lib.rs

/* This Source Code Form is subject to the terms of the Mozilla Public
 * License, v. 2.0. If a copy of the MPL was not distributed with this
 * file, You can obtain one at http://mozilla.org/MPL/2.0/. */

//! This crate implements a prefs file parser.
//!
//! Pref files have the following grammar. Note that there are slight
//! differences between the grammar for a default prefs files and a user prefs
//! file.
//!
//! ```text
//! <pref-file>   = <pref>*
//! <pref>        = <pref-spec> "(" <pref-name> "," <pref-value> <pref-attrs> ")" ";"
//! <pref-spec>   = "user_pref" | "pref" | "sticky_pref" // in default pref files
//! <pref-spec>   = "user_pref"                          // in user pref files
//! <pref-name>   = <string-literal>
//! <pref-value>  = <string-literal> | "true" | "false" | <int-value>
//! <int-value>   = <sign>? <int-literal>
//! <sign>        = "+" | "-"
//! <int-literal> = [0-9]+ (and cannot be followed by [A-Za-z_])
//! <string-literal> =
//!   A single or double-quoted string, with the following escape sequences
//!   allowed: \", \', \\, \n, \r, \xNN, \uNNNN, where \xNN gives a raw byte
//!   value that is copied directly into an 8-bit string value, and \uNNNN
//!   gives a UTF-16 code unit that is converted to UTF-8 before being copied
//!   into an 8-bit string value. \x00 and \u0000 are disallowed because they
//!   would cause C++ code handling such strings to misbehave.
//! <pref-attrs>  = ("," <pref-attr>)*      // in default pref files
//!               = <empty>                 // in user pref files
//! <pref-attr>   = "sticky" | "locked"     // default pref files only
//! ```
//!
//! Comments can take three forms:
//! - `# Python-style comments`
//! - `// C++ style comments`
//! - `/* C style comments (non-nested) */`
//!
//! Non-end-of-line whitespace chars are `\t`, `\v`, `\f`, and space.
//!
//! End-of-line sequences can take three forms, each of which is considered as
//! a single EOL:
//! - `\n`
//! - `\r` (without subsequent `\n`)
//! - `\r\n`
//!
//! The valid range for `<int-value>` is -2,147,483,648..2,147,483,647. Values
//! outside that range will result in a parse error.
//!
//! A `\0` char is interpreted as the end of the file. The use of this character
//! in a prefs file is not recommended. Within string literals `\x00` or
//! `\u0000` can be used instead.
//!
//! The parser performs error recovery. On a syntax error, it will scan forward
//! to the next `;` token and then continue parsing. If the syntax error occurs
//! in the middle of a token, it will first finish obtaining the current token
//! in an appropriate fashion.

// This parser uses several important optimizations.
//
// - Because "`\0` means EOF" is part of the grammar (see above), EOF is
//   representable by a u8. If EOF was represented by an out-of-band value such
//   as -1 or 256, we'd have to return a larger type such as `u16` or `i16`
//   from `get_char()`.
//
// - When starting a new token, it uses a lookup table with the first char,
//   which quickly identifies what kind of token it will be. Furthermore, if
//   that token is an unambiguous single-char token (e.g. `(`, `)`, `+`, `,`,
//   `-`, `;`), the parser will return the appropriate token kind value at
//   minimal cost because the single-char tokens have a uniform representation.
//
// - It has a lookup table that identifies chars in string literals that need
//   special handling. This means non-special chars (the common case) can be
//   handled with a single test, rather than testing for the multiple special
//   cases.
//
// - It pre-scans string literals for special chars. If none are present, it
//   bulk copies the string literal into a Vec, which is faster than doing a
//   char-by-char copy.
//
// - It reuses Vecs to avoid creating a new one for each string literal.

use std::os::raw::{c_char, c_uchar};

//---------------------------------------------------------------------------
// The public interface
//---------------------------------------------------------------------------

/// Keep this in sync with PrefType in Preferences.cpp.
#[derive(Clone, Copy, Debug, PartialEq)]
#[repr(u8)]
pub enum PrefType {
    None,
    String,
    Int,
    Bool,
}

