From 0f20660cb95abd4f407bdafa2c45e01829fa971f Mon Sep 17 00:00:00 2001
From: Titus Wormer <tituswormer@gmail.com>
Date: Tue, 19 Jul 2022 17:56:57 +0200
Subject: Refactor to remove cloning in `edit_map`
---
src/construct/attention.rs | 2 +-
src/construct/heading_atx.rs | 2 +-
src/construct/heading_setext.rs | 2 +-
src/construct/label_end.rs | 17 +++++++---------
src/construct/list.rs | 8 +++-----
src/construct/paragraph.rs | 2 +-
src/construct/partial_data.rs | 2 +-
src/content/document.rs | 8 ++++----
src/subtokenize.rs | 4 ++--
src/tokenizer.rs | 1 +
src/util/edit_map.rs | 44 +++++++++++++++++++++++------------------
11 files changed, 47 insertions(+), 45 deletions(-)
(limited to 'src')
diff --git a/src/construct/attention.rs b/src/construct/attention.rs
index 3e15f9a..7e99600 100644
--- a/src/construct/attention.rs
+++ b/src/construct/attention.rs
@@ -523,7 +523,7 @@ fn resolve_attention(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
/// Classify whether a character code represents whitespace, punctuation, or
diff --git a/src/construct/heading_atx.rs b/src/construct/heading_atx.rs
index 68a6be7..4546924 100644
--- a/src/construct/heading_atx.rs
+++ b/src/construct/heading_atx.rs
@@ -258,5 +258,5 @@ pub fn resolve(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/construct/heading_setext.rs b/src/construct/heading_setext.rs
index 7cd259b..841bf53 100644
--- a/src/construct/heading_setext.rs
+++ b/src/construct/heading_setext.rs
@@ -236,5 +236,5 @@ pub fn resolve(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/construct/label_end.rs b/src/construct/label_end.rs
index 2ac2500..2124681 100644
--- a/src/construct/label_end.rs
+++ b/src/construct/label_end.rs
@@ -154,9 +154,7 @@ use crate::construct::{
partial_title::{start as title, Options as TitleOptions},
};
use crate::token::Token;
-use crate::tokenizer::{
- Code, Event, EventType, LabelStart, Media, State, StateFnResult, Tokenizer,
-};
+use crate::tokenizer::{Code, Event, EventType, Media, State, StateFnResult, Tokenizer};
use crate::util::{
edit_map::EditMap,
normalize_identifier::normalize_identifier,
@@ -325,10 +323,9 @@ fn reference_not_full(tokenizer: &mut Tokenizer, code: Code, info: Info) -> Stat
/// ```
fn ok(tokenizer: &mut Tokenizer, code: Code, mut info: Info) -> StateFnResult {
// Remove this one and everything after it.
- let mut left: Vec<LabelStart> = tokenizer
+ let mut left = tokenizer
.label_start_stack
- .drain(info.label_start_index..)
- .collect();
+ .split_off(info.label_start_index);
// Remove this one from `left`, as we’ll move it to `media_list`.
left.remove(0);
tokenizer.label_start_list_loose.append(&mut left);
@@ -616,9 +613,9 @@ fn collapsed_reference_open(tokenizer: &mut Tokenizer, code: Code) -> StateFnRes
/// images, or turns them back into data.
#[allow(clippy::too_many_lines)]
pub fn resolve_media(tokenizer: &mut Tokenizer) -> Vec<Event> {
- let mut left: Vec<LabelStart> = tokenizer.label_start_list_loose.drain(..).collect();
- let mut left_2: Vec<LabelStart> = tokenizer.label_start_stack.drain(..).collect();
- let media: Vec<Media> = tokenizer.media_list.drain(..).collect();
+ let mut left = tokenizer.label_start_list_loose.split_off(0);
+ let mut left_2 = tokenizer.label_start_stack.split_off(0);
+ let media = tokenizer.media_list.split_off(0);
left.append(&mut left_2);
let mut edit_map = EditMap::new();
@@ -776,5 +773,5 @@ pub fn resolve_media(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/construct/list.rs b/src/construct/list.rs
index 9cd3b62..db8af36 100644
--- a/src/construct/list.rs
+++ b/src/construct/list.rs
@@ -432,8 +432,7 @@ pub fn resolve_list_item(tokenizer: &mut Tokenizer) -> Vec<Event> {
if previous.0 == current.0 && previous.1 == current.1 && before == current.2 {
let previous_mut = &mut lists_wip[list_index];
previous_mut.3 = current.3;
- let mut remainder = lists_wip.drain((list_index + 1)..).collect::<Vec<_>>();
- lists.append(&mut remainder);
+ lists.append(&mut lists_wip.split_off(list_index + 1));
matched = true;
break;
}
@@ -457,8 +456,7 @@ pub fn resolve_list_item(tokenizer: &mut Tokenizer) -> Vec<Event> {
}
if let Some(exit) = exit {
- let mut remainder = lists_wip.drain(exit..).collect::<Vec<_>>();
- lists.append(&mut remainder);
+ lists.append(&mut lists_wip.split_off(exit));
}
lists_wip.push(current);
@@ -494,5 +492,5 @@ pub fn resolve_list_item(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/construct/paragraph.rs b/src/construct/paragraph.rs
index 74dca87..53030f4 100644
--- a/src/construct/paragraph.rs
+++ b/src/construct/paragraph.rs
@@ -142,5 +142,5 @@ pub fn resolve(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/construct/partial_data.rs b/src/construct/partial_data.rs
index 3bc8c1c..b59bb76 100644
--- a/src/construct/partial_data.rs
+++ b/src/construct/partial_data.rs
@@ -114,5 +114,5 @@ pub fn resolve_data(tokenizer: &mut Tokenizer) -> Vec<Event> {
index += 1;
}
- edit_map.consume(&mut tokenizer.events)
+ edit_map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/content/document.rs b/src/content/document.rs
index 63c2476..53e58c4 100644
--- a/src/content/document.rs
+++ b/src/content/document.rs
@@ -352,7 +352,7 @@ fn containers_after(
.last_mut()
.unwrap()
.0
- .append(&mut tokenizer.events.drain(info.index..).collect::<Vec<_>>());
+ .append(&mut tokenizer.events.split_off(info.index));
tokenizer.lazy = info.continued != info.stack.len();
tokenizer.interrupt = info.interrupt_before;
@@ -433,12 +433,12 @@ fn exit_containers(
mut info: DocumentInfo,
phase: &Phase,
) -> DocumentInfo {
- let mut stack_close = info.stack.drain(info.continued..).collect::<Vec<_>>();
+ let mut stack_close = info.stack.split_off(info.continued);
// So, we’re at the end of a line, but we need to close the *previous* line.
