Merge #227

227: Correctly map between UTF-8 and UTF-16 positions r=aochagavia a=aochagavia

Fixes #202 

Co-authored-by: Adolfo Ochagavía <github@adolfo.ochagavia.xyz>
Co-authored-by: Adolfo Ochagavía <aochagavia92@gmail.com>

crates/ra_editor/src/line_index.rs

@@ -1,43 +1,124 @@
 use crate::TextUnit;
+use rustc_hash::FxHashMap;
 use superslice::Ext;
 
-#[derive(Clone, Debug, Hash, PartialEq, Eq)]
+#[derive(Clone, Debug, PartialEq, Eq)]
 pub struct LineIndex {
     newlines: Vec<TextUnit>,
+    utf16_lines: FxHashMap<u32, Vec<Utf16Char>>,
 }
 
 #[derive(Clone, Copy, Debug, PartialEq, Eq, Hash)]
 pub struct LineCol {
     pub line: u32,
-    pub col: TextUnit,
+    pub col_utf16: u32,
+}
+
+#[derive(Clone, Debug, Hash, PartialEq, Eq)]
+struct Utf16Char {
+    start: TextUnit,
+    end: TextUnit,
+}
+
+impl Utf16Char {
+    fn len(&self) -> TextUnit {
+        self.end - self.start
+    }
 }
 
 impl LineIndex {
     pub fn new(text: &str) -> LineIndex {
+        let mut utf16_lines = FxHashMap::default();
+        let mut utf16_chars = Vec::new();
+
         let mut newlines = vec![0.into()];
-        let mut curr = 0.into();
+        let mut curr_row = 0.into();
+        let mut curr_col = 0.into();
+        let mut line = 0;
         for c in text.chars() {
-            curr += TextUnit::of_char(c);
+            curr_row += TextUnit::of_char(c);
             if c == '\n' {
-                newlines.push(curr);
+                newlines.push(curr_row);
+
+                // Save any utf-16 characters seen in the previous line
+                if utf16_chars.len() > 0 {
+                    utf16_lines.insert(line, utf16_chars);
+                    utf16_chars = Vec::new();
+                }
+
+                // Prepare for processing the next line
+                curr_col = 0.into();
+                line += 1;
+                continue;
             }
+
+            let char_len = TextUnit::of_char(c);
+            if char_len.to_usize() > 1 {
+                utf16_chars.push(Utf16Char {
+                    start: curr_col,
+                    end: curr_col + char_len,
+                });
+            }
+
+            curr_col += char_len;
+        }
+        LineIndex {
+            newlines,
+            utf16_lines,
         }
-        LineIndex { newlines }
     }
 
     pub fn line_col(&self, offset: TextUnit) -> LineCol {
         let line = self.newlines.upper_bound(&offset) - 1;
         let line_start_offset = self.newlines[line];
         let col = offset - line_start_offset;
+
         LineCol {
             line: line as u32,
-            col,
+            col_utf16: self.utf8_to_utf16_col(line as u32, col) as u32,
         }
     }
 
     pub fn offset(&self, line_col: LineCol) -> TextUnit {
         //TODO: return Result
-        self.newlines[line_col.line as usize] + line_col.col
+        let col = self.utf16_to_utf8_col(line_col.line, line_col.col_utf16);
+        self.newlines[line_col.line as usize] + col
+    }
+
+    fn utf8_to_utf16_col(&self, line: u32, mut col: TextUnit) -> usize {
+        if let Some(utf16_chars) = self.utf16_lines.get(&line) {
+            let mut correction = TextUnit::from_usize(0);
+            for c in utf16_chars {
+                if col >= c.end {
+                    correction += c.len() - TextUnit::from_usize(1);
+                } else {
+                    // From here on, all utf16 characters come *after* the character we are mapping,
+                    // so we don't need to take them into account
+                    break;
+                }
+            }
+
+            col -= correction;
+        }
+
+        col.to_usize()
+    }
+
+    fn utf16_to_utf8_col(&self, line: u32, col: u32) -> TextUnit {
+        let mut col: TextUnit = col.into();
+        if let Some(utf16_chars) = self.utf16_lines.get(&line) {
+            for c in utf16_chars {
+                if col >= c.start {
+                    col += c.len() - TextUnit::from_usize(1);
+                } else {
+                    // From here on, all utf16 characters come *after* the character we are mapping,
+                    // so we don't need to take them into account
+                    break;
+                }
+            }
+        }
+
+        col
     }
 }
 
@@ -49,63 +130,63 @@ fn test_line_index() {
         index.line_col(0.into()),
         LineCol {
             line: 0,
-            col: 0.into()
+            col_utf16: 0
         }
     );
     assert_eq!(
         index.line_col(1.into()),
         LineCol {
             line: 0,
-            col: 1.into()
+            col_utf16: 1
         }
     );
     assert_eq!(
         index.line_col(5.into()),
         LineCol {
             line: 0,
-            col: 5.into()
+            col_utf16: 5
         }
     );
     assert_eq!(
         index.line_col(6.into()),
         LineCol {
             line: 1,
-            col: 0.into()
+            col_utf16: 0
         }
     );
     assert_eq!(
         index.line_col(7.into()),
         LineCol {
             line: 1,
-            col: 1.into()
+            col_utf16: 1
         }
     );
     assert_eq!(
         index.line_col(8.into()),
         LineCol {
             line: 1,
-            col: 2.into()
+            col_utf16: 2
         }
     );
     assert_eq!(
         index.line_col(10.into()),
         LineCol {
             line: 1,
-            col: 4.into()
+            col_utf16: 4
         }
     );
     assert_eq!(
         index.line_col(11.into()),
         LineCol {
             line: 1,
-            col: 5.into()
+            col_utf16: 5
         }
     );
     assert_eq!(
         index.line_col(12.into()),
         LineCol {
             line: 1,
-            col: 6.into()
+            col_utf16: 6
         }
     );
 
