|
| 1 | +use ropey::RopeSlice; |
| 2 | + |
| 3 | +/// Given a slice of text, return the text re-wrapped to fit |
| 4 | +/// it within the given width. |
| 5 | +pub fn reflow_hard_wrap(text: RopeSlice, max_line_len: usize) -> String { |
| 6 | + // TODO: We should handle CRLF line endings. |
| 7 | + |
| 8 | + let text = String::from(text); |
| 9 | + let prefix = detect_prefix(&text); |
| 10 | + |
| 11 | + // The current algorithm eats a single trailing newline character. This |
| 12 | + // is a hacky way to put it back at the end. Ideally there would be a |
| 13 | + // cleaner way to do this. It also happens to matter a lot because Helix |
| 14 | + // currently selects lines *including* the trailing newline. |
| 15 | + let ends_with_newline = text.chars().rev().next() == Some('\n'); |
| 16 | + |
| 17 | + let flow_pieces = separate_flow_pieces(&text, &prefix).into_iter().peekable(); |
| 18 | + |
| 19 | + let mut new_flow = String::new(); |
| 20 | + let mut current_line_len: usize = 0; |
| 21 | + |
| 22 | + for piece in flow_pieces { |
| 23 | + match piece { |
| 24 | + Piece::NoReflow(piece) => { |
| 25 | + new_flow.push('\n'); |
| 26 | + new_flow.push_str(piece); |
| 27 | + current_line_len = 0; |
| 28 | + continue; |
| 29 | + } |
| 30 | + Piece::Reflow(piece) => { |
| 31 | + use unicode_segmentation::UnicodeSegmentation; |
| 32 | + let piece_len = UnicodeSegmentation::graphemes(piece, true).count(); |
| 33 | + |
| 34 | + let piece_will_fit = current_line_len + piece_len <= max_line_len; |
| 35 | + if !piece_will_fit && !new_flow.is_empty() { |
| 36 | + new_flow.push('\n'); |
| 37 | + current_line_len = 0; |
| 38 | + } |
| 39 | + |
| 40 | + if current_line_len == 0 { |
| 41 | + new_flow.push_str(&prefix); |
| 42 | + } |
| 43 | + |
| 44 | + new_flow.push_str(piece); |
| 45 | + |
| 46 | + current_line_len += piece_len; |
| 47 | + } |
| 48 | + } |
| 49 | + } |
| 50 | + |
| 51 | + if ends_with_newline { |
| 52 | + new_flow.push('\n'); |
| 53 | + } |
| 54 | + |
| 55 | + new_flow |
| 56 | +} |
| 57 | + |
| 58 | +#[cfg_attr(test, derive(Debug, PartialEq))] |
| 59 | +enum Piece<'a> { |
| 60 | + Reflow(&'a str), |
| 61 | + NoReflow(&'a str), |
| 62 | +} |
| 63 | + |
| 64 | +fn detect_prefix(text: &str) -> String { |
| 65 | + // TODO: We should detect the configured comment prefixes here. |
| 66 | + |
| 67 | + if text.is_empty() { |
| 68 | + return String::new(); |
| 69 | + } |
| 70 | + |
| 71 | + // For now, we'll only detect the prefix by the string (possibly empty) |
| 72 | + // that is present at the beginning of every selected line, excluding |
| 73 | + // lines that are entirely blank. |
| 74 | + |
| 75 | + // UNWRAP: This is OK because we already checked to see if the string is empty. |
| 76 | + let first_line = text.lines().next().unwrap(); |
| 77 | + text.lines() |
| 78 | + // TODO(decide): Should `is_whitespace` be `is_ascii_whitespace` in this next line? |
| 79 | + .filter(|line| !line.chars().all(|ch| ch.is_whitespace())) |
| 80 | + .fold(first_line, |prefix, line| { |
| 81 | + // If the prefix is already empty, we won't bother checking what |
| 82 | + // portion of it matches with the current line. |
| 83 | + if prefix.is_empty() { |
| 84 | + return prefix; |
| 85 | + } |
| 86 | + |
| 87 | + let matched_until_excl: usize = prefix |
