mirror of
https://github.com/denoland/deno.git
synced 2024-11-24 15:19:26 -05:00
530 lines
13 KiB
Rust
530 lines
13 KiB
Rust
// Copyright 2018-2022 the Deno authors. All rights reserved. MIT license.
|
|
|
|
use deno_core::error::custom_error;
|
|
use deno_core::error::AnyError;
|
|
use dissimilar::diff;
|
|
use dissimilar::Chunk;
|
|
use std::collections::HashMap;
|
|
use text_size::TextRange;
|
|
use text_size::TextSize;
|
|
use tower_lsp::jsonrpc;
|
|
use tower_lsp::lsp_types as lsp;
|
|
use tower_lsp::lsp_types::TextEdit;
|
|
|
|
fn partition_point<T, P>(slice: &[T], mut predicate: P) -> usize
|
|
where
|
|
P: FnMut(&T) -> bool,
|
|
{
|
|
let mut left = 0;
|
|
let mut right = slice.len() - 1;
|
|
|
|
while left != right {
|
|
let mid = left + (right - left) / 2;
|
|
// SAFETY:
|
|
// When left < right, left <= mid < right.
|
|
// Therefore left always increases and right always decreases,
|
|
// and either of them is selected.
|
|
// In both cases left <= right is satisfied.
|
|
// Therefore if left < right in a step,
|
|
// left <= right is satisfied in the next step.
|
|
// Therefore as long as left != right, 0 <= left < right < len is satisfied
|
|
// and if this case 0 <= mid < len is satisfied too.
|
|
let value = unsafe { slice.get_unchecked(mid) };
|
|
if predicate(value) {
|
|
left = mid + 1;
|
|
} else {
|
|
right = mid;
|
|
}
|
|
}
|
|
|
|
left
|
|
}
|
|
|
|
#[derive(Debug, Clone, Eq, PartialEq, Hash)]
|
|
pub struct Utf16Char {
|
|
pub start: TextSize,
|
|
pub end: TextSize,
|
|
}
|
|
|
|
impl Utf16Char {
|
|
fn len(&self) -> TextSize {
|
|
self.end - self.start
|
|
}
|
|
|
|
fn len_utf16(&self) -> usize {
|
|
if self.len() == TextSize::from(4) {
|
|
2
|
|
} else {
|
|
1
|
|
}
|
|
}
|
|
}
|
|
|
|
#[derive(Debug, Clone, Default, Eq, PartialEq)]
|
|
pub struct LineIndex {
|
|
utf8_offsets: Vec<TextSize>,
|
|
utf16_lines: HashMap<u32, Vec<Utf16Char>>,
|
|
utf16_offsets: Vec<TextSize>,
|
|
}
|
|
|
|
impl LineIndex {
|
|
pub fn new(text: &str) -> LineIndex {
|
|
let mut utf16_lines = HashMap::new();
|
|
let mut utf16_chars = Vec::new();
|
|
|
|
let mut utf8_offsets = vec![0.into()];
|
|
let mut utf16_offsets = vec![0.into()];
|
|
let mut curr_row = 0.into();
|
|
let mut curr_col = 0.into();
|
|
let mut curr_offset_u16 = 0.into();
|
|
let mut line = 0;
|
|
for c in text.chars() {
|
|
let c_len = TextSize::of(c);
|
|
curr_row += c_len;
|
|
curr_offset_u16 += TextSize::from(c.len_utf16() as u32);
|
|
if c == '\n' {
|
|
utf8_offsets.push(curr_row);
|
|
utf16_offsets.push(curr_offset_u16);
|
|
|
|
if !utf16_chars.is_empty() {
|
|
utf16_lines.insert(line, utf16_chars);
|
|
utf16_chars = Vec::new();
|
|
}
|
|
|
|
curr_col = 0.into();
|
|
line += 1;
|
|
continue;
|
|
}
|
|
|
|
if !c.is_ascii() {
|
|
utf16_chars.push(Utf16Char {
|
|
start: curr_col,
|
|
end: curr_col + c_len,
|
|
});
|
|
}
|
|
curr_col += c_len;
|
|
}
|
|
|
|
// utf8_offsets and utf16_offsets length is equal to (# of lines + 1)
|
|
utf8_offsets.push(curr_row);
|
|
utf16_offsets.push(curr_offset_u16);
|
|
|
|
if !utf16_chars.is_empty() {
|
|
utf16_lines.insert(line, utf16_chars);
|
|
}
|
|
|
|
LineIndex {
|
|
utf8_offsets,
|
|
utf16_lines,
|
|
utf16_offsets,
|
|
}
|
|
}
|
|
|
|
/// Convert a u16 based range to a u8 TextRange.