/// Keep this in sync with PrefValueKind in Preferences.h.
#[derive(Clone, Copy, Debug, PartialEq)]
#[repr(u8)]
pub enum PrefValueKind {
    Default,
    User,
}

/// Keep this in sync with PrefValue in Preferences.cpp.
#[repr(C)]
pub union PrefValue {
    pub string_val: *const c_char,
    pub int_val: i32,
    pub bool_val: bool,
}

/// Keep this in sync with PrefsParserPrefFn in Preferences.cpp.
type PrefFn = unsafe extern "C" fn(
    pref_name: *const c_char,
    pref_type: PrefType,
    pref_value_kind: PrefValueKind,
    pref_value: PrefValue,
    is_sticky: bool,
    is_locked: bool,
);

/// Keep this in sync with PrefsParserErrorFn in Preferences.cpp.
type ErrorFn = unsafe extern "C" fn(msg: *const c_char);

/// Parse the contents of a prefs file.
///
/// `buf` is a null-terminated string. `len` is its length, excluding the
/// null terminator.
///
/// `pref_fn` is called once for each successfully parsed pref.
///
/// `error_fn` is called once for each parse error detected.
///
/// Keep this in sync with the prefs_parser_parse() declaration in
/// Preferences.cpp.
#[no_mangle]
pub unsafe extern "C" fn prefs_parser_parse(
    path: *const c_char,
    kind: PrefValueKind,
    buf: *const c_char,
    len: usize,
    pref_fn: PrefFn,
    error_fn: ErrorFn,
) -> bool {
    let path = std::ffi::CStr::from_ptr(path)
        .to_string_lossy()
        .into_owned();

    // Make sure `buf` ends in a '\0', and include that in the length, because
    // it represents EOF.
    let buf = std::slice::from_raw_parts(buf as *const c_uchar, len + 1);
    assert!(buf.last() == Some(&EOF));

    let mut parser = Parser::new(&path, kind, &buf, pref_fn, error_fn);
    parser.parse()
}

//---------------------------------------------------------------------------
// The implementation
//---------------------------------------------------------------------------

#[derive(Clone, Copy, Debug, PartialEq)]
enum Token {
    // Unambiguous single-char tokens.
    SingleChar(u8),

    // Keywords
    Pref,       // pref
    StickyPref, // sticky_pref
    UserPref,   // user_pref
    True,       // true
    False,      // false
    Sticky,     // sticky
    Locked,     // locked

    // String literal, e.g. '"string"'. The value is stored elsewhere.
    String,

    // Unsigned integer literal, e.g. '123'. Although libpref uses i32 values,
    // any '-' and '+' before an integer literal are treated as separate
    // tokens, so these token values are always positive. Furthermore, we
    // tokenize int literals as u32 so that 2147483648 (which doesn't fit into
    // an i32) can be subsequently negated to -2147483648 (which does fit into
    // an i32) if a '-' token precedes it.
    Int(u32),

    // Malformed token.
    Error(&'static str),

    // Malformed token at a particular line number. For use when
    // Parser::line_num might not be the right line number when the error is
    // reported. E.g. if a multi-line string has a bad escape sequence on the
    // first line, we don't report the error until the string's end has been
    // reached.
    ErrorAtLine(&'static str, u32),
}

// We categorize every char by what action should be taken when it appears at
// the start of a new token.
#[derive(Clone, Copy, PartialEq)]
enum CharKind {
    // These are ordered by frequency. See the comment in GetToken().
    SingleChar, // Unambiguous single-char tokens: [()+,-] or EOF
    SpaceNL,    // [\t\v\f \n]
    Keyword,    // [A-Za-z_]
    Quote,      // ["']
    Slash,      // /
    Digit,      // [0-9]
    Hash,       // #
    CR,         // \r
    Other,      // Everything else; invalid except within strings and comments.
}

const C_SINGL: CharKind = CharKind::SingleChar;
const C_SPCNL: CharKind = CharKind::SpaceNL;
const C_KEYWD: CharKind = CharKind::Keyword;
const C_QUOTE: CharKind = CharKind::Quote;
const C_SLASH: CharKind = CharKind::Slash;
const C_DIGIT: CharKind = CharKind::Digit;
const C_HASH_: CharKind = CharKind::Hash;
const C_CR___: CharKind = CharKind::CR;
const C______: CharKind = CharKind::Other;

