1use std::{fmt, ops::Range};
2
3
4/// An error signaling that a different kind of token was expected. Returned by
5/// the various `TryFrom` impls.
6#[derive(Debug, Clone, Copy)]
7pub struct InvalidToken {
8 pub(crate) expected: TokenKind,
9 pub(crate) actual: TokenKind,
10 pub(crate) span: Span,
11}
12
13impl InvalidToken {
14 /// Returns a token stream representing `compile_error!("msg");` where
15 /// `"msg"` is the output of `self.to_string()`. **Panics if called outside
16 /// of a proc-macro context!**
17 pub fn to_compile_error(&self) -> proc_macro::TokenStream {
18 use proc_macro::{Delimiter, Ident, Group, Punct, Spacing, TokenTree};
19
20 let span = match self.span {
21 Span::One(s) => s,
22 #[cfg(feature = "proc-macro2")]
23 Span::Two(s) => s.unwrap(),
24 };
25 let msg = self.to_string();
26 let tokens = vec![
27 TokenTree::from(Ident::new("compile_error", span)),
28 TokenTree::from(Punct::new('!', Spacing::Alone)),
29 TokenTree::from(Group::new(
30 Delimiter::Parenthesis,
31 TokenTree::from(proc_macro::Literal::string(&msg)).into(),
32 )),
33 ];
34
35
36 tokens.into_iter().map(|mut t| { t.set_span(span); t }).collect()
37 }
38
39 /// Like [`to_compile_error`][Self::to_compile_error], but returns a token
40 /// stream from `proc_macro2` and does not panic outside of a proc-macro
41 /// context.
42 #[cfg(feature = "proc-macro2")]
43 pub fn to_compile_error2(&self) -> proc_macro2::TokenStream {
44 use proc_macro2::{Delimiter, Ident, Group, Punct, Spacing, TokenTree};
45
46 let span = match self.span {
47 Span::One(s) => proc_macro2::Span::from(s),
48 Span::Two(s) => s,
49 };
50 let msg = self.to_string();
51 let tokens = vec![
52 TokenTree::from(Ident::new("compile_error", span)),
53 TokenTree::from(Punct::new('!', Spacing::Alone)),
54 TokenTree::from(Group::new(
55 Delimiter::Parenthesis,
56 TokenTree::from(proc_macro2::Literal::string(&msg)).into(),
57 )),
58 ];
59
60
61 tokens.into_iter().map(|mut t| { t.set_span(span); t }).collect()
62 }
63}
64
65impl std::error::Error for InvalidToken {}
66
67impl fmt::Display for InvalidToken {
68 fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
69 fn kind_desc(kind: TokenKind) -> &'static str {
70 match kind {
71 TokenKind::Punct => "a punctuation character",
72 TokenKind::Ident => "an identifier",
73 TokenKind::Group => "a group",
74 TokenKind::Literal => "a literal",
75 TokenKind::BoolLit => "a bool literal (`true` or `false`)",
76 TokenKind::ByteLit => "a byte literal (e.g. `b'r')",
77 TokenKind::ByteStringLit => r#"a byte string literal (e.g. `b"fox"`)"#,
78 TokenKind::CharLit => "a character literal (e.g. `'P'`)",
79 TokenKind::FloatLit => "a float literal (e.g. `3.14`)",
80 TokenKind::IntegerLit => "an integer literal (e.g. `27`)",
81 TokenKind::StringLit => r#"a string literal (e.g. "Ferris")"#,
82 }
83 }
84
85 write!(f, "expected {}, but found {}", kind_desc(self.expected), kind_desc(self.actual))
86 }
87}
88
89#[derive(Debug, Clone, Copy, PartialEq, Eq)]
90pub(crate) enum TokenKind {
91 Punct,
92 Ident,
93 Group,
94 Literal,
95 BoolLit,
96 ByteLit,
97 ByteStringLit,
98 CharLit,
99 FloatLit,
100 IntegerLit,
101 StringLit,
102}
103
104/// Unfortunately, we have to deal with both cases.
105#[derive(Debug, Clone, Copy)]
106pub(crate) enum Span {
107 One(proc_macro::Span),
108 #[cfg(feature = "proc-macro2")]
109 Two(proc_macro2::Span),
110}
111
112impl From<proc_macro::Span> for Span {
113 fn from(src: proc_macro::Span) -> Self {
114 Self::One(src)
115 }
116}
117
118#[cfg(feature = "proc-macro2")]
119impl From<proc_macro2::Span> for Span {
120 fn from(src: proc_macro2::Span) -> Self {
121 Self::Two(src)
122 }
123}
124
125/// Errors during parsing.