if *phase != Phase::Eof {
tokenizer.define_skip_current();
- let mut current_events = tokenizer.events.drain(info.index..).collect::<Vec<_>>();
+ let mut current_events = tokenizer.events.split_off(info.index);
let next = info.next;
info.next = Box::new(flow); // This is weird but Rust needs a function there.
let result = tokenizer.flush(next);
@@ -537,5 +537,5 @@ fn resolve(tokenizer: &mut Tokenizer, info: &DocumentInfo) -> Vec<Event> {
add,
);
- map.consume(&mut tokenizer.events)
+ map.consume(tokenizer.events.split_off(0))
}
diff --git a/src/subtokenize.rs b/src/subtokenize.rs
index 174ddfe..df7b015 100644
--- a/src/subtokenize.rs
+++ b/src/subtokenize.rs
@@ -61,7 +61,7 @@ pub fn link_to(events: &mut [Event], pevious: usize, next: usize) {
/// Parse linked events.
///
/// Supposed to be called repeatedly, returns `1: true` when done.
-pub fn subtokenize(mut events: Vec<Event>, parse_state: &ParseState) -> (Vec<Event>, bool) {
+pub fn subtokenize(events: Vec<Event>, parse_state: &ParseState) -> (Vec<Event>, bool) {
let mut edit_map = EditMap::new();
let mut done = true;
let mut index = 0;
@@ -178,5 +178,5 @@ pub fn subtokenize(mut events: Vec<Event>, parse_state: &ParseState) -> (Vec<Eve
index += 1;
}
- (edit_map.consume(&mut events), done)
+ (edit_map.consume(events), done)
}
diff --git a/src/tokenizer.rs b/src/tokenizer.rs
index 8f85af0..4e184f4 100644
--- a/src/tokenizer.rs
+++ b/src/tokenizer.rs
@@ -730,6 +730,7 @@ fn feed_impl(
}
// Yield to a higher loop.
+ // To do: do not copy?
check_statefn_result((state, Some(codes[index..].to_vec())))
}
diff --git a/src/util/edit_map.rs b/src/util/edit_map.rs
index 90ff483..1f43a3a 100644
--- a/src/util/edit_map.rs
+++ b/src/util/edit_map.rs
@@ -72,11 +72,9 @@ impl EditMap {
add_impl(self, index, remove, add, true);
}
/// Done, change the events.
- pub fn consume(&mut self, events: &mut [Event]) -> Vec<Event> {
+ pub fn consume(&mut self, mut events: Vec<Event>) -> Vec<Event> {
self.map
.sort_unstable_by(|a, b| a.0.partial_cmp(&b.0).unwrap());
- let mut next_events: Vec<Event> = vec![];
- let mut start = 0;
assert!(!self.consumed, "cannot consume after consuming");
self.consumed = true;
@@ -94,28 +92,36 @@ impl EditMap {
index += 1;
}
- let mut index = 0;
+ let mut index = self.map.len();
+ let mut vecs: Vec<Vec<Event>> = vec![];
+ let mut capacity = 0;
- while index < self.map.len() {
+ while index > 0 {
+ index -= 1;
let at = self.map[index].0;
- let remove = self.map[index].1;
- let mut add = self.map[index].2.drain(..).collect::<Vec<_>>();
- if start < at {
- let append = &mut events[start..at].to_vec();
- shift_links(append, &jumps);
- next_events.append(append);
- }
+ let mut keep = events.split_off(at + self.map[index].1);
+ shift_links(&mut keep, &jumps);
+ capacity += keep.len();
+ vecs.push(keep);
- next_events.append(&mut add);
- start = at + remove;
- index += 1;
+ let add = self.map[index].2.split_off(0);
+ capacity += add.len();
+ vecs.push(add);
+
+ events.truncate(at);
}
- if start < events.len() {
- let append = &mut events[start..].to_vec();
- shift_links(append, &jumps);
- next_events.append(append);
+ shift_links(&mut events, &jumps);
+ capacity += events.len();
+ vecs.push(events);
+
+ let mut next_events: Vec<Event> = Vec::with_capacity(capacity);
+ let mut slice = vecs.pop();
+
+ while let Some(mut x) = slice {
+ next_events.append(&mut x);
+ slice = vecs.pop();
}
next_events
--
cgit