path: root/helix-tui/src/widgets/reflow.rs

use crate::text::StyledGrapheme;
use helix_core::line_ending::str_is_line_ending;
use helix_core::unicode::width::UnicodeWidthStr;
use unicode_segmentation::UnicodeSegmentation;

const NBSP: &str = "\u{00a0}";

/// 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).
pub trait LineComposer<'a> {
    fn next_line(&mut self) -> Option<(&[StyledGrapheme<'a>], u16)>;
}

/// A state machine that wraps lines on word boundaries.
pub struct WordWrapper<'a, 'b> {
    symbols: &'b mut dyn Iterator<Item = StyledGrapheme<'a>>,
    max_line_width: u16,
    current_line: Vec<StyledGrapheme<'a>>,
    next_line: Vec<StyledGrapheme<'a>>,
    /// Removes the leading whitespace from lines
    trim: bool,
}

impl<'a, 'b> WordWrapper<'a, 'b> {
    pub fn new(
        symbols: &'b mut dyn Iterator<Item = StyledGrapheme<'a>>,
        max_line_width: u16,
        trim: bool,
    ) -> WordWrapper<'a, 'b> {
        WordWrapper {
            symbols,
            max_line_width,
            current_line: vec![],
            next_line: vec![],
            trim,
        }
    }
}

impl<'a, 'b> LineComposer<'a> for WordWrapper<'a, 'b> {
    fn next_line(&mut self) -> Option<(&[StyledGrapheme<'a>], u16)> {
        if self.max_line_width == 0 {
            return None;
        }
        std::mem::swap(&mut self.current_line, &mut self.next_line);
        self.next_line.truncate(0);

        let mut current_line_width = self
            .current_line
            .iter()
            .map(|StyledGrapheme { symbol, .. }| symbol.width() as u16)
            .sum();

        let mut symbols_to_last_word_end: usize = 0;
        let mut width_to_last_word_end: u16 = 0;
        let mut prev_whitespace = false;
        let mut symbols_exhausted = true;
        for StyledGrapheme { symbol, style } in &mut self.symbols {
            symbols_exhausted = false;
            let symbol_whitespace = symbol.chars().all(&char::is_whitespace) && symbol != NBSP;

            // Ignore characters wider that the total max width.
            if symbol.width() as u16 > self.max_line_width
                // Skip leading whitespace when trim is enabled.
                || self.trim && symbol_whitespace && !str_is_line_ending(symbol) && current_line_width == 0
            {
                continue;
            }

            // Break on newline and discard it.
            if str_is_line_ending(symbol) {
                if prev_whitespace {
                    current_line_width = width_to_last_word_end;
                    self.current_line.truncate(symbols_to_last_word_end);
                }
                break;
            }

            // Mark the previous symbol as word end.
            if symbol_whitespace && !prev_whitespace {
                symbols_to_last_word_end = self.current_line.len();
                width_to_last_word_end = current_line_width;
            }

            self.current_line.push(StyledGrapheme { symbol, style });
            current_line_width += symbol.width() as u16;

            if current_line_width > self.max_line_width {
                // If there was no word break in the text, wrap at the end of the line.
                let (truncate_at, truncated_width) = if symbols_to_last_word_end != 0 {
                    (symbols_to_last_word_end, width_to_last_word_end)
                } else {
                    (self.current_line.len() - 1, self.max_line_width)
                };

                // Push the remainder to the next line but strip leading whitespace:
                {
                    let remainder = &self.current_line[truncate_at..];
                    if let Some(remainder_nonwhite) =
                        remainder.iter().position(|StyledGrapheme { symbol, .. }| {
                            !symbol.chars().all(&char::is_whitespace)
                        })
                    {
                        self.next_line
                            .extend_from_slice(&remainder[remainder_nonwhite..]);
                    }
                }
                self.current_line.truncate(truncate_at);
                current_line_width = truncated_width;
                break;
            }

            prev_whitespace = symbol_whitespace;
        }

        // Even if the iterator is exhausted, pass the previous remainder.
        if symbols_exhausted && self.current_line.is_empty() {
            None
        } else {
            Some((&self.current_line[..], current_line_width))
        }
    }
}

