mm0_util 0.1.4

//! Embellished String which carries additional data useful for interacting with language server messages.
//!
//! The [`LinedString::apply_changes`] function is used to realize changes made to a string once those
//! changes are received by mm0-rs from a language server. [`LinedString`] Implements associated
//! methods allowing it to be used nicely with the [`Position`] type specified by the language server protocol.

use crate::{Position, Range, Span};
#[cfg(feature = "memory")]
use mm0_deepsize_derive::DeepSizeOf;
use std::ops::{Deref, Index};

/// Wrapper around std's String which stores data about the positions of any newline characters.
///
/// Also contains a boolean indicating whether the string has any unicode characters.
/// Unicode is currently unsupported, but this allows the lexer to gracefully handle
/// errors arising from the presence of unicode characters in input.
/// The indices stored in `lines` are the successors of any newline characters.
#[cfg_attr(feature = "memory", derive(DeepSizeOf))]
#[derive(Default, Clone, Debug)]
pub struct LinedString {
  s: String,
  unicode: bool,
  lines: Vec<usize>,
}

/// Allows [`LinedString`] to be indexed with a [`Span`], since [`Span`] is essentially a range.
impl Index<Span> for LinedString {
  type Output = [u8];
  fn index(&self, s: Span) -> &[u8] { &self.as_bytes()[s.start..s.end] }
}

/// Calculates the largest index `n` (on a UTF8 boundary) such that `s[..n]` is at most `chs` UTF-16 codepoints.
/// Used in [`LinedString::lsp_to_idx`] to account for additional character offset introduced by unicode.
fn lsp_to_idx(s: &str, mut chs: usize) -> usize {
  for (n, c) in s.char_indices() {
    let i = c.len_utf16();
    if chs < i {
      return n
    }
    chs -= i;
  }
  s.len()
}

impl LinedString {
  /// Index a [`LinedString`] with a [`Span`], returning a `str`.
  ///
  /// # Safety
  /// This function uses `str::get_unchecked()` internally, so it is *unsafe* unless the [`Span`] used in the index
  /// was generated from the file that is being indexed.
  #[allow(unused)]
  pub fn str_at(&self, s: Span) -> &str { unsafe { std::str::from_utf8_unchecked(&self[s]) } }

  /// Calculate and store information about the positions of any newline
  /// characters in the string, and set 'unicode' to true if the string contains unicode.
  /// The data in 'lines' is actually the positions of the characters immediately after
  /// the line break (so \n.pos + 1).
  #[must_use]
  fn get_lines(unicode: &mut bool, s: &str) -> Vec<usize> {
    let mut lines = vec![];
    for (b, c) in s.char_indices() {
      if c == '\n' {
        lines.push(b + 1)
      }
      if !c.is_ascii() {
        *unicode = true
      }
    }
    lines
  }

  /// Turn a byte index into an LSP [`Position`]
  ///
  /// # Safety
  /// `idx` must be a valid index in the string.
  #[must_use]
  pub fn to_pos(&self, idx: usize) -> Position {
    use std::convert::TryInto;
    let (pos, line) = match self.lines.binary_search(&idx) {
      Ok(n) => (idx, n + 1),
      Err(n) => (n.checked_sub(1).map_or(0, |i| self.lines[i]), n),
    };
    Position {
      line: line.try_into().expect("too many lines"),
      character: if self.unicode {
        // Safety: we know that `pos` is valid index, and we have assumed that `idx` is
        unsafe { self.s.get_unchecked(pos..idx) }.chars().map(char::len_utf16).sum()
      } else {
        idx - pos
      }
      .try_into()
      .expect("too many characters"),
    }
  }

  /// Turn a [`Span`] into an LSP [`Range`].
  #[must_use]
  pub fn to_range(&self, s: Span) -> Range {
    Range { start: self.to_pos(s.start), end: self.to_pos(s.end) }
  }

  /// Turn a [`FileSpan`](crate::FileSpan) into an LSP [`Location`](lsp_types::Location).
  #[cfg(feature = "server")]
  #[must_use]
  pub fn to_loc(&self, fs: &crate::FileSpan) -> lsp_types::Location {
    lsp_types::Location { uri: fs.file.url().clone(), range: self.to_range(fs.span) }
  }

  /// Get the total number of lines in the file (as a `u32` for LSP compatibility).
  #[must_use]
  pub fn num_lines(&self) -> u32 {
    use std::convert::TryInto;
    self.lines.len().try_into().expect("too many lines")
  }

  /// Get the [`Position`] (line and UTF-16 code unit offset) of the end of the file.
  #[must_use]
  pub fn end(&self) -> Position { self.to_pos(self.s.len()) }

  /// Calculates the byte index of the position `chs` UTF-16 code units after
  /// byte index `start` in the string.
  /// If there's no unicode, we can just use (start + idx).
  /// In the presence of unicode, use the helper function [`lsp_to_idx`](Self::lsp_to_idx)
  /// to account for any additional character offset.
  ///
  /// # Safety
  /// `start` must be a valid index in the string.
  #[must_use]
  fn lsp_to_idx(&self, start: usize, chs: usize) -> usize {
    start + if self.unicode { lsp_to_idx(unsafe { self.get_unchecked(start..) }, chs) } else { chs }
  }

