub fn encoding(
&mut self,
encoding: Option<Encoding>,
) -> &mut SearcherBuilder {
self.config.encoding = encoding;
self
}
}
/// A searcher executes searches over a haystack and writes results to a caller
/// provided sink.
///
/// Matches are detected via implementations of the `Matcher` trait, which must
/// be provided by the caller when executing a search.
///
/// When possible, a searcher should be reused.
#[derive(Clone, Debug)]
pub struct Searcher {
/// The configuration for this searcher.
///
/// We make most of these settings available to users of `Searcher` via
/// public API methods, which can be queried in implementations of `Sink`
/// if necessary.
config: Config,
/// A builder for constructing a streaming reader that transcodes source
/// data according to either an explicitly specified encoding or via an
/// automatically detected encoding via BOM sniffing.
///
/// When no transcoding is needed, then the transcoder built will pass
/// through the underlying bytes with no additional overhead.
decode_builder: DecodeReaderBytesBuilder,
/// A buffer that is used for transcoding scratch space.
decode_buffer: RefCell<Vec<u8>>,
/// A line buffer for use in line oriented searching.
///
/// We wrap it in a RefCell to permit lending out borrows of `Searcher`
/// to sinks. We still require a mutable borrow to execute a search, so
/// we statically prevent callers from causing RefCell to panic at runtime
/// due to a borrowing violation.
line_buffer: RefCell<LineBuffer>,
/// A buffer in which to store the contents of a reader when performing a
/// multi line search. In particular, multi line searches cannot be
/// performed incrementally, and need the entire haystack in memory at
/// once.
multi_line_buffer: RefCell<Vec<u8>>,
}
impl Searcher {
/// Create a new searcher with a default configuration.
///
/// To configure the searcher (e.g., invert matching, enable memory maps,
/// enable contexts, etc.), use the
/// [`SearcherBuilder`](struct.SearcherBuilder.html).
pub fn new() -> Searcher {
SearcherBuilder::new().build()
}
/// Execute a search over the file with the given path and write the
/// results to the given sink.
///
/// If memory maps are enabled and the searcher heuristically believes
/// memory maps will help the search run faster, then this will use
/// memory maps. For this reason, callers should prefer using this method
/// or `search_file` over the more generic `search_reader` when possible.
pub fn search_path<P, M, S>(
&mut self,
matcher: M,
path: P,
write_to: S,
) -> Result<(), S::Error>
where P: AsRef<Path>,
M: Matcher,
S: Sink,
{
let path = path.as_ref();
let file = File::open(path).map_err(S::Error::error_io)?;
self.search_file_maybe_path(matcher, Some(path), &file, write_to)
}
/// Execute a search over a file and write the results to the given sink.
///
/// If memory maps are enabled and the searcher heuristically believes
/// memory maps will help the search run faster, then this will use
/// memory maps. For this reason, callers should prefer using this method
/// or `search_path` over the more generic `search_reader` when possible.
pub fn search_file<M, S>(
&mut self,
matcher: M,
file: &File,
write_to: S,
) -> Result<(), S::Error>
where M: Matcher,
S: Sink,
{
self.search_file_maybe_path(matcher, None, file, write_to)
}
fn search_file_maybe_path<M, S>(
&mut self,
matcher: M,
path: Option<&Path>,
file: &File,
write_to: S,
) -> Result<(), S::Error>
where M: Matcher,
S: Sink,
{
if let Some(mmap) = self.config.mmap.open(file, path) {
trace!("{:?}: searching via memory map", path);
return self.search_slice(matcher, &mmap, write_to);
}
// Fast path for multi-line searches of files when memory maps are
// not enabled. This pre-allocates a buffer roughly the size of the
// file, which isn't possible when searching an arbitrary io::Read.
if self.multi_line_with_matcher(&matcher) {
trace!("{:?}: reading entire file on to heap for mulitline", path);
self.fill_multi_line_buffer_from_file::<S>(file)?;
trace!("{:?}: searching via multiline strategy", path);
MultiLine::new(
self,
matcher,
&*self.multi_line_buffer.borrow(),
write_to,
).run()
} else {
trace!("{:?}: searching using generic reader", path);
self.search_reader(matcher, file, write_to)
}
}
/// Execute a search over any implementation of `io::Read` and write the
/// results to the given sink.
///
/// When possible, this implementation will search the reader incrementally
/// without reading it into memory. In some cases---for example, if multi
/// line search is enabled---an incremental search isn't possible and the
/// given reader is consumed completely and placed on the heap before
/// searching begins. For this reason, when multi line search is enabled,
/// one should try to use higher level APIs (e.g., searching by file or
/// file path) so that memory maps can be used if they are available and
/// enabled.
pub fn search_reader<M, R, S>(
&mut self,
matcher: M,
read_from: R,
write_to: S,
) -> Result<(), S::Error>
where M: Matcher,
R: io::Read,
S: Sink,
{
self.check_config(&matcher).map_err(S::Error::error_config)?;
let mut decode_buffer = self.decode_buffer.borrow_mut();
let read_from = self.decode_builder
.build_with_buffer(read_from, &mut *decode_buffer)
.map_err(S::Error::error_io)?;
if self.multi_line_with_matcher(&matcher) {
trace!("generic reader: reading everything to heap for multiline");
self.fill_multi_line_buffer_from_reader::<_, S>(read_from)?;
trace!("generic reader: searching via multiline strategy");
MultiLine::new(
self,
matcher,
&*self.multi_line_buffer.borrow(),
write_to,
).run()
} else {
let mut line_buffer = self.line_buffer.borrow_mut();
let rdr = LineBufferReader::new(read_from, &mut *line_buffer);
trace!("generic reader: searching via roll buffer strategy");
ReadByLine::new(self, matcher, rdr, write_to).run()
}
}
/// Execute a search over the given slice and write the results to the
/// given sink.
pub fn search_slice<M, S>(
&mut self,
matcher: M,
slice: &[u8],
write_to: S,
) -> Result<(), S::Error>
where M: Matcher,
S: Sink,
{
self.check_config(&matcher).map_err(S::Error::error_config)?;
// We can search the slice directly, unless we need to do transcoding.
if self.slice_needs_transcoding(slice) {
trace!("slice reader: needs transcoding, using generic reader");
return self.search_reader(matcher, slice, write_to);
}
if self.multi_line_with_matcher(&matcher) {
trace!("slice reader: searching via multiline strategy");
MultiLine::new(self, matcher, slice, write_to).run()
} else {
trace!("slice reader: searching via slice-by-line strategy");
SliceByLine::new(self, matcher, slice, write_to).run()
}
}
/// Check that the searcher's configuration and the matcher are consistent
/// with each other.