#[rustfmt::skip]
const CHAR_KINDS: [CharKind; 256] = [
/*         0        1        2        3        4        5        6        7        8        9    */
/*   0+ */ C_SINGL, C______, C______, C______, C______, C______, C______, C______, C______, C_SPCNL,
/*  10+ */ C_SPCNL, C_SPCNL, C_SPCNL, C_CR___, C______, C______, C______, C______, C______, C______,
/*  20+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/*  30+ */ C______, C______, C_SPCNL, C______, C_QUOTE, C_HASH_, C______, C______, C______, C_QUOTE,
/*  40+ */ C_SINGL, C_SINGL, C______, C_SINGL, C_SINGL, C_SINGL, C______, C_SLASH, C_DIGIT, C_DIGIT,
/*  50+ */ C_DIGIT, C_DIGIT, C_DIGIT, C_DIGIT, C_DIGIT, C_DIGIT, C_DIGIT, C_DIGIT, C______, C_SINGL,
/*  60+ */ C______, C______, C______, C______, C______, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD,
/*  70+ */ C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD,
/*  80+ */ C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD,
/*  90+ */ C_KEYWD, C______, C______, C______, C______, C_KEYWD, C______, C_KEYWD, C_KEYWD, C_KEYWD,
/* 100+ */ C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD,
/* 110+ */ C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD, C_KEYWD,
/* 120+ */ C_KEYWD, C_KEYWD, C_KEYWD, C______, C______, C______, C______, C______, C______, C______,
/* 130+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 140+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 150+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 160+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 170+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 180+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 190+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 200+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 210+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 220+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 230+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 240+ */ C______, C______, C______, C______, C______, C______, C______, C______, C______, C______,
/* 250+ */ C______, C______, C______, C______, C______, C______
];

const _______: bool = false;
#[rustfmt::skip]
const SPECIAL_STRING_CHARS: [bool; 256] = [
/*         0        1        2        3        4        5        6        7        8        9    */
/*   0+ */    true, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  10+ */    true, _______, _______,    true, _______, _______, _______, _______, _______, _______,
/*  20+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  30+ */ _______, _______, _______, _______,    true, _______, _______, _______, _______,    true,
/*  40+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  50+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  60+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  70+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  80+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/*  90+ */ _______, _______,    true, _______, _______, _______, _______, _______, _______, _______,
/* 100+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 110+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 120+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 130+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 140+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 150+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 160+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 170+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 180+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 190+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 200+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 210+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 220+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 230+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 240+ */ _______, _______, _______, _______, _______, _______, _______, _______, _______, _______,
/* 250+ */ _______, _______, _______, _______, _______, _______
];

struct KeywordInfo {
    string: &'static [u8],
    token: Token,
}

const KEYWORD_INFOS: [KeywordInfo; 7] = [
    // These are ordered by frequency.
    KeywordInfo {
        string: b"pref",
        token: Token::Pref,
    },
    KeywordInfo {
        string: b"true",
        token: Token::True,
    },
    KeywordInfo {
        string: b"false",
        token: Token::False,
    },
    KeywordInfo {
        string: b"user_pref",
        token: Token::UserPref,
    },
    KeywordInfo {
        string: b"sticky",
        token: Token::Sticky,
    },
    KeywordInfo {
        string: b"locked",
        token: Token::Locked,
    },
    KeywordInfo {
        string: b"sticky_pref",
        token: Token::StickyPref,
    },
];

struct Parser<'t> {
    path: &'t str,       // Path to the file being parsed. Used in error messages.
    kind: PrefValueKind, // Default prefs file or user prefs file?
    buf: &'t [u8],       // Text being parsed.
    i: usize,            // Index of next char to be read.
    line_num: u32,       // Current line number within the text.
    pref_fn: PrefFn,     // Callback for processing each pref.
    error_fn: ErrorFn,   // Callback for parse errors.
    has_errors: bool,    // Have we encountered errors?
}

