rust_analyzer_mcp/
position.rs

1//! Turning the positions the LSP speaks in into offsets into a file.
2//!
3//! A position is a line and a column, both counted from zero, and the column is counted in
4//! UTF-16 code units -- not bytes, and not characters. For a line of ASCII all three agree, which
5//! is why the difference is easy to miss and unforgiving when it bites: on a line with an
6//! accented letter the byte offset runs ahead, and on one with an emoji the UTF-16 count does.
7
8/// The byte offset of the LSP position (`line`, `character`) in `text`.
9///
10/// A position past the end of its line, or past the end of the text, gives the end of what it
11/// points into: that is what editors do with one, and it keeps a rounding error in whoever asked
12/// from becoming a panic here.
13pub fn byte_offset(text: &str, line: u32, character: u32) -> usize {
14    let mut offset = 0;
15
16    for (number, line_text) in text.split_inclusive('\n').enumerate() {
17        if number as u32 == line {
18            return offset + column_offset(line_text, character);
19        }
20        offset += line_text.len();
21    }
22
23    // Past the last line, which a range ending at the end of the file does.
24    text.len()
25}
26
27/// The byte offset of `character`, counted in UTF-16 code units, into one line.
28fn column_offset(line: &str, character: u32) -> usize {
29    let line = line.strip_suffix('\n').unwrap_or(line);
30    let line = line.strip_suffix('\r').unwrap_or(line);
31
32    let mut counted = 0;
33    for (offset, character_here) in line.char_indices() {
34        if counted >= character {
35            return offset;
36        }
37
38        // A column that lands inside a character -- half an emoji -- is not an offset into
39        // anything, so it points at the character it landed in.
40        let after = counted + character_here.len_utf16() as u32;
41        if after > character {
42            return offset;
43        }
44        counted = after;
45    }
46
47    line.len()
48}
49
50#[cfg(test)]
51mod tests {
52    use super::*;
53
54    const TEXT: &str = "fn main() {\n    let x = 1;\n}\n";
55
56    #[test]
57    fn a_position_is_a_line_and_a_column() {
58        assert_eq!(byte_offset(TEXT, 0, 0), 0);
59        assert_eq!(byte_offset(TEXT, 0, 3), 3);
60        assert_eq!(byte_offset(TEXT, 1, 4), "fn main() {\n    ".len());
61        assert_eq!(byte_offset(TEXT, 2, 0), TEXT.len() - 2);
62    }
63
64    #[test]
65    fn columns_are_counted_in_utf16_code_units() {
66        // `é` is one code unit and two bytes; `🦀` is two code units and four bytes.
67        let text = "let é = \"🦀\";";
68
69        assert_eq!(byte_offset(text, 0, 4), "let ".len());
70        assert_eq!(byte_offset(text, 0, 5), "let é".len());
71        assert_eq!(byte_offset(text, 0, 9), "let é = \"".len());
72        // The crab is two code units wide and four bytes long.
73        assert_eq!(byte_offset(text, 0, 11), "let é = \"🦀".len());
74    }
75
76    #[test]
77    fn a_column_inside_a_character_stops_before_it() {
78        // Half an emoji is not an offset; the character it lands in is where it points.
79        let text = "🦀🦀";
80
81        assert_eq!(byte_offset(text, 0, 1), 0);
82        assert_eq!(byte_offset(text, 0, 2), "🦀".len());
83        assert_eq!(byte_offset(text, 0, 3), "🦀".len());
84    }
85
86    #[test]
87    fn a_position_past_the_end_is_the_end() {
88        assert_eq!(byte_offset(TEXT, 0, 999), "fn main() {".len());
89        assert_eq!(byte_offset(TEXT, 99, 0), TEXT.len());
90        assert_eq!(byte_offset("", 0, 0), 0);
91    }
92
93    #[test]
94    fn line_endings_are_not_part_of_the_line() {
95        let text = "one\r\ntwo\r\n";
96
97        assert_eq!(byte_offset(text, 0, 99), "one".len());
98        assert_eq!(byte_offset(text, 1, 0), "one\r\n".len());
99        assert_eq!(byte_offset(text, 1, 3), "one\r\ntwo".len());
100    }
101}
rust_analyzer_mcp/position.rs

rust_analyzer_mcp/
position.rs