fn check_config<M: Matcher>(&self, matcher: M) -> Result<(), ConfigError> {
if self.config.heap_limit == Some(0)
&& !self.config.mmap.is_enabled()
{
return Err(ConfigError::SearchUnavailable);
}
let matcher_line_term = match matcher.line_terminator() {
None => return Ok(()),
Some(line_term) => line_term,
};
if matcher_line_term != self.config.line_term {
return Err(ConfigError::MismatchedLineTerminators {
matcher: matcher_line_term,
searcher: self.config.line_term,
});
}
Ok(())
}
/// Returns true if and only if the given slice needs to be transcoded.
fn slice_needs_transcoding(&self, slice: &[u8]) -> bool {
self.config.encoding.is_some() || slice_has_utf16_bom(slice)
}
}
/// The following methods permit querying the configuration of a searcher.
/// These can be useful in generic implementations of
/// [`Sink`](trait.Sink.html),
/// where the output may be tailored based on how the searcher is configured.
impl Searcher {
/// Returns the line terminator used by this searcher.
#[inline]
pub fn line_terminator(&self) -> LineTerminator {
self.config.line_term
}
/// Returns true if and only if this searcher is configured to invert its
/// search results. That is, matching lines are lines that do **not** match
/// the searcher's matcher.
#[inline]
pub fn invert_match(&self) -> bool {
self.config.invert_match
}
/// Returns true if and only if this searcher is configured to count line
/// numbers.
#[inline]
pub fn line_number(&self) -> bool {
self.config.line_number
}
/// Returns true if and only if this searcher is configured to perform
/// multi line search.
#[inline]
pub fn multi_line(&self) -> bool {
self.config.multi_line
}
/// Returns true if and only if this searcher will choose a multi-line
/// strategy given the provided matcher.
///
/// This may diverge from the result of `multi_line` in cases where the
/// searcher has been configured to execute a search that can report
/// matches over multiple lines, but where the matcher guarantees that it
/// will never produce a match over multiple lines.
pub fn multi_line_with_matcher<M: Matcher>(&self, matcher: M) -> bool {
if !self.multi_line() {
return false;
}
if let Some(line_term) = matcher.line_terminator() {
if line_term == self.line_terminator() {
return false;
}
}
if let Some(non_matching) = matcher.non_matching_bytes() {
// If the line terminator is CRLF, we don't actually need to care
// whether the regex can match `\r` or not. Namely, a `\r` is
// neither necessary nor sufficient to terminate a line. A `\n` is
// always required.
if non_matching.contains(self.line_terminator().as_byte()) {
return false;
}
}
true
}
/// Returns the number of "after" context lines to report. When context
/// reporting is not enabled, this returns `0`.
#[inline]
pub fn after_context(&self) -> usize {
self.config.after_context
}
/// Returns the number of "before" context lines to report. When context
/// reporting is not enabled, this returns `0`.
#[inline]
pub fn before_context(&self) -> usize {
self.config.before_context
}
/// Returns true if and only if the searcher has "passthru" mode enabled.
#[inline]
pub fn passthru(&self) -> bool {
self.config.passthru
}
/// Fill the buffer for use with multi-line searching from the given file.
/// This reads from the file until EOF or until an error occurs. If the
/// contents exceed the configured heap limit, then an error is returned.
fn fill_multi_line_buffer_from_file<S: Sink>(
&self,
file: &File,
) -> Result<(), S::Error> {
assert!(self.config.multi_line);
let mut decode_buffer = self.decode_buffer.borrow_mut();
let mut read_from = self.decode_builder
.build_with_buffer(file, &mut *decode_buffer)
.map_err(S::Error::error_io)?;
// If we don't have a heap limit, then we can defer to std's
// read_to_end implementation. fill_multi_line_buffer_from_reader will
// do this too, but since we have a File, we can be a bit smarter about
// pre-allocating here.
//
// If we're transcoding, then our pre-allocation might not be exact,
// but is probably still better than nothing.
if self.config.heap_limit.is_none() {
let mut buf = self.multi_line_buffer.borrow_mut();
buf.clear();
let cap = file
.metadata()
.map(|m| m.len() as usize + 1)
.unwrap_or(0);
buf.reserve(cap);
read_from.read_to_end(&mut *buf).map_err(S::Error::error_io)?;
return Ok(());
}
self.fill_multi_line_buffer_from_reader::<_, S>(read_from)
}
/// Fill the buffer for use with multi-line searching from the given
/// reader. This reads from the reader until EOF or until an error occurs.
/// If the contents exceed the configured heap limit, then an error is
/// returned.
fn fill_multi_line_buffer_from_reader<R: io::Read, S: Sink>(
&self,
mut read_from: R,
) -> Result<(), S::Error> {
assert!(self.config.multi_line);
let mut buf = self.multi_line_buffer.borrow_mut();
buf.clear();
// If we don't have a heap limit, then we can defer to std's
// read_to_end implementation...
let heap_limit = match self.config.heap_limit {
Some(heap_limit) => heap_limit,
None => {
read_from.read_to_end(&mut *buf).map_err(S::Error::error_io)?;
return Ok(());
}
};
if heap_limit == 0 {
return Err(S::Error::error_io(alloc_error(heap_limit)));
}
// ... otherwise we need to roll our own. This is likely quite a bit
// slower than what is optimal, but we avoid worry about memory safety
// until there's a compelling reason to speed this up.
buf.resize(cmp::min(DEFAULT_BUFFER_CAPACITY, heap_limit), 0);
let mut pos = 0;
loop {
let nread = match read_from.read(&mut buf[pos..]) {
Ok(nread) => nread,
Err(ref err) if err.kind() == io::ErrorKind::Interrupted => {
continue;
}
Err(err) => return Err(S::Error::error_io(err)),
};
if nread == 0 {
buf.resize(pos, 0);
return Ok(());
}
pos += nread;
if buf[pos..].is_empty() {
let additional = heap_limit - buf.len();
if additional == 0 {
return Err(S::Error::error_io(alloc_error(heap_limit)));
}
let limit = buf.len() + additional;
let doubled = 2 * buf.len();
buf.resize(cmp::min(doubled, limit), 0);
}
}
}
}
/// Returns true if and only if the given slice begins with a UTF-16 BOM.
///
/// This is used by the searcher to determine if a transcoder is necessary.