| 88 | + .chars() |
| 89 | + .enumerate() |
| 90 | + .zip(line.chars()) |
| 91 | + .take_while(|((_prefix_char_idx, prefix_char), line_char)| { |
| 92 | + // TODO(decide): Do we want to relax the restriction that the "prefix" |
| 93 | + // for these lines must be entirely non-ascii-alphanumeric? |
| 94 | + prefix_char == line_char && !prefix_char.is_ascii_alphanumeric() |
| 95 | + }) |
| 96 | + // add 1 here to be "one past the end" of the prefix |
| 97 | + .map(|((prefix_char_idx, _), _)| prefix_char_idx + 1) |
| 98 | + .last() |
| 99 | + .unwrap_or(0); |
| 100 | + |
| 101 | + prefix.get(..matched_until_excl).unwrap() |
| 102 | + }) |
| 103 | + .into() |
| 104 | +} |
| 105 | + |
| 106 | +// PANIC: This function may panic if the `prefix` does not begin every non- |
| 107 | +// blank line, where "blank" here means a line that has non-whitespace |
| 108 | +// characters *after* the prefix. A line with only the prefix and whitespace |
| 109 | +// characters is considered blank. |
| 110 | +fn separate_flow_pieces<'a>(text: &'a str, prefix: &str) -> Vec<Piece<'a>> { |
| 111 | + use separated::Separated; |
| 112 | + |
| 113 | + let mut flow_pieces = Vec::<Piece>::new(); |
| 114 | + let lines = Separated::new(text, "\n"); |
| 115 | + for line in lines { |
| 116 | + match line.get(..prefix.len()) { |
| 117 | + Some(possible_prefix) if possible_prefix == prefix => { |
| 118 | + // If the rest of the line is whitespace, we count it as an |
| 119 | + // effectively "blank" line, and we don't want to reflow it. |
| 120 | + if line[prefix.len()..].chars().all(char::is_whitespace) { |
| 121 | + flow_pieces.push(Piece::NoReflow(line)); |
| 122 | + } else { |
| 123 | + // UNWRAP: The `detect_prefix` function should ensure that this |
| 124 | + // unwrap is valid. The "prefix" should begin every line except |
| 125 | + // for entirely blank lines. |
| 126 | + let line_no_prefix = line.get(prefix.len()..).unwrap(); |
| 127 | + let new_pieces = Separated::new(line_no_prefix, " ") |
| 128 | + .map(|s| Piece::Reflow(s.trim_end_matches('\n').trim_end_matches('\r'))); |
| 129 | + flow_pieces.extend(new_pieces); |
| 130 | + } |
| 131 | + } |
| 132 | + _ => { |
| 133 | + assert!( |
| 134 | + line.chars().all(char::is_whitespace), |
| 135 | + "lines not matching the prefix should be blank" |
| 136 | + ); |
| 137 | + |
| 138 | + flow_pieces.push(Piece::NoReflow(line)); |
| 139 | + } |
| 140 | + } |
| 141 | + } |
| 142 | + |
| 143 | + flow_pieces |
| 144 | +} |
| 145 | + |
| 146 | +mod separated { |
| 147 | + // Like the std::str::Lines iterator except it doesn't eat the newline characters. |
| 148 | + pub(crate) struct Separated<'a, 'b> { |
| 149 | + s: &'a str, |
| 150 | + pattern: &'b str, |
| 151 | + beg: usize, |
| 152 | + } |
| 153 | + |
| 154 | + impl<'a, 'b> Separated<'a, 'b> { |
| 155 | + pub(crate) fn new(s: &'a str, pattern: &'b str) -> Self { |
| 156 | + Self { s, pattern, beg: 0 } |
| 157 | + } |
| 158 | + } |
| 159 | + |
| 160 | + impl<'a, 'b> Iterator for Separated<'a, 'b> { |
| 161 | + type Item = &'a str; |
| 162 | + |
| 163 | + fn next(&mut self) -> Option<Self::Item> { |
| 164 | + let rest = self.s.get(self.beg..)?; |
| 165 | + |
| 166 | + if rest.is_empty() { |
| 167 | + return None; |
| 168 | + } |
| 169 | + |
| 170 | + let end = rest |