// As described above, we use 0 to represent EOF.
const EOF: u8 = b'\0';

impl<'t> Parser<'t> {
    fn new(
        path: &'t str,
        kind: PrefValueKind,
        buf: &'t [u8],
        pref_fn: PrefFn,
        error_fn: ErrorFn,
    ) -> Parser<'t> {
        // Make sure these tables take up 1 byte per entry.
        assert!(std::mem::size_of_val(&CHAR_KINDS) == 256);
        assert!(std::mem::size_of_val(&SPECIAL_STRING_CHARS) == 256);

        Parser {
            path: path,
            kind: kind,
            buf: buf,
            i: 0,
            line_num: 1,
            pref_fn: pref_fn,
            error_fn: error_fn,
            has_errors: false,
        }
    }

    fn parse(&mut self) -> bool {
        // These are reused, because allocating a new Vec for every string is slow.
        let mut name_str = Vec::with_capacity(128); // For pref names.
        let mut value_str = Vec::with_capacity(512); // For string pref values.
        let mut none_str = Vec::with_capacity(0); // For tokens that shouldn't be strings.

        let mut token = self.get_token(&mut none_str);

        // At the top of the loop we already have a token. In a valid input
        // this will be either the first token of a new pref, or EOF.
        loop {
            // <pref-spec>
            let (pref_value_kind, mut is_sticky) = match token {
                Token::Pref if self.kind == PrefValueKind::Default => {
                    (PrefValueKind::Default, false)
                }
                Token::StickyPref if self.kind == PrefValueKind::Default => {
                    (PrefValueKind::Default, true)
                }
                Token::UserPref => (PrefValueKind::User, false),
                Token::SingleChar(EOF) => return !self.has_errors,
                _ => {
                    token = self.error_and_recover(
                        token,
                        if self.kind == PrefValueKind::Default {
                            "expected pref specifier at start of pref definition"
                        } else {
                            "expected 'user_pref' at start of pref definition"
                        },
                    );
                    continue;
                }
            };

            // "("
            token = self.get_token(&mut none_str);
            if token != Token::SingleChar(b'(') {
                token = self.error_and_recover(token, "expected '(' after pref specifier");
                continue;
            }

            // <pref-name>
            token = self.get_token(&mut name_str);
            let pref_name = if token == Token::String {
                &name_str
            } else {
                token = self.error_and_recover(token, "expected pref name after '('");
                continue;
            };

            // ","
            token = self.get_token(&mut none_str);
            if token != Token::SingleChar(b',') {
                token = self.error_and_recover(token, "expected ',' after pref name");
                continue;
            }