/// Otherwise, it is advantageous to search the slice directly.
fn slice_has_utf16_bom(slice: &[u8]) -> bool {
let enc = match encoding_rs::Encoding::for_bom(slice) {
None => return false,
Some((enc, _)) => enc,
};
[encoding_rs::UTF_16LE, encoding_rs::UTF_16BE].contains(&enc)
}
#[cfg(test)]
mod tests {
use testutil::{KitchenSink, RegexMatcher};
use super::*;
#[test]
fn config_error_heap_limit() {
let matcher = RegexMatcher::new("");
let sink = KitchenSink::new();
let mut searcher = SearcherBuilder::new()
.heap_limit(Some(0))
.build();
let res = searcher.search_slice(matcher, &[], sink);
assert!(res.is_err());
}
#[test]
fn config_error_line_terminator() {
let mut matcher = RegexMatcher::new("");
matcher.set_line_term(Some(LineTerminator::byte(b'z')));
let sink = KitchenSink::new();
let mut searcher = Searcher::new();
let res = searcher.search_slice(matcher, &[], sink);
assert!(res.is_err());
}
}

559
grep-searcher/src/sink.rs Normal file
View File

@@ -0,0 +1,559 @@
use std::fmt;
use std::io;
use grep_matcher::LineTerminator;
use lines::LineIter;
use searcher::{ConfigError, Searcher};
/// A trait that describes errors that can be reported by searchers and
/// implementations of `Sink`.
///
/// Unless you have a specialized use case, you probably don't need to
/// implement this trait explicitly. It's likely that using `io::Error` (which
/// implements this trait) for your error type is good enough, largely because
/// most errors that occur during search will likely be an `io::Error`.
pub trait SinkError: Sized {
/// A constructor for converting any value that satisfies the
/// `fmt::Display` trait into an error.
fn error_message<T: fmt::Display>(message: T) -> Self;
/// A constructor for converting I/O errors that occur while searching into
/// an error of this type.
///
/// By default, this is implemented via the `error_message` constructor.
fn error_io(err: io::Error) -> Self {
Self::error_message(err)
}
/// A constructor for converting configuration errors that occur while
/// building a searcher into an error of this type.
///
/// By default, this is implemented via the `error_message` constructor.
fn error_config(err: ConfigError) -> Self {
Self::error_message(err)
}
}
/// An `io::Error` can be used as an error for `Sink` implementations out of
/// the box.
impl SinkError for io::Error {
fn error_message<T: fmt::Display>(message: T) -> io::Error {
io::Error::new(io::ErrorKind::Other, message.to_string())
}
fn error_io(err: io::Error) -> io::Error {
err
}
}
/// A `Box<std::error::Error>` can be used as an error for `Sink`
/// implementations out of the box.
impl SinkError for Box<::std::error::Error> {
fn error_message<T: fmt::Display>(message: T) -> Box<::std::error::Error> {
Box::<::std::error::Error>::from(message.to_string())
}
}
/// A trait that defines how results from searchers are handled.
///
/// In this crate, a searcher follows the "push" model. What that means is that
/// the searcher drives execution, and pushes results back to the caller. This
/// is in contrast to a "pull" model where the caller drives execution and
/// takes results as they need them. These are also known as "internal" and
/// "external" iteration strategies, respectively.
///
/// For a variety of reasons, including the complexity of the searcher
/// implementation, this crate chooses the "push" or "internal" model of
/// execution. Thus, in order to act on search results, callers must provide
/// an implementation of this trait to a searcher, and the searcher is then
/// responsible for calling the methods on this trait.
///
/// This trait defines several behaviors:
///
/// * What to do when a match is found. Callers must provide this.
/// * What to do when an error occurs. Callers must provide this via the
/// [`SinkError`](trait.SinkError.html) trait. Generally, callers can just
/// use `io::Error` for this, which already implements `SinkError`.
/// * What to do when a contextual line is found. By default, these are
/// ignored.
/// * What to do when a gap between contextual lines has been found. By
/// default, this is ignored.
/// * What to do when a search has started. By default, this does nothing.
/// * What to do when a search has finished successfully. By default, this does
/// nothing.
///
/// Callers must, at minimum, specify the behavior when an error occurs and
/// the behavior when a match occurs. The rest is optional. For each behavior,
/// callers may report an error (say, if writing the result to another
/// location failed) or simply return `false` if they want the search to stop
/// (e.g., when implementing a cap on the number of search results to show).
///
/// When errors are reported (whether in the searcher or in the implementation
/// of `Sink`), then searchers quit immediately without calling `finish`.
///
/// For simpler uses of `Sink`, callers may elect to use one of
/// the more convenient but less flexible implementations in the
/// [`sinks`](sinks/index.html) module.
pub trait Sink {
/// The type of an error that should be reported by a searcher.
///
/// Errors of this type are not only returned by the methods on this
/// trait, but the constructors defined in `SinkError` are also used in
/// the searcher implementation itself. e.g., When a I/O error occurs when
/// reading data from a file.
type Error: SinkError;
/// This method is called whenever a match is found.
///
/// If multi line is enabled on the searcher, then the match reported here
/// may span multiple lines and it may include multiple matches. When multi
/// line is disabled, then the match is guaranteed to span exactly one
/// non-empty line (where a single line is, at minimum, a line terminator).
///
/// If this returns `true`, then searching continues. If this returns
/// `false`, then searching is stopped immediately and `finish` is called.
///
/// If this returns an error, then searching is stopped immediately,
/// `finish` is not called and the error is bubbled back up to the caller
/// of the searcher.
fn matched(
&mut self,
_searcher: &Searcher,
_mat: &SinkMatch,
) -> Result<bool, Self::Error>;
/// This method is called whenever a context line is found, and is optional
/// to implement. By default, it does nothing and returns `true`.
///
/// In all cases, the context given is guaranteed to span exactly one
/// non-empty line (where a single line is, at minimum, a line terminator).
///
/// If this returns `true`, then searching continues. If this returns
/// `false`, then searching is stopped immediately and `finish` is called.
///
/// If this returns an error, then searching is stopped immediately,
/// `finish` is not called and the error is bubbled back up to the caller
/// of the searcher.
#[inline]
fn context(
&mut self,
_searcher: &Searcher,
_context: &SinkContext,
) -> Result<bool, Self::Error> {
Ok(true)
}
/// This method is called whenever a break in contextual lines is found,
/// and is optional to implement. By default, it does nothing and returns
/// `true`.