|
|
pub fn get_text_range(
|
|
&self,
|
|
range: lsp::Range,
|
|
) -> Result<TextRange, AnyError> {
|
|
let start = self.offset(range.start)?;
|
|
let end = self.offset(range.end)?;
|
|
Ok(TextRange::new(start, end))
|
|
}
|
|
|
|
/// Return a u8 offset based on a u16 position.
|
|
pub fn offset(&self, position: lsp::Position) -> Result<TextSize, AnyError> {
|
|
let col = self.utf16_to_utf8_col(position.line, position.character);
|
|
if let Some(line_offset) = self.utf8_offsets.get(position.line as usize) {
|
|
Ok(line_offset + col)
|
|
} else {
|
|
Err(custom_error("OutOfRange", "The position is out of range."))
|
|
}
|
|
}
|
|
|
|
/// Convert an lsp Position into a tsc/TypeScript "position", which is really
|
|
/// an u16 byte offset from the start of the string represented as an u32.
|
|
pub fn offset_tsc(&self, position: lsp::Position) -> jsonrpc::Result<u32> {
|
|
self
|
|
.offset_utf16(position)
|
|
.map(|ts| ts.into())
|
|
.map_err(|err| jsonrpc::Error::invalid_params(err.to_string()))
|
|
}
|
|
|
|
fn offset_utf16(
|
|
&self,
|
|
position: lsp::Position,
|
|
) -> Result<TextSize, AnyError> {
|
|
if let Some(line_offset) = self.utf16_offsets.get(position.line as usize) {
|
|
Ok(line_offset + TextSize::from(position.character))
|
|
} else {
|
|
Err(custom_error("OutOfRange", "The position is out of range."))
|
|
}
|
|
}
|
|
|
|
/// Returns a u16 position based on a u16 offset, which TypeScript offsets are
|
|
/// returned as u16.
|
|
pub fn position_tsc(&self, offset: TextSize) -> lsp::Position {
|
|
let line = partition_point(&self.utf16_offsets, |&it| it <= offset) - 1;
|
|
let line_start_offset = self.utf16_offsets[line];
|
|
let col = offset - line_start_offset;
|
|
|
|
lsp::Position {
|
|
line: line as u32,
|
|
character: col.into(),
|
|
}
|
|
}
|
|
|
|
/// Returns a u16 position based on a u8 offset.
|
|
pub fn position_utf16(&self, offset: TextSize) -> lsp::Position {
|
|
let line = partition_point(&self.utf16_offsets, |&it| it <= offset) - 1;
|
|
let line_start_offset = self.utf16_offsets[line];
|
|
let col = offset - line_start_offset;
|
|
|
|
lsp::Position {
|
|
line: line as u32,
|
|
character: col.into(),
|
|
}
|
|
}
|
|
|
|
pub fn text_content_length_utf16(&self) -> TextSize {
|
|
*self.utf16_offsets.last().unwrap()
|
|
}
|
|
|
|
fn utf16_to_utf8_col(&self, line: u32, mut col: u32) -> TextSize {
|
|
if let Some(utf16_chars) = self.utf16_lines.get(&line) {
|
|
for c in utf16_chars {
|
|
if col > u32::from(c.start) {
|
|
col += u32::from(c.len()) - c.len_utf16() as u32;
|
|
} else {
|
|
break;
|
|
}
|
|
}
|
|
}
|
|
|
|
col.into()
|
|
}
|
|
}
|
|
|
|
/// Compare two strings and return a vector of text edit records which are
|
|
/// supported by the Language Server Protocol.