            // <pref-value>
            token = self.get_token(&mut value_str);
            let (pref_type, pref_value) = match token {
                Token::True => (PrefType::Bool, PrefValue { bool_val: true }),
                Token::False => (PrefType::Bool, PrefValue { bool_val: false }),
                Token::String => (
                    PrefType::String,
                    PrefValue {
                        string_val: value_str.as_ptr() as *const c_char,
                    },
                ),
                Token::Int(u) => {
                    // Accept u <= 2147483647; anything larger will overflow i32.
                    if u <= std::i32::MAX as u32 {
                        (PrefType::Int, PrefValue { int_val: u as i32 })
                    } else {
                        token =
                            self.error_and_recover(Token::Error("integer literal overflowed"), "");
                        continue;
                    }
                }
                Token::SingleChar(b'-') => {
                    token = self.get_token(&mut none_str);
                    if let Token::Int(u) = token {
                        // Accept u <= 2147483648; anything larger will overflow i32 once negated.
                        if u <= std::i32::MAX as u32 {
                            (
                                PrefType::Int,
                                PrefValue {
                                    int_val: -(u as i32),
                                },
                            )
                        } else if u == std::i32::MAX as u32 + 1 {
                            (
                                PrefType::Int,
                                PrefValue {
                                    int_val: std::i32::MIN,
                                },
                            )
                        } else {
                            token = self
                                .error_and_recover(Token::Error("integer literal overflowed"), "");
                            continue;
                        }
                    } else {
                        token = self.error_and_recover(token, "expected integer literal after '-'");
                        continue;
                    }
                }
                Token::SingleChar(b'+') => {
                    token = self.get_token(&mut none_str);
                    if let Token::Int(u) = token {
                        // Accept u <= 2147483647; anything larger will overflow i32.
                        if u <= std::i32::MAX as u32 {
                            (PrefType::Int, PrefValue { int_val: u as i32 })
                        } else {
                            token = self
                                .error_and_recover(Token::Error("integer literal overflowed"), "");
                            continue;
                        }
                    } else {
                        token = self.error_and_recover(token, "expected integer literal after '+'");
                        continue;
                    }
                }
                _ => {
                    token = self.error_and_recover(token, "expected pref value after ','");
                    continue;
                }
            };

            // ("," <pref-attr>)*   // default pref files only
            let mut is_locked = false;
            let mut has_attrs = false;
            if self.kind == PrefValueKind::Default {
                let ok = loop {
                    // ","
                    token = self.get_token(&mut none_str);
                    if token != Token::SingleChar(b',') {
                        break true;
                    }

                    // <pref-attr>
                    token = self.get_token(&mut none_str);
                    match token {
                        Token::Sticky => is_sticky = true,
                        Token::Locked => is_locked = true,
                        _ => {
                            token =
                                self.error_and_recover(token, "expected pref attribute after ','");
                            break false;
                        }
                    }
                    has_attrs = true;
                };
                if !ok {
                    continue;
                }
            } else {
                token = self.get_token(&mut none_str);
            }

            // ")"
            if token != Token::SingleChar(b')') {
                let expected_msg = if self.kind == PrefValueKind::Default {
                    if has_attrs {
                        "expected ',' or ')' after pref attribute"
                    } else {
                        "expected ',' or ')' after pref value"
                    }
                } else {
                    "expected ')' after pref value"
                };
                token = self.error_and_recover(token, expected_msg);
                continue;
            }

            // ";"
            token = self.get_token(&mut none_str);
            if token != Token::SingleChar(b';') {
                token = self.error_and_recover(token, "expected ';' after ')'");
                continue;
            }

            unsafe {
                (self.pref_fn)(
                    pref_name.as_ptr() as *const c_char,
                    pref_type,
                    pref_value_kind,
                    pref_value,
                    is_sticky,
                    is_locked,
                )
            };

            token = self.get_token(&mut none_str);
        }
    }

    fn error_and_recover(&mut self, token: Token, msg: &str) -> Token {
        self.has_errors = true;

        // If `token` is a Token::{Error,ErrorAtLine}, it's a lexing error and
        // the error message is within `token`. Otherwise, it's a parsing error
        // and the error message is in `msg`.
        let (msg, line_num) = match token {
            Token::Error(token_msg) => (token_msg, self.line_num),
            Token::ErrorAtLine(token_msg, line_num) => (token_msg, line_num),
            _ => (msg, self.line_num),
        };
        let msg = format!("{}:{}: prefs parse error: {}", self.path, line_num, msg);
        let msg = std::ffi::CString::new(msg).unwrap();
        unsafe { (self.error_fn)(msg.as_ptr() as *const c_char) };

        // "Panic-mode" recovery: consume tokens until one of the following
        // occurs.
        // - We hit a semicolon, whereupon we return the following token.
        // - We hit EOF, whereupon we return EOF.
        //
        // For this to work, if the lexing functions hit EOF in an error case
        // they must unget it so we can safely reget it here.
        //
        // If the starting token (passed in above) is EOF we must not get
        // another token otherwise we will read past the end of `self.buf`.
        let mut dummy_str = Vec::with_capacity(128);
        let mut token = token;
        loop {
            match token {
                Token::SingleChar(b';') => return self.get_token(&mut dummy_str),
                Token::SingleChar(EOF) => return token,
                _ => {}
            }
            token = self.get_token(&mut dummy_str);
        }
    }