  /// Turn an LSP [`Position`] into a usize index. [`Position`] is already zero-based,
  /// but [`LinedString.lines`] stores `1 + position` of the actual linebreak characters,
  /// so `lines[0]` points to the start of line 1, `lines[1]` points to the start of line 2, etc.
  /// with the start of line 0 just being s.0.
  #[must_use]
  pub fn to_idx(&self, pos: Position) -> Option<usize> {
    match pos.line.checked_sub(1) {
      None => Some(self.lsp_to_idx(0, pos.character as usize)),
      Some(n) =>
        self.lines.get(n as usize).map(|&idx| self.lsp_to_idx(idx, pos.character as usize)),
    }
  }

  /// Extend a [`LinedString`] with the contents of a `&str`, adding additional newline info as needed.
  pub fn extend(&mut self, s: &str) {
    let len = self.s.len();
    self.s.push_str(s);
    for (b, c) in s.char_indices() {
      if c == '\n' {
        self.lines.push(b + len + 1)
      }
      if !c.is_ascii() {
        self.unicode = true
      }
    }
  }

  /// Extends a [`LinedString`]'s contents with the contents of a passed string slice
  /// until it reaches some [`Position`]. Returns the portion of the passed string slice
  /// that was not added to the [`LinedString`].
  pub fn extend_until<'a>(&mut self, unicode: bool, s: &'a str, pos: Position) -> &'a str {
    self.unicode |= unicode;
    let end = self.end();
    debug_assert!(end <= pos);
    let (chs, off) = if pos.line == end.line {
      ((pos.character - end.character) as usize, 0)
    } else {
      let len = self.s.len();
      let mut it = s.char_indices();
      (
        pos.character as usize,
        loop {
          if let Some((b, c)) = it.next() {
            if c == '\n' {
              self.lines.push(b + len + 1);
              if pos.line == self.num_lines() {
                break b + 1
              }
            }
          } else {
            break s.len()
          }
        },
      )
    };
    let tail = unsafe { s.get_unchecked(off..) };
    let idx = if unicode { lsp_to_idx(tail, chs) } else { chs };
    let len = self.s.len() + off;
    for (b, c) in unsafe { tail.get_unchecked(..idx) }.char_indices() {
      if c == '\n' {
        self.lines.push(b + len + 1)
      }
    }
    let (left, right) = s.split_at(off + idx);
    self.s.push_str(left);
    right
  }

  /// Truncates a [`LinedString`]'s contents so that it's equal to the character position
  /// indicated by some lsp [`Position`], discarding any unneeded newline data.
  /// Does nothing if the [`LinedString`]'s contents were already less than or equal
  /// in length to the [`Position`]'s index.
  pub fn truncate(&mut self, pos: Position) {
    if let Some(idx) = self.to_idx(pos) {
      if idx < self.s.len() {
        self.s.truncate(idx);
        self.lines.truncate(pos.line as usize);
      }
    }
  }

  /// Does a bunch of string juggling to actually realize the contents of an iterator
  /// containing a sequence of [`TextDocumentContentChangeEvent`] messages.
  ///
  /// [`TextDocumentContentChangeEvent`]: lsp_types::TextDocumentContentChangeEvent
  #[cfg(feature = "server")]
  #[must_use]
  pub fn apply_changes(
    &self, changes: impl Iterator<Item = lsp_types::TextDocumentContentChangeEvent>,
  ) -> (Position, LinedString) {
    let mut old: LinedString;
    let mut out = LinedString::default();
    let mut uncopied: &str = &self.s;
    let mut first_change = None;
    for e in changes {
      if let Some(Range { start, end }) = e.range {
        if first_change.map_or(true, |c| start < c) {
          first_change = Some(start)
        }
        if out.end() > start {
          out.extend(uncopied);
          old = std::mem::take(&mut out);
          uncopied = &old;
        }
        uncopied = out.extend_until(self.unicode, uncopied, end);
        out.truncate(start);
        out.extend(&e.text);
      } else {
        out = e.text.into();
        first_change = Some(Position::default());
        uncopied = "";
      }
    }
    out.extend(uncopied);
    if let Some(pos) = first_change {
      let start = out.to_idx(pos).expect("change out of range");
      let from = unsafe { self.s.get_unchecked(start..) };
      let to = unsafe { out.s.get_unchecked(start..) };
      for ((b, c1), c2) in from.char_indices().zip(to.chars()) {
        if c1 != c2 {
          return (out.to_pos(b + start), out)
        }
      }
    }
    (out.end(), out)
  }
}

impl Deref for LinedString {
  type Target = String;
  fn deref(&self) -> &String { &self.s }
}

impl From<String> for LinedString {
  fn from(s: String) -> LinedString {
    let mut unicode = false;
    let lines = LinedString::get_lines(&mut unicode, &s);
    LinedString { unicode, lines, s }
  }
}