|
| 1 | +use crate::impl_::concat::slice_copy_from_slice; |
| 2 | + |
| 3 | +/// This is a reimplementation of the `indoc` crate's unindent functionality: |
| 4 | +/// |
| 5 | +/// 1. Count the leading spaces of each line, ignoring the first line and any lines that are empty or contain spaces only. |
| 6 | +/// 2. Take the minimum. |
| 7 | +/// 3. If the first line is empty i.e. the string begins with a newline, remove the first line. |
| 8 | +/// 4. Remove the computed number of spaces from the beginning of each line. |
| 9 | +const fn unindent_bytes(bytes: &mut [u8]) -> usize { |
| 10 | + // (1) + (2) - count leading spaces, take the minimum |
| 11 | + let Some(to_unindent) = get_minimum_leading_spaces(bytes) else { |
| 12 | + // all lines were empty, nothing to unindent |
| 13 | + return bytes.len(); |
| 14 | + }; |
| 15 | + |
| 16 | + // now copy from the original buffer, bringing values forward as needed |
| 17 | + let mut read_idx = 0; |
| 18 | + let mut write_idx = 0; |
| 19 | + |
| 20 | + // (3) - remove first line if it is empty |
| 21 | + match consume_eol(bytes, read_idx) { |
| 22 | + // skip empty first line |
| 23 | + Some(eol) => read_idx = eol, |
| 24 | + // copy non-empty first line as-is |
| 25 | + None => { |
| 26 | + (read_idx, write_idx) = copy_forward_until_eol(bytes, read_idx, write_idx); |
| 27 | + } |
| 28 | + }; |
| 29 | + |
| 30 | + // (4) - unindent remaining lines |
| 31 | + while read_idx < bytes.len() { |
| 32 | + let leading_spaces = count_spaces(bytes, read_idx); |
| 33 | + |
| 34 | + if leading_spaces < to_unindent { |
| 35 | + read_idx += leading_spaces; |
| 36 | + assert!( |
| 37 | + consume_eol(bytes, read_idx).is_some(), |
| 38 | + "removed fewer spaces than expected on non-empty line" |
| 39 | + ); |
| 40 | + } else { |
| 41 | + // leading_spaces may be equal to or larger than to_unindent, only need to unindent |
| 42 | + // the required amount, additional indentation is meaningful |
| 43 | + read_idx += to_unindent; |
| 44 | + } |
| 45 | + |
| 46 | + // copy remainder of line |
| 47 | + (read_idx, write_idx) = copy_forward_until_eol(bytes, read_idx, write_idx); |
| 48 | + } |
| 49 | + |
| 50 | + write_idx |
| 51 | +} |
| 52 | + |
| 53 | +/// Counts the minimum leading spaces of all non-empty lines except the first line. |
| 54 | +/// |
| 55 | +/// Returns `None` if there are no non-empty lines except the first line. |
| 56 | +const fn get_minimum_leading_spaces(bytes: &[u8]) -> Option<usize> { |
| 57 | + // scan for leading spaces (ignoring first line and empty lines) |
| 58 | + let mut i = 0; |
| 59 | + |
| 60 | + // skip first line |
| 61 | + i = advance_to_next_line(bytes, i); |
| 62 | + |
| 63 | + let mut to_unindent = None; |
| 64 | + |
| 65 | + // for remaining lines, count leading spaces |
| 66 | + while i < bytes.len() { |
| 67 | + let line_leading_spaces = count_spaces(bytes, i); |
| 68 | + i += line_leading_spaces; |
| 69 | + |
| 70 | + // line only had spaces, ignore for the count |
| 71 | + if let Some(eol) = consume_eol(bytes, i) { |
| 72 | + i = eol; |
| 73 | + continue; |
| 74 | + } |
| 75 | + |