| 171 | + .find(self.pattern) |
| 172 | + .unwrap_or_else(|| rest.len().saturating_sub(1)); |
| 173 | + |
| 174 | + self.beg += end + 1; |
| 175 | + Some(&rest[..=end]) |
| 176 | + } |
| 177 | + } |
| 178 | +} |
| 179 | + |
| 180 | +#[cfg(test)] |
| 181 | +mod tests { |
| 182 | + use super::*; |
| 183 | + use ropey::Rope; |
| 184 | + |
| 185 | + #[test] |
| 186 | + fn reflow_basic_to_one_line() { |
| 187 | + let text = "hello my name \nis helix"; |
| 188 | + let text = Rope::from(text); |
| 189 | + let reflow = reflow_hard_wrap(text.slice(..), 100); |
| 190 | + assert_eq!(reflow, "hello my name is helix"); |
| 191 | + } |
| 192 | + |
| 193 | + #[test] |
| 194 | + fn reflow_basic_to_many_lines() { |
| 195 | + let text = "hello my name is helix"; |
| 196 | + let text = Rope::from(text); |
| 197 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 198 | + assert_eq!(reflow, "hello my \nname is \nhelix"); |
| 199 | + } |
| 200 | + |
| 201 | + #[test] |
| 202 | + fn reflow_with_blank_empty_line() { |
| 203 | + let text = "hello\n\nmy name is helix"; |
| 204 | + let text = Rope::from(text); |
| 205 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 206 | + assert_eq!(reflow, "hello\n\nmy name \nis helix"); |
| 207 | + } |
| 208 | + |
| 209 | + #[test] |
| 210 | + fn reflow_with_blank_whitespace_line() { |
| 211 | + let text = "hello\n \nmy name is helix"; |
| 212 | + let text = Rope::from(text); |
| 213 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 214 | + assert_eq!(reflow, "hello\n \nmy name \nis helix"); |
| 215 | + } |
| 216 | + |
| 217 | + #[test] |
| 218 | + fn reflow_end_with_blank_line() { |
| 219 | + let text = "hello my name is helix\n"; |
| 220 | + let text = Rope::from(text); |
| 221 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 222 | + assert_eq!(reflow, "hello my \nname is \nhelix\n"); |
| 223 | + } |
| 224 | + |
| 225 | + #[test] |
| 226 | + fn reflow_with_blank_lines_and_prefix() { |
| 227 | + let text = " hello\n\nmy name is helix"; |
| 228 | + let text = Rope::from(text); |
| 229 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 230 | + assert_eq!(reflow, " hello\n\nmy name \nis helix"); |
| 231 | + } |
| 232 | + |
| 233 | + #[test] |
| 234 | + fn reflow_to_many_lines_with_whitespace_prefix() { |
| 235 | + let text = Rope::from("\t Text indented. \n\t Still indented."); |
| 236 | + let expected_reflow = "\t Text indented. Still indented."; |
| 237 | + let reflow = reflow_hard_wrap(text.slice(..), 80); |
| 238 | + assert_eq!(reflow, expected_reflow); |
| 239 | + } |
| 240 | + |
| 241 | + #[test] |
| 242 | + fn reflow_to_many_lines_with_whitespace_and_comment_prefix() { |
| 243 | + let text = Rope::from("// Text indented. \n// Still indented."); |
| 244 | + let expected_reflow = "// Text indented. Still indented."; |
| 245 | + let text = Rope::from(text); |
| 246 | + let reflow = reflow_hard_wrap(text.slice(..), 80); |
| 247 | + assert_eq!(reflow, expected_reflow); |
| 248 | + } |
| 249 | + |
| 250 | + #[test] |
| 251 | + fn reflow_empty() { |
| 252 | + let text = Rope::from(""); |
| 253 | + let reflow = reflow_hard_wrap(text.slice(..), 10); |
| 254 | + assert_eq!(reflow, ""); |
| 255 | + } |
| 256 | + |
| 257 | + #[test] |
| 258 | + fn reflow_max_line_length_zero() { |