126///
127/// This type should be seen primarily for error reporting and not for catching
128/// specific cases. The span and error kind are not guaranteed to be stable
129/// over different versions of this library, meaning that a returned error can
130/// change from one version to the next. There are simply too many fringe cases
131/// that are not easy to classify as a specific error kind. It depends entirely
132/// on the specific parser code how an invalid input is categorized.
133///
134/// Consider these examples:
135/// - `'\` can be seen as
136/// - invalid escape in character literal, or
137/// - unterminated character literal.
138/// - `'''` can be seen as
139/// - empty character literal, or
140/// - unescaped quote character in character literal.
141/// - `0b64` can be seen as
142/// - binary integer literal with invalid digit 6, or
143/// - binary integer literal with invalid digit 4, or
144/// - decimal integer literal with invalid digit b, or
145/// - decimal integer literal 0 with unknown type suffix `b64`.
146///
147/// If you want to see more if these examples, feel free to check out the unit
148/// tests of this library.
149///
150/// While this library does its best to emit sensible and precise errors, and to
151/// keep the returned errors as stable as possible, full stability cannot be
152/// guaranteed.
153#[derive(Debug, Clone)]
154pub struct ParseError {
155 pub(crate) span: Option<Range<usize>>,
156 pub(crate) kind: ParseErrorKind,
157}
158
159impl ParseError {
160 /// Returns a span of this error, if available. **Note**: the returned span
161 /// might change in future versions of this library. See [the documentation
162 /// of this type][ParseError] for more information.
163 pub fn span(&self) -> Option<Range<usize>> {
164 self.span.clone()
165 }
166}
167
168/// This is a free standing function instead of an associated one to reduce
169/// noise around parsing code. There are lots of places that create errors, we
170/// I wanna keep them as short as possible.
171pub(crate) fn perr(span: impl SpanLike, kind: ParseErrorKind) -> ParseError {
172 ParseError {
173 span: span.into_span(),
174 kind,
175 }
176}
177
178pub(crate) trait SpanLike {
179 fn into_span(self) -> Option<Range<usize>>;
180}
181
182impl SpanLike for Option<Range<usize>> {
183 #[inline(always)]
184 fn into_span(self) -> Option<Range<usize>> {
185 self
186 }
187}
188impl SpanLike for Range<usize> {
189 #[inline(always)]
190 fn into_span(self) -> Option<Range<usize>> {
191 Some(self)
192 }
193}
194impl SpanLike for usize {
195 #[inline(always)]
196 fn into_span(self) -> Option<Range<usize>> {
197 Some(self..self + 1)
198 }
199}
200
201
202/// Kinds of errors.
203#[derive(Debug, Clone, Copy, PartialEq, Eq)]
204#[non_exhaustive]
205pub(crate) enum ParseErrorKind {
206 /// The input was an empty string
207 Empty,
208
209 /// An unexpected char was encountered.
210 UnexpectedChar,
211
212 /// Literal was not recognized.
213 InvalidLiteral,
214
215 /// Input does not start with decimal digit when trying to parse an integer.
216 DoesNotStartWithDigit,
217
218 /// A digit invalid for the specified integer base was found.
219 InvalidDigit,
220
221 /// Integer literal does not contain any valid digits.
222 NoDigits,
223
224 /// Exponent of a float literal does not contain any digits.
225 NoExponentDigits,
226
227 /// An unknown escape code, e.g. `\b`.
228 UnknownEscape,
229
230 /// A started escape sequence where the input ended before the escape was
231 /// finished.
232 UnterminatedEscape,
233
234 /// An `\x` escape where the two digits are not valid hex digits.
235 InvalidXEscape,
236
237 /// A string or character literal using the `\xNN` escape where `NN > 0x7F`.
238 NonAsciiXEscape,
239
240 /// A `\u{...}` escape in a byte or byte string literal.
241 UnicodeEscapeInByteLiteral,
242
243 /// A Unicode escape that does not start with a hex digit.
244 InvalidStartOfUnicodeEscape,
245
246 /// A `\u{...}` escape that lacks the opening brace.
247 UnicodeEscapeWithoutBrace,
248
249 /// In a `\u{...}` escape, a non-hex digit and non-underscore character was
250 /// found.
251 NonHexDigitInUnicodeEscape,
252
253 /// More than 6 digits found in unicode escape.
254 TooManyDigitInUnicodeEscape,
255
256 /// The value from a unicode escape does not represent a valid character.
257 InvalidUnicodeEscapeChar,
258
259 /// A `\u{..` escape that is not terminated (lacks the closing brace).
260 UnterminatedUnicodeEscape,
261
262 /// A character literal that's not terminated.
263 UnterminatedCharLiteral,
264
265 /// A character literal that contains more than one character.
266 OverlongCharLiteral,
267
268 /// An empty character literal, i.e. `''`.