///
/// A break can only occur when context reporting is enabled (that is,
/// either or both of `before_context` or `after_context` are greater than
/// `0`). More precisely, a break occurs between non-contiguous groups of
/// lines.
///
/// If this returns `true`, then searching continues. If this returns
/// `false`, then searching is stopped immediately and `finish` is called.
///
/// If this returns an error, then searching is stopped immediately,
/// `finish` is not called and the error is bubbled back up to the caller
/// of the searcher.
#[inline]
fn context_break(
&mut self,
_searcher: &Searcher,
) -> Result<bool, Self::Error> {
Ok(true)
}
/// This method is called when a search has begun, before any search is
/// executed. By default, this does nothing.
///
/// If this returns `true`, then searching continues. If this returns
/// `false`, then searching is stopped immediately and `finish` is called.
///
/// If this returns an error, then searching is stopped immediately,
/// `finish` is not called and the error is bubbled back up to the caller
/// of the searcher.
#[inline]
fn begin(
&mut self,
_searcher: &Searcher,
) -> Result<bool, Self::Error> {
Ok(true)
}
/// This method is called when a search has completed. By default, this
/// does nothing.
///
/// If this returns an error, the error is bubbled back up to the caller of
/// the searcher.
#[inline]
fn finish(
&mut self,
_searcher: &Searcher,
_: &SinkFinish,
) -> Result<(), Self::Error> {
Ok(())
}
}
impl<'a, S: Sink> Sink for &'a mut S {
type Error = S::Error;
#[inline]
fn matched(
&mut self,
searcher: &Searcher,
mat: &SinkMatch,
) -> Result<bool, S::Error> {
(**self).matched(searcher, mat)
}
#[inline]
fn context(
&mut self,
searcher: &Searcher,
context: &SinkContext,
) -> Result<bool, S::Error> {
(**self).context(searcher, context)
}
#[inline]
fn context_break(
&mut self,
searcher: &Searcher,
) -> Result<bool, S::Error> {
(**self).context_break(searcher)
}
#[inline]
fn begin(
&mut self,
searcher: &Searcher,
) -> Result<bool, S::Error> {
(**self).begin(searcher)
}
#[inline]
fn finish(
&mut self,
searcher: &Searcher,
sink_finish: &SinkFinish,
) -> Result<(), S::Error> {
(**self).finish(searcher, sink_finish)
}
}
/// Summary data reported at the end of a search.
///
/// This reports data such as the total number of bytes searched and the
/// absolute offset of the first occurrence of binary data, if any were found.
///
/// A searcher that stops early because of an error does not call `finish`.
/// A searcher that stops early because the `Sink` implementor instructed it
/// to will still call `finish`.
#[derive(Clone, Debug)]
pub struct SinkFinish {
pub(crate) byte_count: u64,
pub(crate) binary_byte_offset: Option<u64>,
}
impl SinkFinish {
/// Return the total number of bytes searched.
#[inline]
pub fn byte_count(&self) -> u64 {
self.byte_count
}
/// If binary detection is enabled and if binary data was found, then this
/// returns the absolute byte offset of the first detected byte of binary
/// data.
///
/// Note that since this is an absolute byte offset, it cannot be relied
/// upon to index into any addressable memory.
#[inline]
pub fn binary_byte_offset(&self) -> Option<u64> {
self.binary_byte_offset
}
}
/// A type that describes a match reported by a searcher.
#[derive(Clone, Debug)]
pub struct SinkMatch<'b> {
pub(crate) line_term: LineTerminator,
pub(crate) bytes: &'b [u8],
pub(crate) absolute_byte_offset: u64,
pub(crate) line_number: Option<u64>,
}
impl<'b> SinkMatch<'b> {
/// Returns the bytes for all matching lines, including the line
/// terminators, if they exist.
#[inline]
pub fn bytes(&self) -> &'b [u8] {
self.bytes
}
/// Return an iterator over the lines in this match.
///
/// If multi line search is enabled, then this may yield more than one
/// line (but always at least one line). If multi line search is disabled,
/// then this always reports exactly one line (but may consist of just
/// the line terminator).
///
/// Lines yielded by this iterator include their terminators.
#[inline]
pub fn lines(&self) -> LineIter<'b> {
LineIter::new(self.line_term.as_byte(), self.bytes)
}
/// Returns the absolute byte offset of the start of this match. This
/// offset is absolute in that it is relative to the very beginning of the
/// input in a search, and can never be relied upon to be a valid index
/// into an in-memory slice.
#[inline]
pub fn absolute_byte_offset(&self) -> u64 {
self.absolute_byte_offset
}
/// Returns the line number of the first line in this match, if available.
///
/// Line numbers are only available when the search builder is instructed
/// to compute them.
#[inline]
pub fn line_number(&self) -> Option<u64> {
self.line_number
}
}
/// The type of context reported by a searcher.
#[derive(Clone, Debug, Eq, PartialEq)]
pub enum SinkContextKind {
/// The line reported occurred before a match.
Before,
/// The line reported occurred after a match.
After,
/// Any other type of context reported, e.g., as a result of a searcher's
/// "passthru" mode.
Other,
}
/// A type that describes a contextual line reported by a searcher.
#[derive(Clone, Debug)]
pub struct SinkContext<'b> {
pub(crate) line_term: LineTerminator,
pub(crate) bytes: &'b [u8],
pub(crate) kind: SinkContextKind,
pub(crate) absolute_byte_offset: u64,
pub(crate) line_number: Option<u64>,
}
impl<'b> SinkContext<'b> {
/// Returns the context bytes, including line terminators.
#[inline]
pub fn bytes(&self) -> &'b [u8] {
self.bytes
}
/// Returns the type of context.
#[inline]
pub fn kind(&self) -> &SinkContextKind {
&self.kind
}
/// Return an iterator over the lines in this match.
///
/// This always yields exactly one line (and that one line may contain just
/// the line terminator).
///
/// Lines yielded by this iterator include their terminators.
#[cfg(test)]
pub(crate) fn lines(&self) -> LineIter<'b> {
LineIter::new(self.line_term.as_byte(), self.bytes)
}
/// Returns the absolute byte offset of the start of this context. This
/// offset is absolute in that it is relative to the very beginning of the
/// input in a search, and can never be relied upon to be a valid index
/// into an in-memory slice.
#[inline]
pub fn absolute_byte_offset(&self) -> u64 {
self.absolute_byte_offset
}
/// Returns the line number of the first line in this context, if
/// available.
///
/// Line numbers are only available when the search builder is instructed
/// to compute them.
