refactor: simplify WordWrapper implementation (#1193)

This commit is contained in:
tranzystorekk 2024-06-25 21:14:32 +02:00 committed by GitHub
parent 36d49e549b
commit 32a0b26525
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
2 changed files with 143 additions and 149 deletions

View file

@ -1,5 +1,8 @@
use crate::{prelude::*, style::Styled};
const NBSP: &str = "\u{00a0}";
const ZWSP: &str = "\u{200b}";
/// A grapheme associated to a style.
/// Note that, although `StyledGrapheme` is the smallest divisible unit of text,
/// it actually is not a member of the text type hierarchy (`Text` -> `Line` -> `Span`).
@ -22,6 +25,11 @@ impl<'a> StyledGrapheme<'a> {
style: style.into(),
}
}
pub(crate) fn is_whitespace(&self) -> bool {
let symbol = self.symbol;
symbol == ZWSP || symbol.chars().all(char::is_whitespace) && symbol != NBSP
}
}
impl<'a> Styled for StyledGrapheme<'a> {

View file

@ -5,9 +5,6 @@ use unicode_width::UnicodeWidthStr;
use crate::{layout::Alignment, text::StyledGrapheme};
const NBSP: &str = "\u{00a0}";
const ZWSP: &str = "\u{200b}";
/// A state machine to pack styled symbols into lines.
/// Cannot implement it as Iterator since it yields slices of the internal buffer (need streaming
/// iterators for that).
@ -59,6 +56,124 @@ where
trim,
}
}
fn next_cached_line(&mut self) -> Option<Vec<StyledGrapheme<'a>>> {
self.wrapped_lines.as_mut()?.next()
}
/// Split an input line (`line_symbols`) into wrapped lines
/// and cache them to be emitted later
fn process_input(&mut self, line_symbols: impl IntoIterator<Item = StyledGrapheme<'a>>) {
let mut result_lines = vec![];
let mut pending_line = vec![];
let mut line_width = 0;
let mut pending_word = vec![];
let mut word_width = 0;
let mut pending_whitespace: VecDeque<StyledGrapheme> = VecDeque::new();
let mut whitespace_width = 0;
let mut non_whitespace_previous = false;
for grapheme in line_symbols {
let is_whitespace = grapheme.is_whitespace();
let symbol_width = grapheme.symbol.width() as u16;
// ignore symbols wider than line limit
if symbol_width > self.max_line_width {
continue;
}
let word_found = non_whitespace_previous && is_whitespace;
// current word would overflow after removing whitespace
let trimmed_overflow = pending_line.is_empty()
&& self.trim
&& word_width + symbol_width > self.max_line_width;
// separated whitespace would overflow on its own
let whitespace_overflow = pending_line.is_empty()
&& self.trim
&& whitespace_width + symbol_width > self.max_line_width;
// current full word (including whitespace) would overflow
let untrimmed_overflow = pending_line.is_empty()
&& !self.trim
&& word_width + whitespace_width + symbol_width > self.max_line_width;
// append finished segment to current line
if word_found || trimmed_overflow || whitespace_overflow || untrimmed_overflow {
if !pending_line.is_empty() || !self.trim {
pending_line.extend(pending_whitespace.drain(..));
line_width += whitespace_width;
}
pending_line.append(&mut pending_word);
line_width += word_width;
pending_whitespace.clear();
whitespace_width = 0;
word_width = 0;
}
// pending line fills up limit
let line_full = line_width >= self.max_line_width;
// pending word would overflow line limit
let pending_word_overflow = symbol_width > 0
&& line_width + whitespace_width + word_width >= self.max_line_width;
// add finished wrapped line to remaining lines
if line_full || pending_word_overflow {
let mut remaining_width = u16::saturating_sub(self.max_line_width, line_width);
result_lines.push(std::mem::take(&mut pending_line));
line_width = 0;
// remove whitespace up to the end of line
while let Some(grapheme) = pending_whitespace.front() {
let width = grapheme.symbol.width() as u16;
if width > remaining_width {
break;
}
whitespace_width -= width;
remaining_width -= width;
pending_whitespace.pop_front();
}
// don't count first whitespace toward next word
if is_whitespace && pending_whitespace.is_empty() {
continue;
}
}
// append symbol to a pending buffer
if is_whitespace {
whitespace_width += symbol_width;
pending_whitespace.push_back(grapheme);
} else {
word_width += symbol_width;
pending_word.push(grapheme);
}
non_whitespace_previous = !is_whitespace;
}
// append remaining text parts
if pending_line.is_empty() && pending_word.is_empty() && !pending_whitespace.is_empty() {
result_lines.push(vec![]);
}
if !pending_line.is_empty() || !self.trim {
pending_line.extend(pending_whitespace);
}
pending_line.extend(pending_word);
if !pending_line.is_empty() {
result_lines.push(pending_line);
}
if result_lines.is_empty() {
result_lines.push(vec![]);
}
// save processed lines for emitting later
self.wrapped_lines = Some(result_lines.into_iter());
}
}
impl<'a, O, I> LineComposer<'a> for WordWrapper<'a, O, I>
@ -72,155 +187,26 @@ where
return None;
}
let mut current_line: Option<Vec<StyledGrapheme<'a>>> = None;
let mut line_width: u16 = 0;
// Try to repeatedly retrieve next line
while current_line.is_none() {
// Retrieve next preprocessed wrapped line
if let Some(line_iterator) = &mut self.wrapped_lines {