    #[inline(always)]
    fn get_char(&mut self) -> u8 {
        // We do the bounds check ourselves so we can return EOF on failure.
        // (Although the buffer is guaranteed to end in an EOF char, we might
        // go one char past that, whereupon we must return EOF again.)
        if self.i < self.buf.len() {
            let c = unsafe { *self.buf.get_unchecked(self.i) };
            self.i += 1;
            c
        } else {
            debug_assert!(self.i == self.buf.len());
            EOF
        }
    }

    // This function skips the bounds check in optimized builds. Using it at
    // the hottest two call sites gives a ~15% parsing speed boost.
    #[inline(always)]
    unsafe fn get_char_unchecked(&mut self) -> u8 {
        debug_assert!(self.i < self.buf.len());
        let c = *self.buf.get_unchecked(self.i);
        self.i += 1;
        c
    }

    #[inline(always)]
    fn unget_char(&mut self) {
        debug_assert!(self.i > 0);
        self.i -= 1;
    }

    #[inline(always)]
    fn match_char(&mut self, c: u8) -> bool {
        if self.buf[self.i] == c {
            self.i += 1;
            return true;
        }
        false
    }

    #[inline(always)]
    fn match_single_line_comment(&mut self) {
        loop {
            // To reach here, the previous char must have been '/' (if this is
            // the first loop iteration) or non-special (if this is the second
            // or subsequent iteration), and assertions elsewhere ensure that
            // there must be at least one subsequent char after those chars
            // (the '\0' for EOF).
            let c = unsafe { self.get_char_unchecked() };

            // All the special chars have value <= b'\r'.
            if c > b'\r' {
                continue;
            }
            match c {
                b'\n' => {
                    self.line_num += 1;
                    break;
                }
                b'\r' => {
                    self.line_num += 1;
                    self.match_char(b'\n');
                    break;
                }
                EOF => {
                    break;
                }
                _ => continue,
            }
        }
    }

    // Returns false if we hit EOF without closing the comment.
    fn match_multi_line_comment(&mut self) -> bool {
        loop {
            match self.get_char() {
                b'*' => {
                    if self.match_char(b'/') {
                        return true;
                    }
                }
                b'\n' => {
                    self.line_num += 1;
                }
                b'\r' => {
                    self.line_num += 1;
                    self.match_char(b'\n');
                }
                EOF => return false,
                _ => continue,
            }
        }
    }

    fn match_hex_digits(&mut self, ndigits: i32) -> Option<u16> {
        debug_assert!(ndigits == 2 || ndigits == 4);
        let mut value: u16 = 0;
        for _ in 0..ndigits {
            value = value << 4;
            match self.get_char() {
                c @ b'0'..=b'9' => value += (c - b'0') as u16,
                c @ b'A'..=b'F' => value += (c - b'A') as u16 + 10,
                c @ b'a'..=b'f' => value += (c - b'a') as u16 + 10,
                _ => {
                    self.unget_char();
                    return None;
                }
            }
        }
        Some(value)
    }

    #[inline(always)]
    fn char_kind(c: u8) -> CharKind {
        // Use get_unchecked() because a u8 index cannot exceed this table's
        // bounds.
        unsafe { *CHAR_KINDS.get_unchecked(c as usize) }
    }

    #[inline(always)]
    fn is_special_string_char(c: u8) -> bool {
        // Use get_unchecked() because a u8 index cannot exceed this table's
        // bounds.
        unsafe { *SPECIAL_STRING_CHARS.get_unchecked(c as usize) }
    }