#[inline]
pub fn line_number(&self) -> Option<u64> {
self.line_number
}
}
/// A collection of convenience implementations of `Sink`.
///
/// Each implementation in this module makes some kind of sacrifice in the name
/// of making common cases easier to use. Most frequently, each type is a
/// wrapper around a closure specified by the caller that provides limited
/// access to the full suite of information available to implementors of
/// `Sink`.
///
/// For example, the `UTF8` sink makes the following sacrifices:
///
/// * All matches must be UTF-8. An arbitrary `Sink` does not have this
/// restriction and can deal with arbitrary data. If this sink sees invalid
/// UTF-8, then an error is returned and searching stops. (Use the `Lossy`
/// sink instead to suppress this error.)
/// * The searcher must be configured to report line numbers. If it isn't,
/// an error is reported at the first match and searching stops.
/// * Context lines, context breaks and summary data reported at the end of
/// a search are all ignored.
/// * Implementors are forced to use `io::Error` as their error type.
///
/// If you need more flexibility, then you're advised to implement the `Sink`
/// trait directly.
pub mod sinks {
use std::io;
use std::str;
use searcher::Searcher;
use super::{Sink, SinkError, SinkMatch};
/// A sink that provides line numbers and matches as strings while ignoring
/// everything else.
///
/// This implementation will return an error if a match contains invalid
/// UTF-8 or if the searcher was not configured to count lines. Errors
/// on invalid UTF-8 can be suppressed by using the `Lossy` sink instead
/// of this one.
///
/// The closure accepts two parameters: a line number and a UTF-8 string
/// containing the matched data. The closure returns a
/// `Result<bool, io::Error>`. If the `bool` is `false`, then the search
/// stops immediately. Otherwise, searching continues.
///
/// If multi line mode was enabled, the line number refers to the line
/// number of the first line in the match.
#[derive(Clone, Debug)]
pub struct UTF8<F>(pub F)
where F: FnMut(u64, &str) -> Result<bool, io::Error>;
impl<F> Sink for UTF8<F>
where F: FnMut(u64, &str) -> Result<bool, io::Error>
{
type Error = io::Error;
fn matched(
&mut self,
_searcher: &Searcher,
mat: &SinkMatch,
) -> Result<bool, io::Error> {
let matched = match str::from_utf8(mat.bytes()) {
Ok(matched) => matched,
Err(err) => return Err(io::Error::error_message(err)),
};
let line_number = match mat.line_number() {
Some(line_number) => line_number,
None => {
let msg = "line numbers not enabled";
return Err(io::Error::error_message(msg));
}
};
(self.0)(line_number, &matched)
}
}
/// A sink that provides line numbers and matches as (lossily converted)
/// strings while ignoring everything else.
///
/// This is like `UTF8`, except that if a match contains invalid UTF-8,
/// then it will be lossily converted to valid UTF-8 by substituting
/// invalid UTF-8 with Unicode replacement characters.
///
/// This implementation will return an error on the first match if the
/// searcher was not configured to count lines.
///
/// The closure accepts two parameters: a line number and a UTF-8 string
/// containing the matched data. The closure returns a
/// `Result<bool, io::Error>`. If the `bool` is `false`, then the search
/// stops immediately. Otherwise, searching continues.
///
/// If multi line mode was enabled, the line number refers to the line
/// number of the first line in the match.
#[derive(Clone, Debug)]
pub struct Lossy<F>(pub F)
where F: FnMut(u64, &str) -> Result<bool, io::Error>;
impl<F> Sink for Lossy<F>
where F: FnMut(u64, &str) -> Result<bool, io::Error>
{
type Error = io::Error;
fn matched(
&mut self,
_searcher: &Searcher,
mat: &SinkMatch,
) -> Result<bool, io::Error> {
use std::borrow::Cow;
let matched = match str::from_utf8(mat.bytes()) {
Ok(matched) => Cow::Borrowed(matched),
// TODO: In theory, it should be possible to amortize
// allocation here, but `std` doesn't provide such an API.
// Regardless, this only happens on matches with invalid UTF-8,
// which should be pretty rare.
Err(_) => String::from_utf8_lossy(mat.bytes()),
};
let line_number = match mat.line_number() {
Some(line_number) => line_number,
None => {
let msg = "line numbers not enabled";
return Err(io::Error::error_message(msg));
}
};
(self.0)(line_number, &matched)
}
}
/// A sink that provides line numbers and matches as raw bytes while
/// ignoring everything else.
///
/// This implementation will return an error on the first match if the
/// searcher was not configured to count lines.
///
/// The closure accepts two parameters: a line number and a raw byte string
/// containing the matched data. The closure returns a `Result<bool,
/// io::Error>`. If the `bool` is `false`, then the search stops
/// immediately. Otherwise, searching continues.
///
/// If multi line mode was enabled, the line number refers to the line
/// number of the first line in the match.
#[derive(Clone, Debug)]
pub struct Bytes<F>(pub F)
where F: FnMut(u64, &[u8]) -> Result<bool, io::Error>;
impl<F> Sink for Bytes<F>
where F: FnMut(u64, &[u8]) -> Result<bool, io::Error>
{
type Error = io::Error;
fn matched(
&mut self,
_searcher: &Searcher,
mat: &SinkMatch,
) -> Result<bool, io::Error> {
let line_number = match mat.line_number() {
Some(line_number) => line_number,
None => {
let msg = "line numbers not enabled";
return Err(io::Error::error_message(msg));
}
};
(self.0)(line_number, mat.bytes())
}
}
}

787
grep-searcher/src/testutil.rs Normal file
View File

@@ -0,0 +1,787 @@
use std::io::{self, Write};
use std::str;
use grep_matcher::{
LineMatchKind, LineTerminator, Match, Matcher, NoCaptures, NoError,
};
use memchr::memchr;
use regex::bytes::{Regex, RegexBuilder};
use searcher::{BinaryDetection, Searcher, SearcherBuilder};
use sink::{Sink, SinkContext, SinkFinish, SinkMatch};
/// A simple regex matcher.
///
/// This supports setting the matcher's line terminator configuration directly,
/// which we use for testing purposes. That is, the caller explicitly
/// determines whether the line terminator optimization is enabled. (In reality
/// this optimization is detected automatically by inspecting and possibly
/// modifying the regex itself.)