if let Some(line) = line_iterator.next() {
line_width = line
loop {
// emit next cached line if present
if let Some(line) = self.next_cached_line() {
let line_width = line
.iter()
.map(|grapheme| grapheme.symbol.width())
.sum::<usize>() as u16;
current_line = Some(line);
}
}
.map(|grapheme| grapheme.symbol.width() as u16)
.sum();
// When no more preprocessed wrapped lines
if current_line.is_none() {
// Try to calculate next wrapped lines based on current whole line
if let Some((line_symbols, line_alignment)) = &mut self.input_lines.next() {
// Save the whole line's alignment
self.current_alignment = *line_alignment;
let mut wrapped_lines = vec![]; // Saves the wrapped lines
// Saves the unfinished wrapped line
let (mut current_line, mut current_line_width) = (vec![], 0);
// Saves the partially processed word
let (mut unfinished_word, mut word_width) = (vec![], 0);
// Saves the whitespaces of the partially unfinished word
let (mut unfinished_whitespaces, mut whitespace_width) =
(VecDeque::<StyledGrapheme>::new(), 0);
let mut has_seen_non_whitespace = false;
for StyledGrapheme { symbol, style } in line_symbols {
let symbol_whitespace = symbol == ZWSP
|| (symbol.chars().all(&char::is_whitespace) && symbol != NBSP);
let symbol_width = symbol.width() as u16;
// Ignore characters wider than the total max width
if symbol_width > self.max_line_width {
continue;
}
// Append finished word to current line
if has_seen_non_whitespace && symbol_whitespace
// Append if trimmed (whitespaces removed) word would overflow
|| word_width + symbol_width > self.max_line_width && current_line.is_empty() && self.trim
// Append if removed whitespace would overflow -> reset whitespace counting to prevent overflow
|| whitespace_width + symbol_width > self.max_line_width && current_line.is_empty() && self.trim
// Append if complete word would overflow
|| word_width + whitespace_width + symbol_width > self.max_line_width && current_line.is_empty() && !self.trim
{
if !current_line.is_empty() || !self.trim {
// Also append whitespaces if not trimming or current line is not
// empty
current_line.extend(
std::mem::take(&mut unfinished_whitespaces).into_iter(),
);
current_line_width += whitespace_width;
}
// Append trimmed word
current_line.append(&mut unfinished_word);
current_line_width += word_width;
// Clear whitespace buffer
unfinished_whitespaces.clear();
whitespace_width = 0;
word_width = 0;
}
// Append the unfinished wrapped line to wrapped lines if it is as wide as
// max line width
if current_line_width >= self.max_line_width
// or if it would be too long with the current partially processed word added
|| current_line_width + whitespace_width + word_width >= self.max_line_width && symbol_width > 0
{
let mut remaining_width = (i32::from(self.max_line_width)
- i32::from(current_line_width))
.max(0) as u16;
wrapped_lines.push(std::mem::take(&mut current_line));
current_line_width = 0;
// Remove all whitespaces till end of just appended wrapped line + next
// whitespace
let mut first_whitespace = unfinished_whitespaces.pop_front();
while let Some(grapheme) = first_whitespace.as_ref() {
let symbol_width = grapheme.symbol.width() as u16;
whitespace_width -= symbol_width;
if symbol_width > remaining_width {
break;
}
remaining_width -= symbol_width;
first_whitespace = unfinished_whitespaces.pop_front();
}
// In case all whitespaces have been exhausted
if symbol_whitespace && first_whitespace.is_none() {
// Prevent first whitespace to count towards next word
continue;
}
}
// Append symbol to unfinished, partially processed word
if symbol_whitespace {
whitespace_width += symbol_width;
unfinished_whitespaces.push_back(StyledGrapheme { symbol, style });
} else {
word_width += symbol_width;
unfinished_word.push(StyledGrapheme { symbol, style });
}
has_seen_non_whitespace = !symbol_whitespace;
}
// Append remaining text parts
if !unfinished_word.is_empty() || !unfinished_whitespaces.is_empty() {
if current_line.is_empty() && unfinished_word.is_empty() {
wrapped_lines.push(vec![]);
} else if !self.trim || !current_line.is_empty() {
current_line.extend(unfinished_whitespaces.into_iter());
} else {
// TODO: explain why this else branch is ok.
// See clippy::else_if_without_else
}
current_line.append(&mut unfinished_word);
}
if !current_line.is_empty() {
wrapped_lines.push(current_line);
}
if wrapped_lines.is_empty() {
// Append empty line if there was nothing to wrap in the first place
wrapped_lines.push(vec![]);
}
self.wrapped_lines = Some(wrapped_lines.into_iter());
} else {
// No more whole lines available -> stop repeatedly retrieving next wrapped line
break;
}
}
}
if let Some(line) = current_line {
self.current_line = line;
Some(WrappedLine {
return Some(WrappedLine {
line: &self.current_line,
width: line_width,
alignment: self.current_alignment,
})
} else {
None
});
}
// otherwise, process pending wrapped lines from input
let (line_symbols, line_alignment) = self.input_lines.next()?;
self.current_alignment = line_alignment;
self.process_input(line_symbols);
}
}
}