| 76 | + // this line has content, consider its leading spaces |
| 77 | + if let Some(current) = to_unindent { |
| 78 | + // .unwrap_or(usize::MAX) not available in const fn |
| 79 | + if line_leading_spaces < current { |
| 80 | + to_unindent = Some(line_leading_spaces); |
| 81 | + } |
| 82 | + } else { |
| 83 | + to_unindent = Some(line_leading_spaces); |
| 84 | + } |
| 85 | + |
| 86 | + i = advance_to_next_line(bytes, i); |
| 87 | + } |
| 88 | + |
| 89 | + to_unindent |
| 90 | +} |
| 91 | + |
| 92 | +const fn advance_to_next_line(bytes: &[u8], mut i: usize) -> usize { |
| 93 | + while i < bytes.len() { |
| 94 | + if let Some(eol) = consume_eol(bytes, i) { |
| 95 | + return eol; |
| 96 | + } |
| 97 | + i += 1; |
| 98 | + } |
| 99 | + i |
| 100 | +} |
| 101 | + |
| 102 | +/// Brings elements in `bytes` forward until `\n` (inclusive) or end of `source`. |
| 103 | +/// |
| 104 | +/// `read_idx` must be greater than or equal to `write_idx`. |
| 105 | +const fn copy_forward_until_eol( |
| 106 | + bytes: &mut [u8], |
| 107 | + mut read_idx: usize, |
| 108 | + mut write_idx: usize, |
| 109 | +) -> (usize, usize) { |
| 110 | + assert!(read_idx >= write_idx); |
| 111 | + while read_idx < bytes.len() { |
| 112 | + let value = bytes[read_idx]; |
| 113 | + bytes[write_idx] = value; |
| 114 | + read_idx += 1; |
| 115 | + write_idx += 1; |
| 116 | + if value == b'\n' { |
| 117 | + break; |
| 118 | + } |
| 119 | + } |
| 120 | + (read_idx, write_idx) |
| 121 | +} |
| 122 | + |
| 123 | +const fn count_spaces(bytes: &[u8], mut i: usize) -> usize { |
| 124 | + let mut count = 0; |
| 125 | + while i < bytes.len() && bytes[i] == b' ' { |
| 126 | + count += 1; |
| 127 | + i += 1; |
| 128 | + } |
| 129 | + count |
| 130 | +} |
| 131 | + |
| 132 | +const fn consume_eol(bytes: &[u8], i: usize) -> Option<usize> { |
| 133 | + if bytes.len() == i { |
| 134 | + // special case: treat end of buffer as EOL without consuming anything |
| 135 | + Some(i) |
| 136 | + } else if bytes.len() > i && bytes[i] == b'\n' { |
| 137 | + Some(i + 1) |
| 138 | + } else if bytes[i] == b'\r' && bytes.len() > i + 1 && bytes[i + 1] == b'\n' { |
| 139 | + Some(i + 2) |
| 140 | + } else { |
| 141 | + None |
| 142 | + } |
| 143 | +} |
| 144 | + |
| 145 | +pub const fn unindent_sized<const N: usize>(src: &[u8]) -> ([u8; N], usize) { |
| 146 | + let mut out: [u8; N] = [0; N]; |
| 147 | + slice_copy_from_slice(&mut out, src); |
| 148 | + let new_len = unindent_bytes(&mut out); |
| 149 | + (out, new_len) |
| 150 | +} |
| 151 | + |
| 152 | +/// Helper for `py_run!` macro which unindents a string at compile time. |
| 153 | +#[macro_export] |
| 154 | +#[doc(hidden)] |
| 155 | +macro_rules! unindent { |
| 156 | + ($value:expr) => {{ |
| 157 | + const RAW: &str = $value; |
| 158 | + const LEN: usize = RAW.len(); |
| 159 | + const UNINDENTED: ([u8; LEN], usize) = |
| 160 | + $crate::impl_::unindent::unindent_sized::<LEN>(RAW.as_bytes()); |
| 161 | + // SAFETY: this removes only spaces and preserves all other contents |
| 162 | + unsafe { ::core::str::from_utf8_unchecked(UNINDENTED.0.split_at(UNINDENTED.1).0) } |
| 163 | + }}; |
| 164 | +} |
| 165 | + |
