/mypy/lex.py

"""Lexical analyzer for mypy.

Translate a string that represents a file or a compilation unit to a list of
tokens.

This module can be run as a script (lex.py FILE).
"""

import re

from mypy.util import short_type


class Token:
    """Base class for all tokens"""
    pre = '' # Space, comments etc. before token
    string = None   # Token string
    line = None     # Token line number
    
    def __init__(self, string, pre=''):
        self.string = string
        self.pre = pre
    
    def __repr__(self):
        """The representation is of form Keyword('  if')."""
        t = short_type(self)
        return t + '(' + self.fix(self.pre) + self.fix(self.string) + ')'
    
    def rep(self):
        return self.pre + self.string
    
    def fix(self, s):
        """Replace common non-printable chars with escape sequences.

        Do not use repr() since we don't want do duplicate backslashes.
        """
        return s.replace('\n', '\\n').replace('\t', '\\t').replace('\r', '\\r')


# Token classes


class Break(Token):
    """Statement break (line break or semicolon)"""


class Indent(Token):
    """Increase block indent level."""


class Dedent(Token):
    """Decrease block indent level."""


class Eof(Token):
    """End of file"""


class Keyword(Token):
    """Reserved word (other than keyword operators; they use Op).

    Examples: if, class, while, def.
    """


class Name(Token):
    """An alphanumeric identifier"""


class IntLit(Token):
    """Integer literal"""


class StrLit(Token):
    """String literal"""
    def parsed(self):
        """Return the parsed contents of the literal."""
        return _parse_str_literal(self.string)


class BytesLit(Token):
    """Bytes literal"""
    def parsed(self):
        """Return the parsed contents of the literal."""
        return _parse_str_literal(self.string)


class FloatLit(Token):
    """Float literal"""


class Punct(Token):
    """Punctuator (e.g. comma, '(' or '=')"""


class Colon(Token):
    pass


class Op(Token):
    """Operator (e.g. '+' or 'in')"""


class Bom(Token):
    """Byte order mark (at the start of a file)"""


class LexError(Token):
    """Lexer error token"""
    type = None # One of the error types below
    
    def __init__(self, string, type):
        super().__init__(string)
        self.type = type


# Lexer error types
NUMERIC_LITERAL_ERROR = 0
UNTERMINATED_STRING_LITERAL = 1
INVALID_CHARACTER = 2
NON_ASCII_CHARACTER_IN_COMMENT = 3
NON_ASCII_CHARACTER_IN_STRING = 4
INVALID_UTF8_SEQUENCE = 5
INVALID_BACKSLASH = 6
INVALID_DEDENT = 7

# Encoding contexts
STR_CONTEXT = 1
COMMENT_CONTEXT = 2


def lex(string):
    """Analyze string and return an array of token objects.

    The last token is always Eof.
    """
    l = Lexer()
    l.lex(string)
    return l.tok


# Reserved words (not including operators)
keywords = set([
    'any', 'as', 'assert', 'break', 'class', 'continue', 'def', 'del', 'elif',
    'else', 'except', 'finally', 'from', 'for', 'global', 'if', 'import',
    'interface', 'lambda', 'pass', 'raise', 'return', 'try', 'while', 'with',
    'yield'])

# Alphabetical operators (reserved words)
alpha_operators = set(['in', 'is', 'not', 'and', 'or'])

# String literal prefixes
str_prefixes = set(['r', 'b', 'br'])  

# List of regular expressions that match non-alphabetical operators
operators = [re.compile('[-+*/<>.%&|^~]'),
                           re.compile('==|!=|<=|>=|\\*\\*|//|<<|>>')]

# List of regular expressions that match punctuator tokens
punctuators = [re.compile('[=,()@]'),
                             re.compile('\\['),
                             re.compile(']'),
                             re.compile('([-+*/%&|^]|\\*\\*|//|<<|>>)=')]


# Source file encodings
DEFAULT_ENCODING = 0
ASCII_ENCODING = 1
LATIN1_ENCODING = 2
UTF8_ENCODING = 3


# Map single-character string escape sequences to corresponding characters.
escape_map = {'a': '\x07',
              'b': '\x08',
              'f': '\x0c',
              'n': '\x0a',
              'r': '\x0d',
              't': '\x09',
              'v': '\x0b',
              '"': '"',
              "'": "'"}


# Matches the optional prefix of a string literal, e.g. the 'r' in r"foo".
str_prefix_re = re.compile('[rRbB]*')

# Matches an escape sequence in a string, e.g. \n or \x4F.
escape_re = re.compile(
    "\\\\([abfnrtv'\"]|x[0-9a-fA-F]{2}|u[0-9a-fA-F]{4}|[0-7]{1,3})")


def _parse_str_literal(string):
    """Translate escape sequences in str literal to the corresponding chars.