    // If the obtained Token has a value, it is put within the Token, unless
    // it's a string, in which case it's put in `str_buf`. This avoids
    // allocating a new Vec for every string, which is slow.
    fn get_token(&mut self, str_buf: &mut Vec<u8>) -> Token {
        loop {
            // Note: the following tests are ordered by frequency when parsing
            // greprefs.js:
            // - SingleChar      36.7%
            // - SpaceNL         27.7% (14.9% for spaces, 12.8% for NL)
            // - Keyword         13.4%
            // - Quote           11.4%
            // - Slash            8.1%
            // - Digit            2.7%
            // - Hash, CR, Other  0.0%

            let c = self.get_char();
            match Parser::char_kind(c) {
                CharKind::SingleChar => {
                    return Token::SingleChar(c);
                }
                CharKind::SpaceNL => {
                    // It's slightly faster to combine the handling of the
                    // space chars with NL than to handle them separately; we
                    // have an extra test for this case, but one fewer test for
                    // all the subsequent CharKinds.
                    if c == b'\n' {
                        self.line_num += 1;
                    }
                    continue;
                }
                CharKind::Keyword => {
                    let start = self.i - 1;
                    loop {
                        let c = self.get_char();
                        if Parser::char_kind(c) != CharKind::Keyword {
                            self.unget_char();
                            break;
                        }
                    }
                    for info in KEYWORD_INFOS.iter() {
                        if &self.buf[start..self.i] == info.string {
                            return info.token;
                        }
                    }
                    return Token::Error("unknown keyword");
                }
                CharKind::Quote => {
                    return self.get_string_token(c, str_buf);
                }
                CharKind::Slash => {
                    match self.get_char() {
                        b'/' => {
                            self.match_single_line_comment();
                        }
                        b'*' => {
                            if !self.match_multi_line_comment() {
                                return Token::Error("unterminated /* comment");
                            }
                        }
                        c @ _ => {
                            if c == b'\n' || c == b'\r' {
                                // Unget the newline char; the outer loop will
                                // reget it and adjust self.line_num
                                // appropriately.
                                self.unget_char();
                            }
                            return Token::Error("expected '/' or '*' after '/'");
                        }
                    }
                    continue;
                }
                CharKind::Digit => {
                    let mut value = Some((c - b'0') as u32);
                    loop {
                        let c = self.get_char();
                        match Parser::char_kind(c) {
                            CharKind::Digit => {
                                fn add_digit(value: Option<u32>, c: u8) -> Option<u32> {
                                    value?.checked_mul(10)?.checked_add((c - b'0') as u32)
                                }
                                value = add_digit(value, c);
                            }
                            CharKind::Keyword => {
                                // Reject things like "123foo". Error recovery
                                // will retokenize from "foo" onward.
                                self.unget_char();
                                return Token::Error("unexpected character in integer literal");
                            }
                            _ => {
                                self.unget_char();
                                break;
                            }
                        }
                    }
                    return match value {
                        Some(v) => Token::Int(v),
                        None => Token::Error("integer literal overflowed"),
                    };
                }
                CharKind::Hash => {
                    self.match_single_line_comment();
                    continue;
                }
                CharKind::CR => {
                    self.match_char(b'\n');
                    self.line_num += 1;
                    continue;
                }
                // Error recovery will retokenize from the next character.
                _ => return Token::Error("unexpected character"),
            }
        }
    }

    fn string_error_token(&self, token: &mut Token, msg: &'static str) {
        // We only want to capture the first tokenization error within a string.
        if *token == Token::String {
            *token = Token::ErrorAtLine(msg, self.line_num);
        }
    }

    // Always inline this because it has a single call site.
    #[inline(always)]
    fn get_string_token(&mut self, quote_char: u8, str_buf: &mut Vec<u8>) -> Token {
        // First scan through the string to see if it contains any chars that
        // need special handling.
        let start = self.i;
        let has_special_chars = loop {
            // To reach here, the previous char must have been a quote
            // (quote_char), and assertions elsewhere ensure that there must be
            // at least one subsequent char (the '\0' for EOF).
            let c = unsafe { self.get_char_unchecked() };
            if Parser::is_special_string_char(c) {
                break c != quote_char;
            }
        };