#[derive(Clone, Debug)]
pub struct RegexMatcher {
regex: Regex,
line_term: Option<LineTerminator>,
every_line_is_candidate: bool,
}
impl RegexMatcher {
/// Create a new regex matcher.
pub fn new(pattern: &str) -> RegexMatcher {
let regex = RegexBuilder::new(pattern)
.multi_line(true) // permits ^ and $ to match at \n boundaries
.build()
.unwrap();
RegexMatcher {
regex: regex,
line_term: None,
every_line_is_candidate: false,
}
}
/// Forcefully set the line terminator of this matcher.
///
/// By default, this matcher has no line terminator set.
pub fn set_line_term(
&mut self,
line_term: Option<LineTerminator>,
) -> &mut RegexMatcher {
self.line_term = line_term;
self
}
/// Whether to return every line as a candidate or not.
///
/// This forces searchers to handle the case of reporting a false positive.
pub fn every_line_is_candidate(
&mut self,
yes: bool,
) -> &mut RegexMatcher {
self.every_line_is_candidate = yes;
self
}
}
impl Matcher for RegexMatcher {
type Captures = NoCaptures;
type Error = NoError;
fn find_at(
&self,
haystack: &[u8],
at: usize,
) -> Result<Option<Match>, NoError> {
Ok(self.regex
.find_at(haystack, at)
.map(|m| Match::new(m.start(), m.end())))
}
fn new_captures(&self) -> Result<NoCaptures, NoError> {
Ok(NoCaptures::new())
}
fn line_terminator(&self) -> Option<LineTerminator> {
self.line_term
}
fn find_candidate_line(
&self,
haystack: &[u8],
) -> Result<Option<LineMatchKind>, NoError> {
if self.every_line_is_candidate {
assert!(self.line_term.is_some());
if haystack.is_empty() {
return Ok(None);
}
// Make it interesting and return the last byte in the current
// line.
let i = memchr(self.line_term.unwrap().as_byte(), haystack)
.map(|i| i)
.unwrap_or(haystack.len() - 1);
Ok(Some(LineMatchKind::Candidate(i)))
} else {
Ok(self.shortest_match(haystack)?.map(LineMatchKind::Confirmed))
}
}
}
/// An implementation of Sink that prints all available information.
///
/// This is useful for tests because it lets us easily confirm whether data
/// is being passed to Sink correctly.
#[derive(Clone, Debug)]
pub struct KitchenSink(Vec<u8>);
impl KitchenSink {
/// Create a new implementation of Sink that includes everything in the
/// kitchen.
pub fn new() -> KitchenSink {
KitchenSink(vec![])
}
/// Return the data written to this sink.
pub fn as_bytes(&self) -> &[u8] {
&self.0
}
}
impl Sink for KitchenSink {
type Error = io::Error;
fn matched(
&mut self,
_searcher: &Searcher,
mat: &SinkMatch,
) -> Result<bool, io::Error> {
assert!(!mat.bytes().is_empty());
assert!(mat.lines().count() >= 1);
let mut line_number = mat.line_number();
let mut byte_offset = mat.absolute_byte_offset();
for line in mat.lines() {
if let Some(ref mut n) = line_number {
write!(self.0, "{}:", n)?;
*n += 1;
}
write!(self.0, "{}:", byte_offset)?;
byte_offset += line.len() as u64;
self.0.write_all(line)?;
}
Ok(true)
}
fn context(
&mut self,
_searcher: &Searcher,
context: &SinkContext,
) -> Result<bool, io::Error> {
assert!(!context.bytes().is_empty());
assert!(context.lines().count() == 1);
if let Some(line_number) = context.line_number() {
write!(self.0, "{}-", line_number)?;
}
write!(self.0, "{}-", context.absolute_byte_offset)?;
self.0.write_all(context.bytes())?;
Ok(true)
}
fn context_break(
&mut self,
_searcher: &Searcher,
) -> Result<bool, io::Error> {
self.0.write_all(b"--\n")?;
Ok(true)
}
fn finish(
&mut self,
_searcher: &Searcher,
sink_finish: &SinkFinish,
) -> Result<(), io::Error> {
writeln!(self.0, "")?;
writeln!(self.0, "byte count:{}", sink_finish.byte_count())?;
if let Some(offset) = sink_finish.binary_byte_offset() {
writeln!(self.0, "binary offset:{}", offset)?;
}
Ok(())
}
}
/// A type for expressing tests on a searcher.
///
/// The searcher code has a lot of different code paths, mostly for the
/// purposes of optimizing a bunch of different use cases. The intent of the
/// searcher is to pick the best code path based on the configuration, which
/// means there is no obviously direct way to ask that a specific code path
/// be exercised. Thus, the purpose of this tester is to explicitly check as
/// many code paths that make sense.
///
/// The tester works by assuming you want to test all pertinent code paths.
/// These can be trimmed down as necessary via the various builder methods.
#[derive(Debug)]
pub struct SearcherTester {
haystack: String,
pattern: String,
filter: Option<::regex::Regex>,
print_labels: bool,
expected_no_line_number: Option<String>,
expected_with_line_number: Option<String>,
expected_slice_no_line_number: Option<String>,
expected_slice_with_line_number: Option<String>,
by_line: bool,
multi_line: bool,
invert_match: bool,
line_number: bool,
binary: BinaryDetection,
auto_heap_limit: bool,
after_context: usize,
before_context: usize,
passthru: bool,
}
impl SearcherTester {
/// Create a new tester for testing searchers.
pub fn new(haystack: &str, pattern: &str) -> SearcherTester {
SearcherTester {
haystack: haystack.to_string(),
pattern: pattern.to_string(),
filter: None,
print_labels: false,
expected_no_line_number: None,
expected_with_line_number: None,
expected_slice_no_line_number: None,
expected_slice_with_line_number: None,
by_line: true,
multi_line: true,
invert_match: false,
line_number: true,
binary: BinaryDetection::none(),
auto_heap_limit: true,
after_context: 0,
before_context: 0,
passthru: false,
}
}
/// Execute the test. If the test succeeds, then this returns successfully.
/// If the test fails, then it panics with an informative message.
pub fn test(&self) {
// Check for configuration errors.
if self.expected_no_line_number.is_none() {
panic!("an 'expected' string with NO line numbers must be given");
}
if self.line_number && self.expected_with_line_number.is_none() {
panic!("an 'expected' string with line numbers must be given, \
or disable testing with line numbers");
}
let configs = self.configs();
if configs.is_empty() {
panic!("test configuration resulted in nothing being tested");
}
if self.print_labels {
for config in &configs {
let labels = vec![
format!("reader-{}", config.label),
format!("slice-{}", config.label),
];
for label in &labels {
if self.include(label) {
println!("{}", label);
} else {
println!("{} (ignored)", label);
}
}
}
}
for config in &configs {
let label = format!("reader-{}", config.label);
if self.include(&label) {
let got = config.search_reader(&self.haystack);
assert_eq_printed!(config.expected_reader, got, "{}", label);
}
let label = format!("slice-{}", config.label);
if self.include(&label) {
let got = config.search_slice(&self.haystack);
assert_eq_printed!(config.expected_slice, got, "{}", label);
}
}
}
/// Set a regex pattern to filter the tests that are run.
