gst-plugins-rs/text/json/src/line_reader.rs

324 lines
11 KiB
Rust

// Copyright (C) 2018 Sebastian Dröge <sebastian@centricular.com>
//
// This Source Code Form is subject to the terms of the Mozilla Public License, v2.0.
// If a copy of the MPL was not distributed with this file, You can obtain one at
// <https://mozilla.org/MPL/2.0/>.
//
// SPDX-License-Identifier: MPL-2.0
use std::collections::VecDeque;
#[derive(Debug)]
pub struct LineReader<T: AsRef<[u8]>> {
queue: VecDeque<T>,
// Read position into the queue in bytes
read_pos: usize,
// Offset into queue where we have to look for a newline
// All previous items don't contain a newline
search_pos: usize,
buf: Vec<u8>,
}
impl<T: AsRef<[u8]>> LineReader<T> {
pub fn new() -> LineReader<T> {
Self {
queue: VecDeque::new(),
read_pos: 0,
search_pos: 0,
buf: Vec::new(),
}
}
pub fn push(&mut self, b: T) {
self.queue.push_back(b);
}
/// Drops everything from the internal queue that was previously returned, i.e.
/// if previously a line was returned we drop this whole line so that we can
/// proceed with the next line
fn drop_previous_line(&mut self) {
// Drop everything we read the last time now
while self.read_pos > 0
&& self.read_pos >= self.queue.front().map(|f| f.as_ref().len()).unwrap_or(0)
{
self.read_pos -= self.queue.front().map(|f| f.as_ref().len()).unwrap_or(0);
self.queue.pop_front();
if self.search_pos > 0 {
self.search_pos -= 1;
}
}
self.buf.clear();
}
#[allow(unused)]
pub fn line_or_drain(&mut self) -> Option<&[u8]> {
self.line_with_drain(true)
}
#[allow(unused)]
pub fn line(&mut self) -> Option<&[u8]> {
self.line_with_drain(false)
}
/// Searches the first '\n' in the currently queued buffers and returns the index in buffers
/// inside the queue and the index in bytes from the beginning of the queue, or None.
///
/// Also updates the search_pos so that we don't look again in the previous buffers on the next
/// call
fn find_newline(&mut self) -> Option<(usize, usize)> {
let mut offset = 0;
for (idx, buf) in self.queue.iter().enumerate() {
let buf = buf.as_ref();
// Fast skip-ahead
if idx < self.search_pos {
offset += buf.len();
continue;
}
let pos = buf
.iter()
.enumerate()
.skip(if idx == 0 { self.read_pos } else { 0 })
.find(|(_, b)| **b == b'\n')
.map(|(idx, _)| idx);
if let Some(pos) = pos {
// On the next call we have to search in this buffer again
// as it might contain a second newline
self.search_pos = idx;
return Some((idx, offset + pos + 1));
}
// This buffer did not contain a newline so we don't have to look
// in it again next time
self.search_pos = idx + 1;
offset += buf.len();
}
None
}
/// Copies length bytes from all buffers from the beginning until last_idx into our internal
/// buffer, and skips the first offset bytes from the first buffer.
fn copy_into_internal_buffer(&mut self, last_idx: usize, offset: usize, len: usize) {
// Reserve space for the whole line beforehand
if self.buf.capacity() < len {
self.buf.reserve(len - self.buf.capacity());
}
// Then iterate over all buffers inside the queue until the one that contains
// the newline character
for (idx, buf) in self.queue.iter().enumerate().take(last_idx + 1) {
let buf = buf.as_ref();
// Calculate how much data we still have to copy
let rem = len - self.buf.len();
assert!(rem > 0);
// For the first index we need to take into account the offset. The first
// bytes might have to be skipped and as such we have fewer bytes available
// than the whole length of the buffer
let buf_len = if idx == 0 {
assert!(offset < buf.len());
buf.len() - offset
} else {
buf.len()
};
// Calculate how much we can copy from this buffer. At most the size of the buffer
// itself, but never more than the amount we still have to copy overall
let copy_len = if rem > buf_len { buf_len } else { rem };
assert!(copy_len > 0);
if idx == 0 {
self.buf
.extend_from_slice(&buf[offset..(offset + copy_len)]);
} else {
self.buf.extend_from_slice(&buf[..copy_len]);
}
}
assert_eq!(self.buf.len(), len);
}
pub fn line_with_drain(&mut self, drain: bool) -> Option<&[u8]> {
// Drop all data from the previous line
self.drop_previous_line();
// read_pos must always be inside the first buffer of our queue here
// or otherwise we went into an inconsistent state: the first buffer(s)
// would've had to be dropped above then as they are not relevant anymore
assert!(
self.read_pos == 0
|| self.read_pos < self.queue.front().map(|f| f.as_ref().len()).unwrap_or(0)
);
// Find the next newline character from our previous position
if let Some((idx, pos)) = self.find_newline() {
// We now need to copy everything from the old read_pos to the new
// pos, and on the next call we have to start from the new pos
let old_read_pos = self.read_pos;
self.read_pos = pos;
assert!(self.read_pos > old_read_pos);
assert!(idx < self.queue.len());
// If the newline is found in the first buffer in our queue, we can directly return
// the slice from it without doing any copying.
