Revision control

Copy as Markdown

Other Tools

use std::{fmt, ops::Range};
/// An error signaling that a different kind of token was expected. Returned by
/// the various `TryFrom` impls.
#[derive(Debug, Clone, Copy)]
pub struct InvalidToken {
pub(crate) expected: TokenKind,
pub(crate) actual: TokenKind,
pub(crate) span: Span,
}
impl InvalidToken {
/// Returns a token stream representing `compile_error!("msg");` where
/// `"msg"` is the output of `self.to_string()`. **Panics if called outside
/// of a proc-macro context!**
pub fn to_compile_error(&self) -> proc_macro::TokenStream {
use proc_macro::{Delimiter, Ident, Group, Punct, Spacing, TokenTree};
let span = match self.span {
Span::One(s) => s,
#[cfg(feature = "proc-macro2")]
Span::Two(s) => s.unwrap(),
};
let msg = self.to_string();
let tokens = vec![
TokenTree::from(Ident::new("compile_error", span)),
TokenTree::from(Punct::new('!', Spacing::Alone)),
TokenTree::from(Group::new(
Delimiter::Parenthesis,
TokenTree::from(proc_macro::Literal::string(&msg)).into(),
)),
];
tokens.into_iter().map(|mut t| { t.set_span(span); t }).collect()
}
/// Like [`to_compile_error`][Self::to_compile_error], but returns a token
/// stream from `proc_macro2` and does not panic outside of a proc-macro
/// context.
#[cfg(feature = "proc-macro2")]
pub fn to_compile_error2(&self) -> proc_macro2::TokenStream {
use proc_macro2::{Delimiter, Ident, Group, Punct, Spacing, TokenTree};
let span = match self.span {
Span::One(s) => proc_macro2::Span::from(s),
Span::Two(s) => s,
};
let msg = self.to_string();
let tokens = vec![
TokenTree::from(Ident::new("compile_error", span)),
TokenTree::from(Punct::new('!', Spacing::Alone)),
TokenTree::from(Group::new(
Delimiter::Parenthesis,
TokenTree::from(proc_macro2::Literal::string(&msg)).into(),
)),
];
tokens.into_iter().map(|mut t| { t.set_span(span); t }).collect()
}
}
impl std::error::Error for InvalidToken {}
impl fmt::Display for InvalidToken {
fn fmt(&self, f: &mut fmt::Formatter<'_>) -> fmt::Result {
fn kind_desc(kind: TokenKind) -> &'static str {
match kind {
TokenKind::Punct => "a punctuation character",
TokenKind::Ident => "an identifier",
TokenKind::Group => "a group",
TokenKind::Literal => "a literal",
TokenKind::BoolLit => "a bool literal (`true` or `false`)",
TokenKind::ByteLit => "a byte literal (e.g. `b'r')",
TokenKind::ByteStringLit => r#"a byte string literal (e.g. `b"fox"`)"#,
TokenKind::CharLit => "a character literal (e.g. `'P'`)",
TokenKind::FloatLit => "a float literal (e.g. `3.14`)",
TokenKind::IntegerLit => "an integer literal (e.g. `27`)",
TokenKind::StringLit => r#"a string literal (e.g. "Ferris")"#,
}
}
write!(f, "expected {}, but found {}", kind_desc(self.expected), kind_desc(self.actual))
}
}
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
pub(crate) enum TokenKind {
Punct,
Ident,
Group,
Literal,
BoolLit,
ByteLit,
ByteStringLit,
CharLit,
FloatLit,
IntegerLit,
StringLit,
}
/// Unfortunately, we have to deal with both cases.
#[derive(Debug, Clone, Copy)]
pub(crate) enum Span {
One(proc_macro::Span),
#[cfg(feature = "proc-macro2")]
Two(proc_macro2::Span),
}
impl From<proc_macro::Span> for Span {
fn from(src: proc_macro::Span) -> Self {
Self::One(src)
}
}
#[cfg(feature = "proc-macro2")]
impl From<proc_macro2::Span> for Span {
fn from(src: proc_macro2::Span) -> Self {
Self::Two(src)
}
}
/// Errors during parsing.