    For example, \t is translated to the tab character (ascii 9).

    Return the translated contents of the literal.  Also handle raw and
    triple-quoted string literals.
    """
    prefix = str_prefix_re.match(string).group(0).lower()
    s = string[len(prefix):]
    if s.startswith("'''") or s.startswith('"""'):
        return s[3:-3]
    elif 'r' in prefix:
        return s[1:-1].replace('\\' + s[0], s[0])
    else:
        return escape_re.sub(lambda m: escape_repl(m, prefix), s[1:-1])


def escape_repl(m, prefix):
    """Translate a string escape sequence, e.g. \t -> the tab character.

    Assume that the Match object is from escape_re.
    """
    seq = m.group(1)
    if len(seq) == 1 and seq in escape_map:
        # Single-character escape sequence, e.g. \n.
        return escape_map[seq]
    elif seq.startswith('x'):
        # Hexadecimal sequence \xNN.
        return chr(int(seq[1:], 16))
    elif seq.startswith('u'):
        # Unicode sequence \uNNNN.
        if 'b' not in prefix:
            return chr(int(seq[1:], 16))
        else:
            return '\\' + seq
    else:
        # Octal sequence.
        ord = int(seq, 8)
        if 'b' in prefix:
            # Make sure code is no larger than 255 for bytes literals.
            ord = ord % 256
        return chr(ord)


class Lexer:
    """Lexical analyzer."""
    i = None     # Current string index (into s)
    s = None     # The string being analyzed
    line = None  # Current line number
    pre_whitespace = ''     # Whitespace and comments before the next token
    enc = DEFAULT_ENCODING  # Encoding TODO implement properly

    # Generated tokens
    tok = None
    
    # Table from byte character value to lexer method. E.g. entry at ord('0')
    # contains the method lex_number().
    map = None

    # Indent levels of currently open blocks, in spaces.
    indents = None
    
    # Open ('s, ['s and {'s without matching closing bracket; used for ignoring
    # newlines within parentheses/brackets.
    open_brackets = None
    
    def __init__(self):
        self.map = [self.unknown_character] * 256
        self.tok = []
        self.indents = [0]
        self.open_brackets = []
        # Fill in the map from valid character codes to relevant lexer methods.
        for seq, method in [('ABCDEFGHIJKLMNOPQRSTUVWXYZ', self.lex_name),
                            ('abcdefghijklmnopqrstuvwxyz_', self.lex_name),
                            ('0123456789', self.lex_number),
                            ('.', self.lex_number_or_dot),
                            (' ' + '\t' + '\x0c', self.lex_space),
                            ('"', self.lex_str_double),
                            ("'", self.lex_str_single),
                            ('\r' + '\n', self.lex_break),
                            (';', self.lex_semicolon),
                            (':', self.lex_colon),
                            ('#', self.lex_comment),
                            ('\\', self.lex_backslash),
                            ('([{', self.lex_open_bracket),
                            (')]}', self.lex_close_bracket),
                            ('-+*/<>%&|^~=!,@', self.lex_misc)]:
            for c in seq:
                self.map[ord(c)] = method
    
    def lex(self, s):
        """Lexically analyze a string, storing the tokens at the tok array."""
        self.s = s
        self.i = 0
        self.line = 1    

        if s.startswith('\xef\xbb\xbf'):
            self.add_token(Bom(s[0:3]))
            
        # Parse initial indent; otherwise first-line indent would not generate
        # an error.
        self.lex_indent()

        # Make a local copy of map as a simple optimization.
        map = self.map
        
        # Lex the file. Repeatedly call the lexer method for the current char.
        while self.i < len(s):
            # Get the character code of the next character to lex.
            c = ord(s[self.i])
            # Dispatch to the relevant lexer method. This will consume some
            # characters in the text, add a token to self.tok and increment
            # self.i.
            map[c]()
        
        # Append a break if there is no statement/block terminator at the end
        # of input.
        if len(self.tok) > 0 and (not isinstance(self.tok[-1], Break) and
                                  not isinstance(self.tok[-1], Dedent)):
            self.add_token(Break(''))

        # Close remaining open blocks with Dedent tokens.
        self.lex_indent()
        
        self.add_token(Eof(''))
    
    def lex_number_or_dot(self):
        """Analyse a token starting with a dot.