        // Clear str_buf's contents without changing its capacity.
        str_buf.clear();

        // If there are no special chars (the common case), we can bulk copy it
        // to str_buf. This is a lot faster than the char-by-char loop below.
        if !has_special_chars {
            str_buf.extend(&self.buf[start..self.i - 1]);
            str_buf.push(b'\0');
            return Token::String;
        }

        // There were special chars. Re-scan the string, filling in str_buf one
        // char at a time.
        //
        // On error, we change `token` to an error token and then keep going to
        // the end of the string literal. `str_buf` won't be used in that case.
        self.i = start;
        let mut token = Token::String;

        loop {
            let c = self.get_char();
            let c2 = if !Parser::is_special_string_char(c) {
                c
            } else if c == quote_char {
                break;
            } else if c == b'\\' {
                match self.get_char() {
                    b'\"' => b'\"',
                    b'\'' => b'\'',
                    b'\\' => b'\\',
                    b'n' => b'\n',
                    b'r' => b'\r',
                    b'x' => {
                        if let Some(value) = self.match_hex_digits(2) {
                            debug_assert!(value <= 0xff);
                            if value != 0 {
                                value as u8
                            } else {
                                self.string_error_token(&mut token, "\\x00 is not allowed");
                                continue;
                            }
                        } else {
                            self.string_error_token(&mut token, "malformed \\x escape sequence");
                            continue;
                        }
                    }
                    b'u' => {
                        if let Some(value) = self.match_hex_digits(4) {
                            let mut utf16 = vec![value];
                            if 0xd800 == (0xfc00 & value) {
                                // High surrogate value. Look for the low surrogate value.
                                if self.match_char(b'\\') && self.match_char(b'u') {
                                    if let Some(lo) = self.match_hex_digits(4) {
                                        if 0xdc00 == (0xfc00 & lo) {
                                            // Found a valid low surrogate.
                                            utf16.push(lo);
                                        } else {
                                            self.string_error_token(
                                                &mut token,
                                                "invalid low surrogate after high surrogate",
                                            );
                                            continue;
                                        }
                                    }
                                }
                                if utf16.len() != 2 {
                                    self.string_error_token(
                                        &mut token,
                                        "expected low surrogate after high surrogate",
                                    );
                                    continue;
                                }
                            } else if 0xdc00 == (0xfc00 & value) {
                                // Unaccompanied low surrogate value.
                                self.string_error_token(
                                    &mut token,
                                    "expected high surrogate before low surrogate",
                                );
                                continue;
                            } else if value == 0 {
                                self.string_error_token(&mut token, "\\u0000 is not allowed");
                                continue;
                            }

                            // Insert the UTF-16 sequence as UTF-8.
                            let utf8 = String::from_utf16(&utf16).unwrap();
                            str_buf.extend(utf8.as_bytes());
                        } else {
                            self.string_error_token(&mut token, "malformed \\u escape sequence");
                            continue;
                        }
                        continue; // We don't want to str_buf.push(c2) below.
                    }
                    c @ _ => {
                        if c == b'\n' || c == b'\r' {
                            // Unget the newline char; the outer loop will
                            // reget it and adjust self.line_num appropriately.
                            self.unget_char();
                        }
                        self.string_error_token(
                            &mut token,
                            "unexpected escape sequence character after '\\'",
                        );
                        continue;
                    }
                }
            } else if c == b'\n' {
                self.line_num += 1;
                c
            } else if c == b'\r' {
                self.line_num += 1;
                if self.match_char(b'\n') {
                    str_buf.push(b'\r');
                    b'\n'
                } else {
                    c
                }
            } else if c == EOF {
                self.string_error_token(&mut token, "unterminated string literal");
                break;
            } else {
                // This case is only hit for the non-closing quote char.
                debug_assert!((c == b'\'' || c == b'\"') && c != quote_char);
                c
            };
            str_buf.push(c2);
        }
        str_buf.push(b'\0');

        token
    }
}