helix-mirror/helix-core/src/graphemes.rs

// Based on https://github.com/cessen/led/blob/c4fa72405f510b7fd16052f90a598c429b3104a6/src/graphemes.rs
use ropey::{iter::Chunks, str_utils::byte_to_char_idx, RopeSlice};
use unicode_segmentation::{GraphemeCursor, GraphemeIncomplete};
use unicode_width::UnicodeWidthStr;

use std::fmt;

#[must_use]
pub fn grapheme_width(g: &str) -> usize {
    if g.as_bytes()[0] <= 127 {
        // Fast-path ascii.
        // Point 1: theoretically, ascii control characters should have zero
        // width, but in our case we actually want them to have width: if they
        // show up in text, we want to treat them as textual elements that can
        // be editied.  So we can get away with making all ascii single width
        // here.
        // Point 2: we're only examining the first codepoint here, which means
        // we're ignoring graphemes formed with combining characters.  However,
        // if it starts with ascii, it's going to be a single-width grapeheme
        // regardless, so, again, we can get away with that here.
        // Point 3: we're only examining the first _byte_.  But for utf8, when
        // checking for ascii range values only, that works.
        1
    } else {
        // We use max(1) here because all grapeheme clusters--even illformed
        // ones--should have at least some width so they can be edited
        // properly.
        UnicodeWidthStr::width(g).max(1)
    }
}

#[must_use]
pub fn nth_prev_grapheme_boundary(slice: RopeSlice, char_idx: usize, n: usize) -> usize {
    // Bounds check
    debug_assert!(char_idx <= slice.len_chars());

    // We work with bytes for this, so convert.
    let mut byte_idx = slice.char_to_byte(char_idx);

    // Get the chunk with our byte index in it.
    let (mut chunk, mut chunk_byte_idx, mut chunk_char_idx, _) = slice.chunk_at_byte(byte_idx);

    // Set up the grapheme cursor.
    let mut gc = GraphemeCursor::new(byte_idx, slice.len_bytes(), true);

    // Find the previous grapheme cluster boundary.
    for _ in 0..n {
        loop {
            match gc.prev_boundary(chunk, chunk_byte_idx) {
                Ok(None) => return 0,
                Ok(Some(n)) => {
                    byte_idx = n;
                    break;
                }
                Err(GraphemeIncomplete::PrevChunk) => {
                    let (a, b, c, _) = slice.chunk_at_byte(chunk_byte_idx - 1);
                    chunk = a;
                    chunk_byte_idx = b;
                    chunk_char_idx = c;
                }
                Err(GraphemeIncomplete::PreContext(n)) => {
                    let ctx_chunk = slice.chunk_at_byte(n - 1).0;
                    gc.provide_context(ctx_chunk, n - ctx_chunk.len());
                }
                _ => unreachable!(),
            }
        }
    }
    let tmp = byte_to_char_idx(chunk, byte_idx - chunk_byte_idx);
    chunk_char_idx + tmp
}

/// Finds the previous grapheme boundary before the given char position.
pub fn prev_grapheme_boundary(slice: RopeSlice, char_idx: usize) -> usize {
    nth_prev_grapheme_boundary(slice, char_idx, 1)
}

#[must_use]
pub fn nth_next_grapheme_boundary(slice: RopeSlice, char_idx: usize, n: usize) -> usize {
    // Bounds check
    debug_assert!(char_idx <= slice.len_chars());

    // We work with bytes for this, so convert.
    let mut byte_idx = slice.char_to_byte(char_idx);

    // Get the chunk with our byte index in it.
    let (mut chunk, mut chunk_byte_idx, mut chunk_char_idx, _) = slice.chunk_at_byte(byte_idx);

    // Set up the grapheme cursor.
    let mut gc = GraphemeCursor::new(byte_idx, slice.len_bytes(), true);

    // Find the nth next grapheme cluster boundary.
    for _ in 0..n {
        loop {
            match gc.next_boundary(chunk, chunk_byte_idx) {
                Ok(None) => return slice.len_chars(),
                Ok(Some(n)) => {
                    byte_idx = n;
                    break;
                }
                Err(GraphemeIncomplete::NextChunk) => {
                    chunk_byte_idx += chunk.len();
                    let (a, _, c, _) = slice.chunk_at_byte(chunk_byte_idx);
                    chunk = a;
                    chunk_char_idx = c;
                }
                Err(GraphemeIncomplete::PreContext(n)) => {
                    let ctx_chunk = slice.chunk_at_byte(n - 1).0;
                    gc.provide_context(ctx_chunk, n - ctx_chunk.len());
                }
                _ => unreachable!(),
            }
        }
    }
    let tmp = byte_to_char_idx(chunk, byte_idx - chunk_byte_idx);
    chunk_char_idx + tmp
}