        It can be the member access operator or a float literal such as '.123'.
        """
        if self.is_at_number():
            self.lex_number()
        else:
            self.lex_misc()
    
    number_exp = re.compile(r'[0-9]|\.[0-9]')
    
    def is_at_number(self):
        """Is the current location at a numeric literal?"""
        return self.match(self.number_exp) != ''
    
    # Regexps used by lex_number

    # Decimal/hex/octal literal
    number_exp1 = re.compile('0[xXoO][0-9a-fA-F]+|[0-9]+')
    # Float literal, e.g. '1.23' or '12e+34'
    number_exp2 = re.compile(
        r'[0-9]*\.[0-9]*([eE][-+]?[0-9]+)?|[0-9]+[eE][-+]?[0-9]+')
    # These characters must not appear after a number literal.
    name_char_exp = re.compile('[a-zA-Z0-9_]')
    
    def lex_number(self):
        """Analyse an int or float literal.

        Assume that the current location points to one of them.
        """
        s1 = self.match(self.number_exp1)
        s2 = self.match(self.number_exp2)
        
        maxlen = max(len(s1), len(s2))
        if self.name_char_exp.match(
                    self.s[self.i + maxlen:self.i + maxlen + 1]) is not None:
            # Error: alphanumeric character after number literal.
            s3 = self.match(re.compile('[0-9][0-9a-zA-Z_]*'))
            maxlen = max(maxlen, len(s3))
            self.add_token(LexError(' ' * maxlen, NUMERIC_LITERAL_ERROR))
        elif len(s1) > len(s2):
            # Integer literal.
            self.add_token(IntLit(s1))
        else:
            # Float literal.
            self.add_token(FloatLit(s2))
    
    name_exp = re.compile('[a-zA-Z_][a-zA-Z0-9_]*')
    
    def lex_name(self):
        """Analyse a name (an identifier, a keyword or an alphabetical
        operator).  This also deals with prefixed string literals such
        as r'...'.
        """
        s = self.match(self.name_exp)
        if s in keywords:
            self.add_token(Keyword(s))
        elif s in alpha_operators:
            self.add_token(Op(s))
        elif s in str_prefixes and self.match(re.compile('[a-z]+[\'"]')) != '':
            self.lex_prefixed_str(s)
        else:
            self.add_token(Name(s))
    
    # Regexps representing components of string literals

    # Initial part of a single-quoted literal, e.g. b'foo' or b'foo\\\n
    str_exp_single = re.compile(
        r"[a-z]*'([^'\\\r\n]|\\[^\r\n])*('|\\(\n|\r\n?))")
    # Non-initial part of a multiline single-quoted literal, e.g. foo'
    str_exp_single_multi = re.compile(
        r"([^'\\\r\n]|\\[^\r\n])*('|\\(\n|\r\n?))")
    # Initial part of a single-quoted raw literal, e.g. r'foo' or r'foo\\\n
    str_exp_raw_single = re.compile(
        r"[a-z]*'([^'\r\n\\]|\\'|\\[^\n\r])*('|\\(\n|\r\n?))")
    # Non-initial part of a raw multiline single-quoted literal, e.g. foo'
    str_exp_raw_single_multi = re.compile(
        r"([^'\r\n]|'')*('|\\(\n|\r\n?))")

    # Start of a ''' literal, e.g. b'''
    str_exp_single3 = re.compile("[a-z]*'''")
    # End of a ''' literal, e.g. foo'''
    str_exp_single3end = re.compile(r"[^\n\r]*?'''")

    # The following are similar to above (but use double quotes).
    
    str_exp_double = re.compile(
        r'[a-z]*"([^"\\\r\n]|\\[^\r\n])*("|\\(\n|\r\n?))')
    str_exp_double_multi = re.compile(
        r'([^"\\\r\n]|\\[^\r\n])*("|\\(\n|\r\n?))')  
    str_exp_raw_double = re.compile(
        r'[a-z]*"([^"\r\n\\]|\\"|\\[^\n\r])*("|\\(\n|\r\n?))')
    str_exp_raw_double_multi = re.compile(
        r'([^"\r\n]|"")*("|\\(\n|\r\n?))')
    
    str_exp_double3 = re.compile('[a-z]*"""')
    str_exp_double3end = re.compile(r'[^\n\r]*?"""')
    