| 259 | + let text = "hello my name is helix"; |
| 260 | + let text = Rope::from(text); |
| 261 | + let reflow = reflow_hard_wrap(text.slice(..), 0); |
| 262 | + assert_eq!(reflow, "hello \nmy \nname \nis \nhelix"); |
| 263 | + } |
| 264 | + |
| 265 | + #[test] |
| 266 | + fn reflow_comment_after_blank_line() { |
| 267 | + let text = Rope::from("// Text indented. \n\n// Still indented."); |
| 268 | + let expected_reflow = "// Text indented. \n\n// Still indented."; |
| 269 | + let text = Rope::from(text); |
| 270 | + let reflow = reflow_hard_wrap(text.slice(..), 80); |
| 271 | + assert_eq!(reflow, expected_reflow); |
| 272 | + } |
| 273 | + |
| 274 | + #[test] |
| 275 | + fn detect_prefix_no_indent() { |
| 276 | + let text = Rope::from("hello my name is helix"); |
| 277 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 278 | + assert_eq!(prefix, ""); |
| 279 | + } |
| 280 | + |
| 281 | + #[test] |
| 282 | + fn detect_prefix_spaces_indent() { |
| 283 | + let text = Rope::from(" hello my name is helix"); |
| 284 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 285 | + assert_eq!(prefix, " "); |
| 286 | + } |
| 287 | + |
| 288 | + #[test] |
| 289 | + fn detect_prefix_tabs_indent() { |
| 290 | + let text = Rope::from("\t\t\thello my name is helix"); |
| 291 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 292 | + assert_eq!(prefix, "\t\t\t"); |
| 293 | + } |
| 294 | + |
| 295 | + #[test] |
| 296 | + fn detect_prefix_spaces_with_tabs() { |
| 297 | + let text = Rope::from(" \t\thello my name is helix"); |
| 298 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 299 | + assert_eq!(prefix, " \t\t"); |
| 300 | + } |
| 301 | + |
| 302 | + #[test] |
| 303 | + fn detect_prefix_tabs_with_spaces_indent() { |
| 304 | + let text = Rope::from("\t\t hello my name is helix"); |
| 305 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 306 | + assert_eq!(prefix, "\t\t "); |
| 307 | + } |
| 308 | + |
| 309 | + #[test] |
| 310 | + fn detect_prefix_many_lines_with_comment_then_space() { |
| 311 | + let text = Rope::from("// Text indented.\n// Still indented."); |
| 312 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 313 | + assert_eq!(prefix, "// "); |
| 314 | + } |
| 315 | + |
| 316 | + #[test] |
| 317 | + fn detect_prefix_unfinished_final_line() { |
| 318 | + let text = Rope::from("// Text indented.\n// Still indented.\n"); |
| 319 | + let prefix = detect_prefix(text.to_string().as_str()); |
| 320 | + assert_eq!(prefix, "// "); |
| 321 | + } |
| 322 | + |
| 323 | + #[test] |
| 324 | + fn flow_pieces_basic() { |
| 325 | + let text = "one two three"; |
| 326 | + let pieces = separate_flow_pieces(text, ""); |
| 327 | + assert_eq!( |
| 328 | + pieces, |
| 329 | + Vec::from([ |
| 330 | + Piece::Reflow("one "), |
| 331 | + Piece::Reflow("two "), |
| 332 | + Piece::Reflow("three") |
| 333 | + ]) |
| 334 | + ); |
| 335 | + } |
| 336 | + |
| 337 | + #[test] |
| 338 | + fn flow_pieces_with_blank_line() { |
| 339 | + let text = "one\n\ntwo"; |
| 340 | + let pieces = separate_flow_pieces(text, ""); |
| 341 | + assert_eq!( |
| 342 | + pieces, |
| 343 | + Vec::from([ |
| 344 | + Piece::Reflow("one"), |
| 345 | + Piece::NoReflow("\n"), |
| 346 | + Piece::Reflow("two"), |
| 347 | + ]) |
| 348 | + ); |
| 349 | + } |
| 350 | +} |
0 commit comments