mod classes; mod comments; mod numbers; mod ptr; mod strings; use { SyntaxKind::{self, *}, TextUnit, }; use self::{ classes::*, comments::{scan_comment, scan_shebang}, numbers::scan_number, ptr::Ptr, strings::{ is_string_literal_start, scan_byte_char_or_string, scan_char, scan_raw_string, scan_string, }, }; /// A token of Rust source. #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord, Hash)] pub struct Token { /// The kind of token. pub kind: SyntaxKind, /// The length of the token. pub len: TextUnit, } /// Break a string up into its component tokens pub fn tokenize(text: &str) -> Vec { let mut text = text; let mut acc = Vec::new(); while !text.is_empty() { let token = next_token(text); acc.push(token); let len: u32 = token.len.into(); text = &text[len as usize..]; } acc } /// Get the next token from a string pub fn next_token(text: &str) -> Token { assert!(!text.is_empty()); let mut ptr = Ptr::new(text); let c = ptr.bump().unwrap(); let kind = next_token_inner(c, &mut ptr); let len = ptr.into_len(); Token { kind, len } } fn next_token_inner(c: char, ptr: &mut Ptr) -> SyntaxKind { if is_whitespace(c) { ptr.bump_while(is_whitespace); return WHITESPACE; } match c { '#' => if scan_shebang(ptr) { return SHEBANG; }, '/' => if let Some(kind) = scan_comment(ptr) { return kind; }, _ => (), } let ident_start = is_ident_start(c) && !is_string_literal_start(c, ptr.next(), ptr.nnext()); if ident_start { return scan_ident(c, ptr); } if is_dec_digit(c) { let kind = scan_number(c, ptr); scan_literal_suffix(ptr); return kind; } // One-byte tokens. match c { ';' => return SEMI, ',' => return COMMA, '(' => return L_PAREN, ')' => return R_PAREN, '{' => return L_CURLY, '}' => return R_CURLY, '[' => return L_BRACK, ']' => return R_BRACK, '<' => return L_ANGLE, '>' => return R_ANGLE, '@' => return AT, '#' => return POUND, '~' => return TILDE, '?' => return QUESTION, '$' => return DOLLAR, '&' => return AMPERSAND, '|' => return PIPE, '+' => return PLUS, '*' => return STAR, '/' => return SLASH, '^' => return CARET, '%' => return PERCENT, // Multi-byte tokens. '.' => { return match (ptr.next(), ptr.nnext()) { (Some('.'), Some('.')) => { ptr.bump(); ptr.bump(); DOTDOTDOT } (Some('.'), Some('=')) => { ptr.bump(); ptr.bump(); DOTDOTEQ } (Some('.'), _) => { ptr.bump(); DOTDOT } _ => DOT, }; } ':' => { return match ptr.next() { Some(':') => { ptr.bump(); COLONCOLON } _ => COLON, }; } '=' => { return match ptr.next() { Some('=') => { ptr.bump(); EQEQ } Some('>') => { ptr.bump(); FAT_ARROW } _ => EQ, }; } '!' => { return match ptr.next() { Some('=') => { ptr.bump(); NEQ } _ => EXCL, }; } '-' => { return if ptr.next_is('>') { ptr.bump(); THIN_ARROW } else { MINUS }; } // If the character is an ident start not followed by another single // quote, then this is a lifetime name: '\'' => { return if ptr.next_is_p(is_ident_start) && !ptr.nnext_is('\'') { ptr.bump(); while ptr.next_is_p(is_ident_continue) { ptr.bump(); } // lifetimes shouldn't end with a single quote // if we find one, then this is an invalid character literal if ptr.next_is('\'') { ptr.bump(); return CHAR; // TODO: error reporting } LIFETIME } else { scan_char(ptr); scan_literal_suffix(ptr); CHAR }; } 'b' => { let kind = scan_byte_char_or_string(ptr); scan_literal_suffix(ptr); return kind; } '"' => { scan_string(ptr); scan_literal_suffix(ptr); return STRING; } 'r' => { scan_raw_string(ptr); scan_literal_suffix(ptr); return RAW_STRING; } _ => (), } ERROR } fn scan_ident(c: char, ptr: &mut Ptr) -> SyntaxKind { let is_single_letter = match ptr.next() { None => true, Some(c) if !is_ident_continue(c) => true, _ => false, }; if is_single_letter { return if c == '_' { UNDERSCORE } else { IDENT }; } ptr.bump_while(is_ident_continue); if let Some(kind) = SyntaxKind::from_keyword(ptr.current_token_text()) { return kind; } IDENT } fn scan_literal_suffix(ptr: &mut Ptr) { if ptr.next_is_p(is_ident_start) { ptr.bump(); } ptr.bump_while(is_ident_continue); }