|
|
pub fn get_edits(a: &str, b: &str, line_index: &LineIndex) -> Vec<TextEdit> {
|
|
if a == b {
|
|
return vec![];
|
|
}
|
|
let chunks = diff(a, b);
|
|
let mut text_edits = Vec::<TextEdit>::new();
|
|
let mut iter = chunks.iter().peekable();
|
|
let mut a_pos = TextSize::from(0);
|
|
loop {
|
|
let chunk = iter.next();
|
|
match chunk {
|
|
None => break,
|
|
Some(Chunk::Equal(e)) => {
|
|
a_pos += TextSize::from(e.encode_utf16().count() as u32);
|
|
}
|
|
Some(Chunk::Delete(d)) => {
|
|
let start = line_index.position_utf16(a_pos);
|
|
a_pos += TextSize::from(d.encode_utf16().count() as u32);
|
|
let end = line_index.position_utf16(a_pos);
|
|
let range = lsp::Range { start, end };
|
|
match iter.peek() {
|
|
Some(Chunk::Insert(i)) => {
|
|
iter.next();
|
|
text_edits.push(TextEdit {
|
|
range,
|
|
new_text: i.to_string(),
|
|
});
|
|
}
|
|
_ => text_edits.push(TextEdit {
|
|
range,
|
|
new_text: "".to_string(),
|
|
}),
|
|
}
|
|
}
|
|
Some(Chunk::Insert(i)) => {
|
|
let pos = line_index.position_utf16(a_pos);
|
|
let range = lsp::Range {
|
|
start: pos,
|
|
end: pos,
|
|
};
|
|
text_edits.push(TextEdit {
|
|
range,
|
|
new_text: i.to_string(),
|
|
});
|
|
}
|
|
}
|
|
}
|
|
|
|
text_edits
|
|
}
|
|
|
|
#[cfg(test)]
|
|
mod tests {
|
|
use super::*;
|
|
|
|
#[test]
|
|
fn test_line_index() {
|
|
let text = "hello\nworld";
|
|
let index = LineIndex::new(text);
|
|
assert_eq!(
|
|
index.position_utf16(0.into()),
|
|
lsp::Position {
|
|
line: 0,
|
|
character: 0
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(1.into()),
|
|
lsp::Position {
|
|
line: 0,
|
|
character: 1
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(5.into()),
|
|
lsp::Position {
|
|
line: 0,
|
|
character: 5
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(6.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 0
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(7.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 1
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(8.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 2
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(10.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 4
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(11.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 5
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(12.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 6
|
|
}
|
|
);
|
|
|
|
let text = "\nhello\nworld";
|
|
let index = LineIndex::new(text);
|
|
assert_eq!(
|
|
index.position_utf16(0.into()),
|
|
lsp::Position {
|
|
line: 0,
|
|
character: 0
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(1.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 0
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(2.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 1
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(6.into()),
|
|
lsp::Position {
|
|
line: 1,
|
|
character: 5
|
|
}
|
|
);
|
|
assert_eq!(
|
|
index.position_utf16(7.into()),
|
|
lsp::Position {
|
|
line: 2,
|
|
character: 0
|
|
}
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn test_char_len() {
|
|
assert_eq!('メ'.len_utf8(), 3);
|
|
assert_eq!('メ'.len_utf16(), 1);
|
|
assert_eq!('编'.len_utf8(), 3);
|
|
assert_eq!('编'.len_utf16(), 1);
|
|
assert_eq!('🦕'.len_utf8(), 4);