///
/// This type should be seen primarily for error reporting and not for catching
/// specific cases. The span and error kind are not guaranteed to be stable
/// over different versions of this library, meaning that a returned error can
/// change from one version to the next. There are simply too many fringe cases
/// that are not easy to classify as a specific error kind. It depends entirely
/// on the specific parser code how an invalid input is categorized.
///
/// Consider these examples:
/// - `'\` can be seen as
/// - invalid escape in character literal, or
/// - unterminated character literal.
/// - `'''` can be seen as
/// - empty character literal, or
/// - unescaped quote character in character literal.
/// - `0b64` can be seen as
/// - binary integer literal with invalid digit 6, or
/// - binary integer literal with invalid digit 4, or
/// - decimal integer literal with invalid digit b, or
/// - decimal integer literal 0 with unknown type suffix `b64`.
///
/// If you want to see more if these examples, feel free to check out the unit
/// tests of this library.
///
/// While this library does its best to emit sensible and precise errors, and to
/// keep the returned errors as stable as possible, full stability cannot be
/// guaranteed.
#[derive(Debug, Clone)]
pub struct ParseError {
pub(crate) span: Option<Range<usize>>,
pub(crate) kind: ParseErrorKind,
}
impl ParseError {
/// Returns a span of this error, if available. **Note**: the returned span
/// might change in future versions of this library. See [the documentation
/// of this type][ParseError] for more information.
pub fn span(&self) -> Option<Range<usize>> {
self.span.clone()
}
}
/// This is a free standing function instead of an associated one to reduce
/// noise around parsing code. There are lots of places that create errors, we
/// I wanna keep them as short as possible.
pub(crate) fn perr(span: impl SpanLike, kind: ParseErrorKind) -> ParseError {
ParseError {
span: span.into_span(),
kind,
}
}
pub(crate) trait SpanLike {
fn into_span(self) -> Option<Range<usize>>;
}
impl SpanLike for Option<Range<usize>> {
#[inline(always)]
fn into_span(self) -> Option<Range<usize>> {
self
}
}
impl SpanLike for Range<usize> {
#[inline(always)]
fn into_span(self) -> Option<Range<usize>> {
Some(self)
}
}
impl SpanLike for usize {
#[inline(always)]
fn into_span(self) -> Option<Range<usize>> {
Some(self..self + 1)
}
}
/// Kinds of errors.
#[derive(Debug, Clone, Copy, PartialEq, Eq)]
#[non_exhaustive]
pub(crate) enum ParseErrorKind {
/// The input was an empty string
Empty,
/// An unexpected char was encountered.
UnexpectedChar,
/// Literal was not recognized.
InvalidLiteral,
/// Input does not start with decimal digit when trying to parse an integer.
DoesNotStartWithDigit,
/// A digit invalid for the specified integer base was found.
InvalidDigit,
/// Integer literal does not contain any valid digits.
NoDigits,
/// Exponent of a float literal does not contain any digits.
NoExponentDigits,
/// An unknown escape code, e.g. `\b`.
UnknownEscape,
/// A started escape sequence where the input ended before the escape was
/// finished.
UnterminatedEscape,
/// An `\x` escape where the two digits are not valid hex digits.
InvalidXEscape,
/// A string or character literal using the `\xNN` escape where `NN > 0x7F`.
NonAsciiXEscape,
/// A `\u{...}` escape in a byte or byte string literal.
UnicodeEscapeInByteLiteral,
/// A Unicode escape that does not start with a hex digit.
InvalidStartOfUnicodeEscape,
/// A `\u{...}` escape that lacks the opening brace.
UnicodeEscapeWithoutBrace,
/// In a `\u{...}` escape, a non-hex digit and non-underscore character was
/// found.
NonHexDigitInUnicodeEscape,
/// More than 6 digits found in unicode escape.
TooManyDigitInUnicodeEscape,
/// The value from a unicode escape does not represent a valid character.
InvalidUnicodeEscapeChar,
/// A `\u{..` escape that is not terminated (lacks the closing brace).
UnterminatedUnicodeEscape,
/// A character literal that's not terminated.
UnterminatedCharLiteral,
/// A character literal that contains more than one character.
OverlongCharLiteral,
/// An empty character literal, i.e. `''`.