/// Finds the next grapheme boundary after the given char position.
pub fn next_grapheme_boundary(slice: RopeSlice, char_idx: usize) -> usize {
    nth_next_grapheme_boundary(slice, char_idx, 1)
}

/// Returns the passed char index if it's already a grapheme boundary,
/// or the next grapheme boundary char index if not.
pub fn ensure_grapheme_boundary(slice: RopeSlice, char_idx: usize) -> usize {
    if char_idx == 0 {
        0
    } else {
        next_grapheme_boundary(slice, char_idx - 1)
    }
}

/// Returns the passed byte index if it's already a grapheme boundary,
/// or the next grapheme boundary byte index if not.
pub fn ensure_grapheme_boundary_byte(slice: RopeSlice, byte_idx: usize) -> usize {
    // TODO: we can avoid the byte/char conversions entirely
    // if we also make byte versions of the other functions.
    let char_idx = slice.byte_to_char(byte_idx);
    let fixed_char_idx = ensure_grapheme_boundary(slice, char_idx);
    if fixed_char_idx == char_idx {
        byte_idx
    } else {
        slice.char_to_byte(fixed_char_idx)
    }
}

/// Returns whether the given char position is a grapheme boundary.
pub fn is_grapheme_boundary(slice: RopeSlice, char_idx: usize) -> bool {
    // Bounds check
    debug_assert!(char_idx <= slice.len_chars());

    // We work with bytes for this, so convert.
    let byte_idx = slice.char_to_byte(char_idx);

    // Get the chunk with our byte index in it.
    let (chunk, chunk_byte_idx, _, _) = slice.chunk_at_byte(byte_idx);

    // Set up the grapheme cursor.
    let mut gc = GraphemeCursor::new(byte_idx, slice.len_bytes(), true);

    // Determine if the given position is a grapheme cluster boundary.
    loop {
        match gc.is_boundary(chunk, chunk_byte_idx) {
            Ok(n) => return n,
            Err(GraphemeIncomplete::PreContext(n)) => {
                let (ctx_chunk, ctx_byte_start, _, _) = slice.chunk_at_byte(n - 1);
                gc.provide_context(ctx_chunk, ctx_byte_start);
            }
            Err(_) => unreachable!(),
        }
    }
}

/// An iterator over the graphemes of a `RopeSlice`.
#[derive(Clone)]
pub struct RopeGraphemes<'a> {
    text: RopeSlice<'a>,
    chunks: Chunks<'a>,
    cur_chunk: &'a str,
    cur_chunk_start: usize,
    cursor: GraphemeCursor,
}

impl<'a> fmt::Debug for RopeGraphemes<'a> {
    fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
        f.debug_struct("RopeGraphemes")
            .field("text", &self.text)
            .field("chunks", &self.chunks)
            .field("cur_chunk", &self.cur_chunk)
            .field("cur_chunk_start", &self.cur_chunk_start)
            // .field("cursor", &self.cursor)
            .finish()
    }
}

impl<'a> RopeGraphemes<'a> {
    #[must_use]
    pub fn new(slice: RopeSlice) -> RopeGraphemes {
        let mut chunks = slice.chunks();
        let first_chunk = chunks.next().unwrap_or("");
        RopeGraphemes {
            text: slice,
            chunks,
            cur_chunk: first_chunk,
            cur_chunk_start: 0,
            cursor: GraphemeCursor::new(0, slice.len_bytes(), true),
        }
    }
}

impl<'a> Iterator for RopeGraphemes<'a> {
    type Item = RopeSlice<'a>;

    fn next(&mut self) -> Option<RopeSlice<'a>> {
        let a = self.cursor.cur_cursor();
        let b;
        loop {
            match self
                .cursor
                .next_boundary(self.cur_chunk, self.cur_chunk_start)
            {
                Ok(None) => {
                    return None;
                }
                Ok(Some(n)) => {
                    b = n;
                    break;
                }
                Err(GraphemeIncomplete::NextChunk) => {
                    self.cur_chunk_start += self.cur_chunk.len();
                    self.cur_chunk = self.chunks.next().unwrap_or("");
                }
                Err(GraphemeIncomplete::PreContext(idx)) => {
                    let (chunk, byte_idx, _, _) = self.text.chunk_at_byte(idx.saturating_sub(1));
                    self.cursor.provide_context(chunk, byte_idx);
                }
                _ => unreachable!(),
            }
        }

        if a < self.cur_chunk_start {
            let a_char = self.text.byte_to_char(a);
            let b_char = self.text.byte_to_char(b);

            Some(self.text.slice(a_char..b_char))
        } else {
            let a2 = a - self.cur_chunk_start;
            let b2 = b - self.cur_chunk_start;
            Some((&self.cur_chunk[a2..b2]).into())
        }
    }
}