/// A state machine that truncates overhanging lines.
pub struct LineTruncator<'a, 'b> {
    symbols: &'b mut dyn Iterator<Item = StyledGrapheme<'a>>,
    max_line_width: u16,
    current_line: Vec<StyledGrapheme<'a>>,
    /// Record the offset to skip render
    horizontal_offset: u16,
}

impl<'a, 'b> LineTruncator<'a, 'b> {
    pub fn new(
        symbols: &'b mut dyn Iterator<Item = StyledGrapheme<'a>>,
        max_line_width: u16,
    ) -> LineTruncator<'a, 'b> {
        LineTruncator {
            symbols,
            max_line_width,
            horizontal_offset: 0,
            current_line: vec![],
        }
    }

    pub fn set_horizontal_offset(&mut self, horizontal_offset: u16) {
        self.horizontal_offset = horizontal_offset;
    }
}

impl<'a, 'b> LineComposer<'a> for LineTruncator<'a, 'b> {
    fn next_line(&mut self) -> Option<(&[StyledGrapheme<'a>], u16)> {
        if self.max_line_width == 0 {
            return None;
        }

        self.current_line.truncate(0);
        let mut current_line_width = 0;

        let mut skip_rest = false;
        let mut symbols_exhausted = true;
        let mut horizontal_offset = self.horizontal_offset as usize;
        for StyledGrapheme { symbol, style } in &mut self.symbols {
            symbols_exhausted = false;

            // Ignore characters wider that the total max width.
            if symbol.width() as u16 > self.max_line_width {
                continue;
            }

            // Break on newline and discard it.
            if str_is_line_ending(symbol) {
                break;
            }

            if current_line_width + symbol.width() as u16 > self.max_line_width {
                // Exhaust the remainder of the line.
                skip_rest = true;
                break;
            }

            let symbol = if horizontal_offset == 0 {
                symbol
            } else {
                let w = symbol.width();
                if w > horizontal_offset {
                    let t = trim_offset(symbol, horizontal_offset);
                    horizontal_offset = 0;
                    t
                } else {
                    horizontal_offset -= w;
                    ""
                }
            };
            current_line_width += symbol.width() as u16;
            self.current_line.push(StyledGrapheme { symbol, style });
        }

        if skip_rest {
            for StyledGrapheme { symbol, .. } in &mut self.symbols {
                if str_is_line_ending(symbol) {
                    break;
                }
            }
        }

        if symbols_exhausted && self.current_line.is_empty() {
            None
        } else {
            Some((&self.current_line[..], current_line_width))
        }
    }
}

/// This function will return a str slice which start at specified offset.
/// As src is a unicode str, start offset has to be calculated with each character.
fn trim_offset(src: &str, mut offset: usize) -> &str {
    let mut start = 0;
    for c in UnicodeSegmentation::graphemes(src, true) {
        let w = c.width();
        if w <= offset {
            offset -= w;
            start += c.len();
        } else {
            break;
        }
    }
    &src[start..]
}

#[cfg(test)]
mod test {
    use super::*;
    use unicode_segmentation::UnicodeSegmentation;

    enum Composer {
        WordWrapper { trim: bool },
        LineTruncator,
    }

    fn run_composer(which: Composer, text: &str, text_area_width: u16) -> (Vec<String>, Vec<u16>) {
        let style = Default::default();
        let mut styled =
            UnicodeSegmentation::graphemes(text, true).map(|g| StyledGrapheme { symbol: g, style });
        let mut composer: Box<dyn LineComposer> = match which {
            Composer::WordWrapper { trim } => {
                Box::new(WordWrapper::new(&mut styled, text_area_width, trim))
            }
            Composer::LineTruncator => Box::new(LineTruncator::new(&mut styled, text_area_width)),
        };
        let mut lines = vec![];
        let mut widths = vec![];
        while let Some((styled, width)) = composer.next_line() {
            let line = styled
                .iter()
                .map(|StyledGrapheme { symbol, .. }| *symbol)
                .collect::<String>();
            assert!(width <= text_area_width);
            lines.push(line);
            widths.push(width);
        }
        (lines, widths)
    }

    #[test]
    fn line_composer_one_line() {
        let width = 40;
        for i in 1..width {
            let text = "a".repeat(i);
            let (word_wrapper, _) =
                run_composer(Composer::WordWrapper { trim: true }, &text, width as u16);
            let (line_truncator, _) = run_composer(Composer::LineTruncator, &text, width as u16);
            let expected = vec![text];
            assert_eq!(word_wrapper, expected);
            assert_eq!(line_truncator, expected);
        }
    }