///
/// By default, no filter is present. When a filter is set, only test
/// configurations with a label matching the given pattern will be run.
///
/// This is often useful when debugging tests, e.g., when you want to do
/// printf debugging and only want one particular test configuration to
/// execute.
#[allow(dead_code)]
pub fn filter(&mut self, pattern: &str) -> &mut SearcherTester {
self.filter = Some(::regex::Regex::new(pattern).unwrap());
self
}
/// When set, the labels for all test configurations are printed before
/// executing any test.
///
/// Note that in order to see these in tests that aren't failing, you'll
/// want to use `cargo test -- --nocapture`.
#[allow(dead_code)]
pub fn print_labels(&mut self, yes: bool) -> &mut SearcherTester {
self.print_labels = yes;
self
}
/// Set the expected search results, without line numbers.
pub fn expected_no_line_number(
&mut self,
exp: &str,
) -> &mut SearcherTester {
self.expected_no_line_number = Some(exp.to_string());
self
}
/// Set the expected search results, with line numbers.
pub fn expected_with_line_number(
&mut self,
exp: &str,
) -> &mut SearcherTester {
self.expected_with_line_number = Some(exp.to_string());
self
}
/// Set the expected search results, without line numbers, when performing
/// a search on a slice. When not present, `expected_no_line_number` is
/// used instead.
pub fn expected_slice_no_line_number(
&mut self,
exp: &str,
) -> &mut SearcherTester {
self.expected_slice_no_line_number = Some(exp.to_string());
self
}
/// Set the expected search results, with line numbers, when performing a
/// search on a slice. When not present, `expected_with_line_number` is
/// used instead.
#[allow(dead_code)]
pub fn expected_slice_with_line_number(
&mut self,
exp: &str,
) -> &mut SearcherTester {
self.expected_slice_with_line_number = Some(exp.to_string());
self
}
/// Whether to test search with line numbers or not.
///
/// This is enabled by default. When enabled, the string that is expected
/// when line numbers are present must be provided. Otherwise, the expected
/// string isn't required.
pub fn line_number(&mut self, yes: bool) -> &mut SearcherTester {
self.line_number = yes;
self
}
/// Whether to test search using the line-by-line searcher or not.
///
/// By default, this is enabled.
pub fn by_line(&mut self, yes: bool) -> &mut SearcherTester {
self.by_line = yes;
self
}
/// Whether to test search using the multi line searcher or not.
///
/// By default, this is enabled.
#[allow(dead_code)]
pub fn multi_line(&mut self, yes: bool) -> &mut SearcherTester {
self.multi_line = yes;
self
}
/// Whether to perform an inverted search or not.
///
/// By default, this is disabled.
pub fn invert_match(&mut self, yes: bool) -> &mut SearcherTester {
self.invert_match = yes;
self
}
/// Whether to enable binary detection on all searches.
///
/// By default, this is disabled.
pub fn binary_detection(
&mut self,
detection: BinaryDetection,
) -> &mut SearcherTester {
self.binary = detection;
self
}
/// Whether to automatically attempt to test the heap limit setting or not.
///
/// By default, one of the test configurations includes setting the heap
/// limit to its minimal value for normal operation, which checks that
/// everything works even at the extremes. However, in some cases, the heap
/// limit can (expectedly) alter the output slightly. For example, it can
/// impact the number of bytes searched when performing binary detection.
/// For convenience, it can be useful to disable the automatic heap limit
/// test.
pub fn auto_heap_limit(&mut self, yes: bool) -> &mut SearcherTester {
self.auto_heap_limit = yes;
self
}
/// Set the number of lines to include in the "after" context.
///
/// The default is `0`, which is equivalent to not printing any context.
pub fn after_context(&mut self, lines: usize) -> &mut SearcherTester {
self.after_context = lines;
self
}
/// Set the number of lines to include in the "before" context.
///
/// The default is `0`, which is equivalent to not printing any context.
pub fn before_context(&mut self, lines: usize) -> &mut SearcherTester {
self.before_context = lines;
self
}
/// Whether to enable the "passthru" feature or not.
///
/// When passthru is enabled, it effectively treats all non-matching lines
/// as contextual lines. In other words, enabling this is akin to
/// requesting an unbounded number of before and after contextual lines.
///
/// This is disabled by default.
pub fn passthru(&mut self, yes: bool) -> &mut SearcherTester {
self.passthru = yes;
self
}
/// Return the minimum size of a buffer required for a successful search.
///
/// Generally, this corresponds to the maximum length of a line (including
/// its terminator), but if context settings are enabled, then this must
/// include the sum of the longest N lines.
///
/// Note that this must account for whether the test is using multi line
/// search or not, since multi line search requires being able to fit the
/// entire haystack into memory.
fn minimal_heap_limit(&self, multi_line: bool) -> usize {
if multi_line {
1 + self.haystack.len()
} else if self.before_context == 0 && self.after_context == 0 {
1 + self.haystack.lines().map(|s| s.len()).max().unwrap_or(0)
} else {
let mut lens: Vec<usize> =
self.haystack.lines().map(|s| s.len()).collect();
lens.sort();
lens.reverse();
let context_count =
if self.passthru {
self.haystack.lines().count()
} else {
// Why do we add 2 here? Well, we need to add 1 in order to
// have room to search at least one line. We add another
// because the implementation will occasionally include
// an additional line when handling the context. There's
// no particularly good reason, other than keeping the
// implementation simple.
2 + self.before_context + self.after_context
};
// We add 1 to each line since `str::lines` doesn't include the
// line terminator.
lens.into_iter()
.take(context_count)
.map(|len| len + 1)
.sum::<usize>()
}
}
/// Returns true if and only if the given label should be included as part
/// of executing `test`.
