led/src/graphemes.rs

use ropey::{iter::Chunks, RopeSlice};
use unicode_segmentation::{GraphemeCursor, GraphemeIncomplete};
use unicode_width::UnicodeWidthStr;

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)
    }
}

pub fn nth_prev_grapheme_boundary(slice: &RopeSlice, byte_idx: usize, n: usize) -> usize {
    // TODO: implement this more efficiently.  This has to do a lot of
    // re-scanning of rope chunks.  Probably move the main implementation here,
    // and have prev_grapheme_boundary call this instead.
    let mut byte_idx = byte_idx;
    for _ in 0..n {
        byte_idx = prev_grapheme_boundary(slice, byte_idx);
    }
    byte_idx
}

/// Finds the previous grapheme boundary before the given char position.
pub fn prev_grapheme_boundary(slice: &RopeSlice, byte_idx: usize) -> usize {
    // Bounds check
    debug_assert!(byte_idx <= slice.len());

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

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

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

pub fn nth_next_grapheme_boundary(slice: &RopeSlice, byte_idx: usize, n: usize) -> usize {
    // TODO: implement this more efficiently.  This has to do a lot of
    // re-scanning of rope chunks.  Probably move the main implementation here,
    // and have next_grapheme_boundary call this instead.
    let mut byte_idx = byte_idx;
    for _ in 0..n {
        byte_idx = next_grapheme_boundary(slice, byte_idx);
    }
    byte_idx
}

/// Finds the next grapheme boundary after the given char position.
pub fn next_grapheme_boundary(slice: &RopeSlice, byte_idx: usize) -> usize {
    // Bounds check
    debug_assert!(byte_idx <= slice.len());

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

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

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

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

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

    // Set up the grapheme cursor.
    let mut gc = GraphemeCursor::new(byte_idx, slice.len(), 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(n - 1);
                gc.provide_context(ctx_chunk, ctx_byte_start);
            }
            _ => 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> RopeGraphemes<'a> {
    pub fn new<'b>(slice: &RopeSlice<'b>) -> RopeGraphemes<'b> {
        let mut chunks = slice.chunks();
        let first_chunk = chunks.next().unwrap_or("");
        RopeGraphemes {
            text: *slice,
            chunks: chunks,
            cur_chunk: first_chunk,
            cur_chunk_start: 0,
            cursor: GraphemeCursor::new(0, slice.len(), true),
        }
    }
}

impl<'a> Iterator for RopeGraphemes<'a> {
    type Item = std::borrow::Cow<'a, str>;

    fn next(&mut self) -> Option<std::borrow::Cow<'a, str>> {
        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(idx.saturating_sub(1));
                    self.cursor.provide_context(chunk, byte_idx);
                }
                _ => unreachable!(),
            }
        }

        if a >= self.cur_chunk_start && b <= (self.cur_chunk_start + self.cur_chunk.len()) {
            let a_internal = a - self.cur_chunk_start;
            let b_internal = b - self.cur_chunk_start;
            Some(std::borrow::Cow::Borrowed(
                &self.cur_chunk[a_internal..b_internal],
            ))
        } else {
            Some(self.text.slice(a..b).into())
        }
    }
}