    #[test]
    fn line_composer_short_lines() {
        let width = 20;
        let text =
            "abcdefg\nhijklmno\npabcdefg\nhijklmn\nopabcdefghijk\nlmnopabcd\n\n\nefghijklmno";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);

        let wrapped: Vec<&str> = text.split('\n').collect();
        assert_eq!(word_wrapper, wrapped);
        assert_eq!(line_truncator, wrapped);
    }

    #[test]
    fn line_composer_long_word() {
        let width = 20;
        let text = "abcdefghijklmnopabcdefghijklmnopabcdefghijklmnopabcdefghijklmno";
        let (word_wrapper, _) =
            run_composer(Composer::WordWrapper { trim: true }, text, width as u16);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width as u16);

        let wrapped = vec![
            &text[..width],
            &text[width..width * 2],
            &text[width * 2..width * 3],
            &text[width * 3..],
        ];
        assert_eq!(
            word_wrapper, wrapped,
            "WordWrapper should detect the line cannot be broken on word boundary and \
             break it at line width limit."
        );
        assert_eq!(line_truncator, vec![&text[..width]]);
    }

    #[test]
    fn line_composer_long_sentence() {
        let width = 20;
        let text =
            "abcd efghij klmnopabcd efgh ijklmnopabcdefg hijkl mnopab c d e f g h i j k l m n o";
        let text_multi_space =
            "abcd efghij    klmnopabcd efgh     ijklmnopabcdefg hijkl mnopab c d e f g h i j k l \
             m n o";
        let (word_wrapper_single_space, _) =
            run_composer(Composer::WordWrapper { trim: true }, text, width as u16);
        let (word_wrapper_multi_space, _) = run_composer(
            Composer::WordWrapper { trim: true },
            text_multi_space,
            width as u16,
        );
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width as u16);

        let word_wrapped = vec![
            "abcd efghij",
            "klmnopabcd efgh",
            "ijklmnopabcdefg",
            "hijkl mnopab c d e f",
            "g h i j k l m n o",
        ];
        assert_eq!(word_wrapper_single_space, word_wrapped);
        assert_eq!(word_wrapper_multi_space, word_wrapped);

        assert_eq!(line_truncator, vec![&text[..width]]);
    }

    #[test]
    fn line_composer_zero_width() {
        let width = 0;
        let text = "abcd efghij klmnopabcd efgh ijklmnopabcdefg hijkl mnopab ";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);

        let expected: Vec<&str> = Vec::new();
        assert_eq!(word_wrapper, expected);
        assert_eq!(line_truncator, expected);
    }

    #[test]
    fn line_composer_max_line_width_of_1() {
        let width = 1;
        let text = "abcd efghij klmnopabcd efgh ijklmnopabcdefg hijkl mnopab ";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);

        let expected: Vec<&str> = UnicodeSegmentation::graphemes(text, true)
            .filter(|g| g.chars().any(|c| !c.is_whitespace()))
            .collect();
        assert_eq!(word_wrapper, expected);
        assert_eq!(line_truncator, vec!["a"]);
    }

    #[test]
    fn line_composer_max_line_width_of_1_double_width_characters() {
        let width = 1;
        let text = "コンピュータ上で文字を扱う場合、典型的には文字\naaaによる通信を行う場合にその\
                    両端点では、";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);
        assert_eq!(word_wrapper, vec!["", "a", "a", "a"]);
        assert_eq!(line_truncator, vec!["", "a"]);
    }

    /// Tests WordWrapper with words some of which exceed line length and some not.
    #[test]
    fn line_composer_word_wrapper_mixed_length() {
        let width = 20;
        let text = "abcd efghij klmnopabcdefghijklmnopabcdefghijkl mnopab cdefghi j klmno";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        assert_eq!(
            word_wrapper,
            vec![
                "abcd efghij",
                "klmnopabcdefghijklmn",
                "opabcdefghijkl",
                "mnopab cdefghi j",
                "klmno",
            ]
        )
    }

    #[test]
    fn line_composer_double_width_chars() {
        let width = 20;
        let text = "コンピュータ上で文字を扱う場合、典型的には文字による通信を行う場合にその両端点\
                    では、";
        let (word_wrapper, word_wrapper_width) =
            run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);
        assert_eq!(line_truncator, vec!["コンピュータ上で文字"]);
        let wrapped = vec![
            "コンピュータ上で文字",
            "を扱う場合、典型的に",
            "は文字による通信を行",
            "う場合にその両端点で",
            "は、",
        ];
        assert_eq!(word_wrapper, wrapped);
        assert_eq!(word_wrapper_width, vec![width, width, width, width, 4]);
    }