269 EmptyCharLiteral,
270
271 UnterminatedByteLiteral,
272 OverlongByteLiteral,
273 EmptyByteLiteral,
274 NonAsciiInByteLiteral,
275
276 /// A `'` character was not escaped in a character or byte literal, or a `"`
277 /// character was not escaped in a string or byte string literal.
278 UnescapedSingleQuote,
279
280 /// A \n, \t or \r raw character in a char or byte literal.
281 UnescapedSpecialWhitespace,
282
283 /// When parsing a character, byte, string or byte string literal directly
284 /// and the input does not start with the corresponding quote character
285 /// (plus optional raw string prefix).
286 DoesNotStartWithQuote,
287
288 /// Unterminated raw string literal.
289 UnterminatedRawString,
290
291 /// String literal without a `"` at the end.
292 UnterminatedString,
293
294 /// Invalid start for a string literal.
295 InvalidStringLiteralStart,
296
297 /// Invalid start for a byte literal.
298 InvalidByteLiteralStart,
299
300 InvalidByteStringLiteralStart,
301
302 /// An literal `\r` character not followed by a `\n` character in a
303 /// (raw) string or byte string literal.
304 IsolatedCr,
305
306 /// Literal suffix is not a valid identifier.
307 InvalidSuffix,
308
309 /// Returned by `Float::parse` if an integer literal (no fractional nor
310 /// exponent part) is passed.
311 UnexpectedIntegerLit,
312
313 /// Integer suffixes cannot start with `e` or `E` as this conflicts with the
314 /// grammar for float literals.
315 IntegerSuffixStartingWithE,
316}
317
318impl std::error::Error for ParseError {}
319
320impl fmt::Display for ParseError {
321 fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
322 use ParseErrorKind::*;
323
324 let description = match self.kind {
325 Empty => "input is empty",
326 UnexpectedChar => "unexpected character",
327 InvalidLiteral => "invalid literal",
328 DoesNotStartWithDigit => "number literal does not start with decimal digit",
329 InvalidDigit => "integer literal contains a digit invalid for its base",
330 NoDigits => "integer literal does not contain any digits",
331 NoExponentDigits => "exponent of floating point literal does not contain any digits",
332 UnknownEscape => "unknown escape",
333 UnterminatedEscape => "unterminated escape: input ended too soon",
334 InvalidXEscape => r"invalid `\x` escape: not followed by two hex digits",
335 NonAsciiXEscape => r"`\x` escape in char/string literal exceed ASCII range",
336 UnicodeEscapeInByteLiteral => r"`\u{...}` escape in byte (string) literal not allowed",
337 InvalidStartOfUnicodeEscape => r"invalid start of `\u{...}` escape",
338 UnicodeEscapeWithoutBrace => r"`Unicode \u{...}` escape without opening brace",
339 NonHexDigitInUnicodeEscape => r"non-hex digit found in `\u{...}` escape",
340 TooManyDigitInUnicodeEscape => r"more than six digits in `\u{...}` escape",
341 InvalidUnicodeEscapeChar => r"value specified in `\u{...}` escape is not a valid char",
342 UnterminatedUnicodeEscape => r"unterminated `\u{...}` escape",
343 UnterminatedCharLiteral => "character literal is not terminated",
344 OverlongCharLiteral => "character literal contains more than one character",
345 EmptyCharLiteral => "empty character literal",
346 UnterminatedByteLiteral => "byte literal is not terminated",
347 OverlongByteLiteral => "byte literal contains more than one byte",
348 EmptyByteLiteral => "empty byte literal",
349 NonAsciiInByteLiteral => "non ASCII character in byte (string) literal",
350 UnescapedSingleQuote => "character literal contains unescaped ' character",
351 UnescapedSpecialWhitespace => r"unescaped newline (\n), tab (\t) or cr (\r) character",
352 DoesNotStartWithQuote => "invalid start for char/byte/string literal",
353 UnterminatedRawString => "unterminated raw (byte) string literal",
354 UnterminatedString => "unterminated (byte) string literal",
355 InvalidStringLiteralStart => "invalid start for string literal",
356 InvalidByteLiteralStart => "invalid start for byte literal",
357 InvalidByteStringLiteralStart => "invalid start for byte string literal",
358 IsolatedCr => r"`\r` not immediately followed by `\n` in string",
359 InvalidSuffix => "literal suffix is not a valid identifier",
360 UnexpectedIntegerLit => "expected float literal, but found integer",
361 IntegerSuffixStartingWithE => "integer literal suffix must not start with 'e' or 'E'",
362 };
363
364 description.fmt(f)?;
365 if let Some(span) = &self.span {
366 write!(f, " (at {}..{})", span.start, span.end)?;
367 }
368
369 Ok(())
370 }
371}
372