EmptyCharLiteral,
UnterminatedByteLiteral,
OverlongByteLiteral,
EmptyByteLiteral,
NonAsciiInByteLiteral,
/// A `'` character was not escaped in a character or byte literal, or a `"`
/// character was not escaped in a string or byte string literal.
UnescapedSingleQuote,
/// A \n, \t or \r raw character in a char or byte literal.
UnescapedSpecialWhitespace,
/// When parsing a character, byte, string or byte string literal directly
/// and the input does not start with the corresponding quote character
/// (plus optional raw string prefix).
DoesNotStartWithQuote,
/// Unterminated raw string literal.
UnterminatedRawString,
/// String literal without a `"` at the end.
UnterminatedString,
/// Invalid start for a string literal.
InvalidStringLiteralStart,
/// Invalid start for a byte literal.
InvalidByteLiteralStart,
InvalidByteStringLiteralStart,
/// An literal `\r` character not followed by a `\n` character in a
/// (raw) string or byte string literal.
IsolatedCr,
/// Literal suffix is not a valid identifier.
InvalidSuffix,
/// Returned by `Float::parse` if an integer literal (no fractional nor
/// exponent part) is passed.
UnexpectedIntegerLit,
/// Integer suffixes cannot start with `e` or `E` as this conflicts with the
/// grammar for float literals.
IntegerSuffixStartingWithE,
}
impl std::error::Error for ParseError {}
impl fmt::Display for ParseError {
fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
use ParseErrorKind::*;
let description = match self.kind {
Empty => "input is empty",
UnexpectedChar => "unexpected character",
InvalidLiteral => "invalid literal",
DoesNotStartWithDigit => "number literal does not start with decimal digit",
InvalidDigit => "integer literal contains a digit invalid for its base",
NoDigits => "integer literal does not contain any digits",
NoExponentDigits => "exponent of floating point literal does not contain any digits",
UnknownEscape => "unknown escape",
UnterminatedEscape => "unterminated escape: input ended too soon",
InvalidXEscape => r"invalid `\x` escape: not followed by two hex digits",
NonAsciiXEscape => r"`\x` escape in char/string literal exceed ASCII range",
UnicodeEscapeInByteLiteral => r"`\u{...}` escape in byte (string) literal not allowed",
InvalidStartOfUnicodeEscape => r"invalid start of `\u{...}` escape",
UnicodeEscapeWithoutBrace => r"`Unicode \u{...}` escape without opening brace",
NonHexDigitInUnicodeEscape => r"non-hex digit found in `\u{...}` escape",
TooManyDigitInUnicodeEscape => r"more than six digits in `\u{...}` escape",
InvalidUnicodeEscapeChar => r"value specified in `\u{...}` escape is not a valid char",
UnterminatedUnicodeEscape => r"unterminated `\u{...}` escape",
UnterminatedCharLiteral => "character literal is not terminated",
OverlongCharLiteral => "character literal contains more than one character",
EmptyCharLiteral => "empty character literal",
UnterminatedByteLiteral => "byte literal is not terminated",
OverlongByteLiteral => "byte literal contains more than one byte",
EmptyByteLiteral => "empty byte literal",
NonAsciiInByteLiteral => "non ASCII character in byte (string) literal",
UnescapedSingleQuote => "character literal contains unescaped ' character",
UnescapedSpecialWhitespace => r"unescaped newline (\n), tab (\t) or cr (\r) character",
DoesNotStartWithQuote => "invalid start for char/byte/string literal",
UnterminatedRawString => "unterminated raw (byte) string literal",
UnterminatedString => "unterminated (byte) string literal",
InvalidStringLiteralStart => "invalid start for string literal",
InvalidByteLiteralStart => "invalid start for byte literal",
InvalidByteStringLiteralStart => "invalid start for byte string literal",
IsolatedCr => r"`\r` not immediately followed by `\n` in string",
InvalidSuffix => "literal suffix is not a valid identifier",
UnexpectedIntegerLit => "expected float literal, but found integer",
IntegerSuffixStartingWithE => "integer literal suffix must not start with 'e' or 'E'",
};
description.fmt(f)?;
if let Some(span) = &self.span {
write!(f, " (at {}..{})", span.start, span.end)?;
}
Ok(())
}
}