    #[test]
    fn line_composer_leading_whitespace_removal() {
        let width = 20;
        let text = "AAAAAAAAAAAAAAAAAAAA    AAA";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);
        assert_eq!(word_wrapper, vec!["AAAAAAAAAAAAAAAAAAAA", "AAA",]);
        assert_eq!(line_truncator, vec!["AAAAAAAAAAAAAAAAAAAA"]);
    }

    /// Tests truncation of leading whitespace.
    #[test]
    fn line_composer_lots_of_spaces() {
        let width = 20;
        let text = "                                                                     ";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);
        assert_eq!(word_wrapper, vec![""]);
        assert_eq!(line_truncator, vec!["                    "]);
    }

    /// Tests an input starting with a letter, followed by spaces - some of the behaviour is
    /// incidental.
    #[test]
    fn line_composer_char_plus_lots_of_spaces() {
        let width = 20;
        let text = "a                                                                     ";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        let (line_truncator, _) = run_composer(Composer::LineTruncator, text, width);
        // What's happening below is: the first line gets consumed, trailing spaces discarded,
        // after 20 of which a word break occurs (probably shouldn't). The second line break
        // discards all whitespace. The result should probably be vec!["a"] but it doesn't matter
        // that much.
        assert_eq!(word_wrapper, vec!["a", ""]);
        assert_eq!(line_truncator, vec!["a                   "]);
    }

    #[test]
    fn line_composer_word_wrapper_double_width_chars_mixed_with_spaces() {
        let width = 20;
        // Japanese seems not to use spaces but we should break on spaces anyway... We're using it
        // to test double-width chars.
        // You are more than welcome to add word boundary detection based of alterations of
        // hiragana and katakana...
        // This happens to also be a test case for mixed width because regular spaces are single width.
        let text = "コンピュ ータ上で文字を扱う場合、 典型的には文 字による 通信を行 う場合にその両端点では、";
        let (word_wrapper, word_wrapper_width) =
            run_composer(Composer::WordWrapper { trim: true }, text, width);
        assert_eq!(
            word_wrapper,
            vec![
                "コンピュ",
                "ータ上で文字を扱う場",
                "合、 典型的には文",
                "字による 通信を行",
                "う場合にその両端点で",
                "は、",
            ]
        );
        // Odd-sized lines have a space in them.
        assert_eq!(word_wrapper_width, vec![8, 20, 17, 17, 20, 4]);
    }

    /// Ensure words separated by nbsp are wrapped as if they were a single one.
    #[test]
    fn line_composer_word_wrapper_nbsp() {
        let width = 20;
        let text = "AAAAAAAAAAAAAAA AAAA\u{00a0}AAA";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: true }, text, width);
        assert_eq!(word_wrapper, vec!["AAAAAAAAAAAAAAA", "AAAA\u{00a0}AAA",]);

        // Ensure that if the character was a regular space, it would be wrapped differently.
        let text_space = text.replace('\u{00a0}', " ");
        let (word_wrapper_space, _) =
            run_composer(Composer::WordWrapper { trim: true }, &text_space, width);
        assert_eq!(word_wrapper_space, vec!["AAAAAAAAAAAAAAA AAAA", "AAA",]);
    }

    #[test]
    fn line_composer_word_wrapper_preserve_indentation() {
        let width = 20;
        let text = "AAAAAAAAAAAAAAAAAAAA    AAA";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: false }, text, width);
        assert_eq!(word_wrapper, vec!["AAAAAAAAAAAAAAAAAAAA", "   AAA",]);
    }

    #[test]
    fn line_composer_word_wrapper_preserve_indentation_with_wrap() {
        let width = 10;
        let text = "AAA AAA AAAAA AA AAAAAA\n B\n  C\n   D";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: false }, text, width);
        assert_eq!(
            word_wrapper,
            vec!["AAA AAA", "AAAAA AA", "AAAAAA", " B", "  C", "   D"]
        );
    }

    #[test]
    fn line_composer_word_wrapper_preserve_indentation_lots_of_whitespace() {
        let width = 10;
        let text = "               4 Indent\n                 must wrap!";
        let (word_wrapper, _) = run_composer(Composer::WordWrapper { trim: false }, text, width);
        assert_eq!(
            word_wrapper,
            vec![
                "          ",
                "    4",
                "Indent",
                "          ",
                "      must",
                "wrap!"
            ]
        );
    }
}