//
// On average this should be the most common case.
if idx == 0 {
let buf = self.queue.front().unwrap().as_ref();
return Some(&buf[old_read_pos..self.read_pos]);
} else {
// Copy into our internal buffer as the current line spans multiple buffers
let len = self.read_pos - old_read_pos;
self.copy_into_internal_buffer(idx, old_read_pos, len);
return Some(&self.buf[0..len]);
}
}
// No newline found above and we're not draining, so let's wait until
// more data is available that might contain a newline character
if !drain {
return None;
}
if self.queue.is_empty() {
return None;
}
// When draining and we only have a single buffer in the queue we can
// directly return a slice into it
if self.queue.len() == 1 {
let res = &self.queue.front().unwrap().as_ref()[self.read_pos..];
self.read_pos += res.len();
self.search_pos = 1;
return Some(res);
}
// Otherwise we have to copy everything that is remaining into our
// internal buffer and then return a slice from that
let len = self.queue.iter().map(|v| v.as_ref().len()).sum();
if self.buf.capacity() < len {
self.buf.reserve(len - self.buf.capacity());
}
for (idx, ref v) in self.queue.iter().enumerate() {
if idx == 0 {
self.buf.extend_from_slice(&v.as_ref()[self.read_pos..]);
} else {
self.buf.extend_from_slice(v.as_ref());
}
}
self.read_pos += self.buf.len();
self.search_pos = self.queue.len();
Some(self.buf.as_ref())
}
pub fn clear(&mut self) {
self.queue.clear();
self.read_pos = 0;
self.search_pos = 0;
self.buf.clear();
}
}
#[cfg(test)]
mod tests {
use super::LineReader;
#[test]
fn test_single_buffer() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nefgh\nijkl\n".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), Some(b"ijkl\n".as_ref()));
assert_eq!(r.line(), None);
}
#[test]
fn test_empty_line() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nefgh\n\nijkl\n".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), Some(b"\n".as_ref()));
assert_eq!(r.line(), Some(b"ijkl\n".as_ref()));
assert_eq!(r.line(), None);
}
#[test]
fn test_multi_buffer_split() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nef".as_ref()));
r.push(Vec::from(b"gh\nijkl\n".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), Some(b"ijkl\n".as_ref()));
assert_eq!(r.line(), None);
}
#[test]
fn test_multi_buffer_split_2() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\ne".as_ref()));
r.push(Vec::from(b"f".as_ref()));
r.push(Vec::from(b"g".as_ref()));
r.push(Vec::from(b"h\nijkl\n".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), Some(b"ijkl\n".as_ref()));
assert_eq!(r.line(), None);
}
#[test]
fn test_single_buffer_drain() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nefgh\nijkl".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), None);
assert_eq!(r.line_or_drain(), Some(b"ijkl".as_ref()));
assert_eq!(r.line_or_drain(), None);
}
#[test]
fn test_single_buffer_drain_multi_line() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nefgh\n".as_ref()));
r.push(Vec::from(b"ijkl".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), None);
assert_eq!(r.line_or_drain(), Some(b"ijkl".as_ref()));
assert_eq!(r.line_or_drain(), None);
}
#[test]
fn test_single_buffer_drain_multi_line_2() {
let mut r = LineReader::new();
r.push(Vec::from(b"abcd\nefgh\ni".as_ref()));
r.push(Vec::from(b"j".as_ref()));
r.push(Vec::from(b"k".as_ref()));
r.push(Vec::from(b"l".as_ref()));
assert_eq!(r.line(), Some(b"abcd\n".as_ref()));
assert_eq!(r.line(), Some(b"efgh\n".as_ref()));
assert_eq!(r.line(), None);
assert_eq!(r.line_or_drain(), Some(b"ijkl".as_ref()));
assert_eq!(r.line_or_drain(), None);
}
}