    def lex_str_single(self):
        """Analyse single-quoted string literal"""
        self.lex_str(self.str_exp_single, self.str_exp_single_multi,
                     self.str_exp_single3, self.str_exp_single3end)
    
    def lex_str_double(self):
        """Analyse double-quoted string literal"""
        self.lex_str(self.str_exp_double, self.str_exp_double_multi,
                     self.str_exp_double3, self.str_exp_double3end)
    
    def lex_prefixed_str(self, prefix):
        """Analyse a string literal with a prefix, such as r'...'."""
        s = self.match(re.compile('[a-z]+[\'"]'))
        if s.endswith("'"):
            re1 = self.str_exp_single
            re2 = self.str_exp_single_multi
            if 'r' in prefix:
                re1 = self.str_exp_raw_single
                re2 = self.str_exp_raw_single_multi
            self.lex_str(re1, re2, self.str_exp_single3,
                         self.str_exp_single3end, prefix)
        else:
            re1 = self.str_exp_double
            re2 = self.str_exp_double_multi
            if 'r' in prefix:
                re1 = self.str_exp_raw_double
                re2 = self.str_exp_raw_double_multi
            self.lex_str(re1, re2, self.str_exp_double3,
                         self.str_exp_double3end, prefix)
    
    def lex_str(self, regex, re2, re3, re3end, prefix=''):
        """Analyse a string literal described by regexps. Assume that
        the current location is at the beginning of the literal. The
        arguments re3 and re3end describe the corresponding
        triple-quoted literals.
        """
        s3 = self.match(re3)
        if s3 != '':
            # Triple-quoted string literal.
            self.lex_triple_quoted_str(re3end, prefix)
        else:
            # Single or double quoted string literal.
            s = self.match(regex)
            if s != '':
                if s.endswith('\n') or s.endswith('\r'):
                    self.lex_multiline_string_literal(re2, s)
                else:
                    self.verify_encoding(s, STR_CONTEXT)
                    if 'b' in prefix:
                        self.add_token(BytesLit(s))
                    else:
                        self.add_token(StrLit(s))
            else:
                # Unterminated string literal.
                s = self.match(re.compile('[^\\n\\r]*'))
                self.add_token(LexError(s, UNTERMINATED_STRING_LITERAL))
    
    def lex_triple_quoted_str(self, re3end, prefix):
        line = self.line
        ss = self.s[self.i:self.i + len(prefix) + 3]
        self.i += len(prefix) + 3
        while True:
            m = re3end.match(self.s, self.i)
            if m is not None:
                break
            m = re.match('[^\\n\\r]*(\\n|\\r\\n?)', self.s[self.i:])
            if m is None:
                self.add_special_token(
                    LexError(ss, UNTERMINATED_STRING_LITERAL), line, 0)
                return 
            s = m.group(0)
            ss += s
            self.line += 1
            self.i += len(s)
        lit = None
        if 'b' in prefix:
            lit = BytesLit(ss + m.group(0))
        else:
            lit = StrLit(ss + m.group(0))
        self.add_special_token(lit, line, len(m.group(0)))
    
    def lex_multiline_string_literal(self, re_end, prefix):
        """Analyze multiline single/double-quoted string literal.

        Use explicit \ for line continuation.
        """
        line = self.line
        self.i += len(prefix)
        ss = prefix
        while True:
            m = self.match(re_end)
            if m == '':
                self.add_special_token(
                    LexError(ss, UNTERMINATED_STRING_LITERAL), line, 0)
                return 
            ss += m
            self.line += 1
            self.i += len(m)        
            if not m.endswith('\n') and not m.endswith('\r'): break
        self.add_special_token(StrLit(ss), line, 0) # TODO bytes
    
    comment_exp = re.compile(r'#[^\n\r]*')
    
    def lex_comment(self):
        """Analyse a comment."""
        s = self.match(self.comment_exp)
        self.verify_encoding(s, COMMENT_CONTEXT)
        self.add_pre_whitespace(s)
    
    backslash_exp = re.compile(r'\\(\n|\r\n?)')
    
    def lex_backslash(self):
        s = self.match(self.backslash_exp)
        if s != '':
            self.add_pre_whitespace(s)
            self.line += 1
        else:
            self.add_token(LexError('\\', INVALID_BACKSLASH))
    
    space_exp = re.compile(r'[ \t\x0c]*')
    indent_exp = re.compile(r'[ \t]*[#\n\r]?')
    
    def lex_space(self):
        """Analyze a run of whitespace characters (within a line, not indents).