|
|
assert_eq!('🦕'.len_utf16(), 2);
|
|
}
|
|
|
|
#[test]
|
|
fn test_empty_index() {
|
|
let col_index = LineIndex::new(
|
|
"
|
|
const C: char = 'x';
|
|
",
|
|
);
|
|
assert_eq!(col_index.utf16_lines.len(), 0);
|
|
}
|
|
|
|
#[test]
|
|
fn test_single_char() {
|
|
let col_index = LineIndex::new(
|
|
"
|
|
const C: char = 'メ';
|
|
",
|
|
);
|
|
|
|
assert_eq!(col_index.utf16_lines.len(), 1);
|
|
assert_eq!(col_index.utf16_lines[&1].len(), 1);
|
|
assert_eq!(
|
|
col_index.utf16_lines[&1][0],
|
|
Utf16Char {
|
|
start: 17.into(),
|
|
end: 20.into()
|
|
}
|
|
);
|
|
|
|
// UTF-16 to UTF-8, no changes
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 15), TextSize::from(15));
|
|
|
|
// UTF-16 to UTF-8
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 19), TextSize::from(21));
|
|
|
|
let col_index = LineIndex::new("a𐐏b");
|
|
assert_eq!(col_index.utf16_to_utf8_col(0, 3), TextSize::from(5));
|
|
}
|
|
|
|
#[test]
|
|
fn test_string() {
|
|
let col_index = LineIndex::new(
|
|
"
|
|
const C: char = \"メ メ\";
|
|
",
|
|
);
|
|
|
|
assert_eq!(col_index.utf16_lines.len(), 1);
|
|
assert_eq!(col_index.utf16_lines[&1].len(), 2);
|
|
assert_eq!(
|
|
col_index.utf16_lines[&1][0],
|
|
Utf16Char {
|
|
start: 17.into(),
|
|
end: 20.into()
|
|
}
|
|
);
|
|
assert_eq!(
|
|
col_index.utf16_lines[&1][1],
|
|
Utf16Char {
|
|
start: 21.into(),
|
|
end: 24.into()
|
|
}
|
|
);
|
|
|
|
// UTF-16 to UTF-8
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 15), TextSize::from(15));
|
|
|
|
// メ UTF-8: 0xE3 0x83 0xA1, UTF-16: 0x30E1
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 17), TextSize::from(17)); // first メ at 17..20
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 18), TextSize::from(20)); // space
|
|
assert_eq!(col_index.utf16_to_utf8_col(1, 19), TextSize::from(21)); // second メ at 21..24
|
|
|
|
assert_eq!(col_index.utf16_to_utf8_col(2, 15), TextSize::from(15));
|
|
}
|
|
|
|
#[test]
|
|
fn test_get_edits() {
|
|
let a = "abcdefg";
|
|
let b = "a\nb\nchije\nfg\n";
|
|
let actual = get_edits(a, b, &LineIndex::new(a));
|
|
assert_eq!(
|
|
actual,
|
|
vec![
|
|
TextEdit {
|
|
range: lsp::Range {
|
|
start: lsp::Position {
|
|
line: 0,
|
|
character: 1
|
|
},
|
|
end: lsp::Position {
|
|
line: 0,
|
|
character: 5
|
|
}
|
|
},
|
|
new_text: "\nb\nchije\n".to_string()
|
|
},
|
|
TextEdit {
|
|
range: lsp::Range {
|
|
start: lsp::Position {
|
|
line: 0,
|
|
character: 7
|
|
},
|
|
end: lsp::Position {
|
|
line: 0,
|
|
character: 7
|
|
}
|
|
},
|
|
new_text: "\n".to_string()
|
|
},
|
|
]
|
|
);
|
|
}
|
|
|
|
#[test]
|
|
fn test_get_edits_mbc() {
|
|
let a = "const bar = \"👍🇺🇸😃\";\nconsole.log('hello deno')\n";
|
|
let b = "const bar = \"👍🇺🇸😃\";\nconsole.log(\"hello deno\");\n";
|
|
let actual = get_edits(a, b, &LineIndex::new(a));
|
|
assert_eq!(
|
|
actual,
|
|
vec![
|
|
TextEdit {
|
|
range: lsp::Range {
|
|
start: lsp::Position {
|
|
line: 1,
|
|
character: 12
|
|
},
|
|
end: lsp::Position {
|
|
line: 1,
|
|
character: 13
|
|
}
|
|
},
|
|
new_text: "\"".to_string()
|
|
},
|
|
TextEdit {
|
|
range: lsp::Range {
|
|
start: lsp::Position {
|
|
line: 1,
|
|
character: 23
|
|
},
|
|
end: lsp::Position {
|
|
line: 1,
|
|
character: 25
|
|
}
|
|
},
|
|
new_text: "\");".to_string()
|
|
},
|
|
]
|
|
)
|
|
}
|
|
}
|