@@ -115,35 +196,129 @@ fn test_line_index() {
         index.line_col(0.into()),
         LineCol {
             line: 0,
-            col: 0.into()
+            col_utf16: 0
         }
     );
     assert_eq!(
         index.line_col(1.into()),
         LineCol {
             line: 1,
-            col: 0.into()
+            col_utf16: 0
         }
     );
     assert_eq!(
         index.line_col(2.into()),
         LineCol {
             line: 1,
-            col: 1.into()
+            col_utf16: 1
         }
     );
     assert_eq!(
         index.line_col(6.into()),
         LineCol {
             line: 1,
-            col: 5.into()
+            col_utf16: 5
         }
     );
     assert_eq!(
         index.line_col(7.into()),
         LineCol {
             line: 2,
-            col: 0.into()
+            col_utf16: 0
         }
     );
 }
+
+#[cfg(test)]
+mod test_utf8_utf16_conv {
+    use super::*;
+
+    #[test]
+    fn test_char_len() {
+        assert_eq!('メ'.len_utf8(), 3);
+        assert_eq!('メ'.len_utf16(), 1);
+    }
+
+    #[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-8 to UTF-16, no changes
+        assert_eq!(col_index.utf8_to_utf16_col(1, 15.into()), 15);
+
+        // UTF-8 to UTF-16
+        assert_eq!(col_index.utf8_to_utf16_col(1, 22.into()), 20);
+
+        // UTF-16 to UTF-8, no changes
+        assert_eq!(col_index.utf16_to_utf8_col(1, 15), TextUnit::from(15));
+
+        // UTF-16 to UTF-8
+        assert_eq!(col_index.utf16_to_utf8_col(1, 19), TextUnit::from(21));
+    }
+
+    #[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-8 to UTF-16
+        assert_eq!(col_index.utf8_to_utf16_col(1, 15.into()), 15);
+
+        assert_eq!(col_index.utf8_to_utf16_col(1, 21.into()), 19);
+        assert_eq!(col_index.utf8_to_utf16_col(1, 25.into()), 21);
+
+        assert!(col_index.utf8_to_utf16_col(2, 15.into()) == 15);
+
+        // UTF-16 to UTF-8
+        assert_eq!(col_index.utf16_to_utf8_col(1, 15), TextUnit::from_usize(15));
+
+        assert_eq!(col_index.utf16_to_utf8_col(1, 18), TextUnit::from_usize(20));
+        assert_eq!(col_index.utf16_to_utf8_col(1, 19), TextUnit::from_usize(23));
+
+        assert_eq!(col_index.utf16_to_utf8_col(2, 15), TextUnit::from_usize(15));
+    }
+}

crates/ra_lsp_server/src/conv.rs

@@ -49,10 +49,9 @@ impl ConvWith for Position {
     type Output = TextUnit;
 
     fn conv_with(self, line_index: &LineIndex) -> TextUnit {
-        // TODO: UTF-16
         let line_col = LineCol {
             line: self.line as u32,
-            col: (self.character as u32).into(),
+            col_utf16: self.character as u32,
         };
         line_index.offset(line_col)
     }
@@ -64,8 +63,10 @@ impl ConvWith for TextUnit {
 
     fn conv_with(self, line_index: &LineIndex) -> Position {
         let line_col = line_index.line_col(self);
-        // TODO: UTF-16
-        Position::new(u64::from(line_col.line), u64::from(u32::from(line_col.col)))
+        Position::new(
+            u64::from(line_col.line),
+            u64::from(u32::from(line_col.col_utf16)),
+        )
     }
 }
 
@@ -203,8 +204,10 @@ impl TryConvWith for SourceChange {
                     .map(|it| it.edits.as_slice())
                     .unwrap_or(&[]);
                 let line_col = translate_offset_with_edit(&*line_index, pos.offset, edits);
-                let position =
-                    Position::new(u64::from(line_col.line), u64::from(u32::from(line_col.col)));
+                let position = Position::new(
+                    u64::from(line_col.line),
+                    u64::from(u32::from(line_col.col_utf16)),
+                );
                 Some(TextDocumentPositionParams {
                     text_document: TextDocumentIdentifier::new(pos.file_id.try_conv_with(world)?),
                     position,
@@ -247,12 +250,12 @@ fn translate_offset_with_edit(
     if in_edit_line_col.line == 0 {
         LineCol {
             line: edit_line_col.line,
-            col: edit_line_col.col + in_edit_line_col.col,
+            col_utf16: edit_line_col.col_utf16 + in_edit_line_col.col_utf16,
         }
     } else {
         LineCol {
             line: edit_line_col.line + in_edit_line_col.line,
-            col: in_edit_line_col.col,
+            col_utf16: in_edit_line_col.col_utf16,
         }
     }
 }