///
/// Inclusion is determined by the filter specified. If no filter has been
/// given, then this always returns `true`.
fn include(&self, label: &str) -> bool {
let re = match self.filter {
None => return true,
Some(ref re) => re,
};
re.is_match(label)
}
/// Configs generates a set of all search configurations that should be
/// tested. The configs generated are based on the configuration in this
/// builder.
fn configs(&self) -> Vec<TesterConfig> {
let mut configs = vec![];
let matcher = RegexMatcher::new(&self.pattern);
let mut builder = SearcherBuilder::new();
builder
.line_number(false)
.invert_match(self.invert_match)
.binary_detection(self.binary.clone())
.after_context(self.after_context)
.before_context(self.before_context)
.passthru(self.passthru);
if self.by_line {
let mut matcher = matcher.clone();
let mut builder = builder.clone();
let expected_reader =
self.expected_no_line_number.as_ref().unwrap().to_string();
let expected_slice = match self.expected_slice_no_line_number {
None => expected_reader.clone(),
Some(ref e) => e.to_string(),
};
configs.push(TesterConfig {
label: "byline-noterm-nonumber".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
if self.auto_heap_limit {
builder.heap_limit(Some(self.minimal_heap_limit(false)));
configs.push(TesterConfig {
label: "byline-noterm-nonumber-heaplimit".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
builder.heap_limit(None);
}
matcher.set_line_term(Some(LineTerminator::byte(b'\n')));
configs.push(TesterConfig {
label: "byline-term-nonumber".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
matcher.every_line_is_candidate(true);
configs.push(TesterConfig {
label: "byline-term-nonumber-candidates".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
}
if self.by_line && self.line_number {
let mut matcher = matcher.clone();
let mut builder = builder.clone();
let expected_reader =
self.expected_with_line_number.as_ref().unwrap().to_string();
let expected_slice = match self.expected_slice_with_line_number {
None => expected_reader.clone(),
Some(ref e) => e.to_string(),
};
builder.line_number(true);
configs.push(TesterConfig {
label: "byline-noterm-number".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
matcher.set_line_term(Some(LineTerminator::byte(b'\n')));
configs.push(TesterConfig {
label: "byline-term-number".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
matcher.every_line_is_candidate(true);
configs.push(TesterConfig {
label: "byline-term-number-candidates".to_string(),
expected_reader: expected_reader.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
}
if self.multi_line {
let mut builder = builder.clone();
let expected_slice = match self.expected_slice_no_line_number {
None => {
self.expected_no_line_number.as_ref().unwrap().to_string()
}
Some(ref e) => e.to_string(),
};
builder.multi_line(true);
configs.push(TesterConfig {
label: "multiline-nonumber".to_string(),
expected_reader: expected_slice.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
if self.auto_heap_limit {
builder.heap_limit(Some(self.minimal_heap_limit(true)));
configs.push(TesterConfig {
label: "multiline-nonumber-heaplimit".to_string(),
expected_reader: expected_slice.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
builder.heap_limit(None);
}
}
if self.multi_line && self.line_number {
let mut builder = builder.clone();
let expected_slice = match self.expected_slice_with_line_number {
None => {
self.expected_with_line_number
.as_ref().unwrap().to_string()
}
Some(ref e) => e.to_string(),
};
builder.multi_line(true);
builder.line_number(true);
configs.push(TesterConfig {
label: "multiline-number".to_string(),
expected_reader: expected_slice.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
builder.heap_limit(Some(self.minimal_heap_limit(true)));
configs.push(TesterConfig {
label: "multiline-number-heaplimit".to_string(),
expected_reader: expected_slice.clone(),
expected_slice: expected_slice.clone(),
builder: builder.clone(),
matcher: matcher.clone(),
});
builder.heap_limit(None);
}
configs
}
}
#[derive(Debug)]
struct TesterConfig {
label: String,
expected_reader: String,
expected_slice: String,
builder: SearcherBuilder,
matcher: RegexMatcher,
}
impl TesterConfig {
/// Execute a search using a reader. This exercises the incremental search
/// strategy, where the entire contents of the corpus aren't necessarily
/// in memory at once.
fn search_reader(&self, haystack: &str) -> String {
let mut sink = KitchenSink::new();
let mut searcher = self.builder.build();
let result = searcher.search_reader(
&self.matcher,
haystack.as_bytes(),
&mut sink,
);
if let Err(err) = result {
let label = format!("reader-{}", self.label);
panic!("error running '{}': {}", label, err);
}
String::from_utf8(sink.as_bytes().to_vec()).unwrap()
}
/// Execute a search using a slice. This exercises the search routines that
/// have the entire contents of the corpus in memory at one time.
fn search_slice(&self, haystack: &str) -> String {
let mut sink = KitchenSink::new();
let mut searcher = self.builder.build();
let result = searcher.search_slice(
&self.matcher,
haystack.as_bytes(),
&mut sink,
);
if let Err(err) = result {
let label = format!("slice-{}", self.label);
panic!("error running '{}': {}", label, err);
}
String::from_utf8(sink.as_bytes().to_vec()).unwrap()
}
}
#[cfg(test)]
mod tests {
use grep_matcher::{Match, Matcher};
use super::*;
fn m(start: usize, end: usize) -> Match {
Match::new(start, end)
}
#[test]
fn empty_line1() {
let haystack = b"";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(0, 0))));
}
#[test]
fn empty_line2() {
let haystack = b"\n";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(0, 0))));
assert_eq!(matcher.find_at(haystack, 1), Ok(Some(m(1, 1))));
}
#[test]
fn empty_line3() {
let haystack = b"\n\n";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(0, 0))));
assert_eq!(matcher.find_at(haystack, 1), Ok(Some(m(1, 1))));
assert_eq!(matcher.find_at(haystack, 2), Ok(Some(m(2, 2))));
}
#[test]
fn empty_line4() {
let haystack = b"a\n\nb\n";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 1), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 2), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 3), Ok(Some(m(5, 5))));
assert_eq!(matcher.find_at(haystack, 4), Ok(Some(m(5, 5))));
assert_eq!(matcher.find_at(haystack, 5), Ok(Some(m(5, 5))));
}
#[test]
fn empty_line5() {
let haystack = b"a\n\nb\nc";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 1), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 2), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 3), Ok(None));
assert_eq!(matcher.find_at(haystack, 4), Ok(None));
assert_eq!(matcher.find_at(haystack, 5), Ok(None));
assert_eq!(matcher.find_at(haystack, 6), Ok(None));
}
#[test]
fn empty_line6() {
let haystack = b"a\n";
let matcher = RegexMatcher::new(r"^$");
assert_eq!(matcher.find_at(haystack, 0), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 1), Ok(Some(m(2, 2))));
assert_eq!(matcher.find_at(haystack, 2), Ok(Some(m(2, 2))));
}
}