| 166 | +pub use crate::unindent; |
| 167 | + |
| 168 | +/// Equivalent of the `unindent!` macro, but works at runtime. |
| 169 | +pub fn unindent(s: &str) -> String { |
| 170 | + let mut bytes = s.as_bytes().to_owned(); |
| 171 | + let unindented_size = unindent_bytes(&mut bytes); |
| 172 | + bytes.resize(unindented_size, 0); |
| 173 | + String::from_utf8(bytes).unwrap() |
| 174 | +} |
| 175 | + |
| 176 | +#[cfg(test)] |
| 177 | +mod tests { |
| 178 | + use super::*; |
| 179 | + |
| 180 | + const SAMPLE_1_WITH_FIRST_LINE: &str = " first line |
| 181 | + line one |
| 182 | +
|
| 183 | + line two |
| 184 | + "; |
| 185 | + |
| 186 | + const UNINDENTED_1: &str = " first line\nline one\n\n line two\n"; |
| 187 | + |
| 188 | + const SAMPLE_2_EMPTY_FIRST_LINE: &str = " |
| 189 | + line one |
| 190 | +
|
| 191 | + line two |
| 192 | + "; |
| 193 | + const UNINDENTED_2: &str = "line one\n\n line two\n"; |
| 194 | + |
| 195 | + const SAMPLE_3_NO_INDENT: &str = " |
| 196 | +no indent |
| 197 | + here"; |
| 198 | + |
| 199 | + const UNINDENTED_3: &str = "no indent\n here"; |
| 200 | + |
| 201 | + const SAMPLE_4_NOOP: &str = "no indent\nhere\n but here"; |
| 202 | + |
| 203 | + const SAMPLE_5_EMPTY: &str = " \n \n"; |
| 204 | + |
| 205 | + const ALL_CASES: &[(&str, &str)] = &[ |
| 206 | + (SAMPLE_1_WITH_FIRST_LINE, UNINDENTED_1), |
| 207 | + (SAMPLE_2_EMPTY_FIRST_LINE, UNINDENTED_2), |
| 208 | + (SAMPLE_3_NO_INDENT, UNINDENTED_3), |
| 209 | + (SAMPLE_4_NOOP, SAMPLE_4_NOOP), |
| 210 | + (SAMPLE_5_EMPTY, SAMPLE_5_EMPTY), |
| 211 | + ]; |
| 212 | + |
| 213 | + // run const tests for each sample to ensure they work at compile time |
| 214 | + |
| 215 | + #[test] |
| 216 | + fn test_unindent_const() { |
| 217 | + const UNINDENTED: &str = unindent!(SAMPLE_1_WITH_FIRST_LINE); |
| 218 | + assert_eq!(UNINDENTED, UNINDENTED_1); |
| 219 | + } |
| 220 | + |
| 221 | + #[test] |
| 222 | + fn test_unindent_const_removes_empty_first_line() { |
| 223 | + const UNINDENTED: &str = unindent!(SAMPLE_2_EMPTY_FIRST_LINE); |
| 224 | + assert_eq!(UNINDENTED, UNINDENTED_2); |
| 225 | + } |
| 226 | + |
| 227 | + #[test] |
| 228 | + fn test_unindent_const_no_indent() { |
| 229 | + const UNINDENTED: &str = unindent!(SAMPLE_3_NO_INDENT); |
| 230 | + assert_eq!(UNINDENTED, UNINDENTED_3); |
| 231 | + } |
| 232 | + |
| 233 | + #[test] |
| 234 | + fn test_unindent_macro_runtime() { |
| 235 | + // this variation on the test ensures full coverage (const eval not included in coverage) |
| 236 | + const INDENTED: &str = SAMPLE_1_WITH_FIRST_LINE; |
| 237 | + const LEN: usize = INDENTED.len(); |
| 238 | + let (unindented, unindented_size) = unindent_sized::<LEN>(INDENTED.as_bytes()); |
| 239 | + let unindented = std::str::from_utf8(&unindented[..unindented_size]).unwrap(); |
| 240 | + assert_eq!(unindented, UNINDENTED_1); |
| 241 | + } |
| 242 | + |
| 243 | + #[test] |
| 244 | + fn test_unindent_function() { |
| 245 | + for (indented, expected) in ALL_CASES { |
| 246 | + let unindented = unindent(indented); |
| 247 | + assert_eq!(&unindented, expected); |
| 248 | + } |
| 249 | + } |
| 250 | +} |
0 commit comments