        Only store them in self.pre_whitespace.
        """
        s = self.match(self.space_exp)
        self.add_pre_whitespace(s)
    
    comment_or_newline = '#' + '\n' + '\r'
    
    def lex_indent(self):
        """Analyze whitespace chars at the beginning of a line (indents)."""
        s = self.match(self.indent_exp)
        if s != '' and s[-1] in self.comment_or_newline:
            # Empty line (whitespace only or comment only).
            self.add_pre_whitespace(s[:-1])
            if s[-1] == '#':
                self.lex_comment()
            else:
                self.lex_break()
            self.lex_indent()
            return 
        indent = self.calc_indent(s)
        if indent == self.indents[-1]:
            # No change in indent: just whitespace.
            self.add_pre_whitespace(s)
        elif indent > self.indents[-1]:
            # An increased indent (new block).
            self.indents.append(indent)
            self.add_token(Indent(s))
        else:
            # Decreased indent (end of one or more blocks).
            pre = self.pre_whitespace
            self.pre_whitespace = ''
            while indent < self.indents[-1]:
                self.add_token(Dedent(''))
                self.indents.pop()
            self.pre_whitespace = pre
            self.add_pre_whitespace(s)            
            if indent != self.indents[-1]:
                # Error: indent level does not match a previous indent level.
                self.add_token(LexError('', INVALID_DEDENT))
    
    def calc_indent(self, s):
        indent = 0
        for ch in s:
            if ch == ' ':
                indent += 1
            else:
                # Tab: 8 spaces (rounded to a multiple of 8).
                indent += 8 - indent % 8
        return indent
    
    break_exp = re.compile(r'\r\n|\r|\n|;')
    
    def lex_break(self):
        """Analyse a line break."""
        s = self.match(self.break_exp)
        if self.ignore_break():
            self.add_pre_whitespace(s)
            self.line += 1
        else:
            self.add_token(Break(s))
            self.line += 1
            self.lex_indent()
    
    def lex_semicolon(self):
        self.add_token(Break(';'))
    
    def lex_colon(self):
        self.add_token(Colon(':'))
    
    open_bracket_exp = re.compile('[[({]')
    
    def lex_open_bracket(self):
        s = self.match(self.open_bracket_exp)
        self.open_brackets.append(s)
        self.add_token(Punct(s))
    
    close_bracket_exp = re.compile('[])}]')
    
    open_bracket = {')': '(', ']': '[', '}': '{'}
    
    def lex_close_bracket(self):
        s = self.match(self.close_bracket_exp)
        if (self.open_brackets != []
                and self.open_bracket[s] == self.open_brackets[-1]):
            self.open_brackets.pop()
        self.add_token(Punct(s))
    
    def lex_misc(self):
        """Analyse a non-alphabetical operator or a punctuator."""
        s = ''
        t = None
        for re_list, type in [(operators, Op), (punctuators, Punct)]:
            for re in re_list:
                s2 = self.match(re)
                if len(s2) > len(s):
                    t = type
                    s = s2
        if s == '':
            # Could not match any token; report an invalid character. This is
            # reached at least if the current character is '!' not followed by
            # '='.
            self.add_token(LexError(self.s[self.i], INVALID_CHARACTER))
        else:
            self.add_token(t(s))
    
    def unknown_character(self):
        """Report an unknown character as a lexical analysis error."""
        self.add_token(LexError(self.s[self.i], INVALID_CHARACTER))
    
    
    # Utility methods
    
    
    def match(self, pattern):
        """If the argument regexp is matched at the current location,
        return the matched string; otherwise return the empty string.
        """
        m = pattern.match(self.s, self.i)
        if m is not None:
            return m.group(0)
        else:
            return ''
    
    def add_pre_whitespace(self, s):
        """Record whitespace and comments before the next token.
        
        The accumulated whitespace/comments will be stored in the next token
        and then it will be cleared.

        This is needed for pretty-printing the original source code while
        preserving comments, indentation, whitespace etc.
        """
        self.pre_whitespace += s
        self.i += len(s)
    
    def add_token(self, tok):
        """Store a token. Update its line number and record preceding
        whitespace characters and comments.
        """
        if (tok.string == '' and not isinstance(tok, Eof)
                and not isinstance(tok, Break)
                and not isinstance(tok, LexError)
                and not isinstance(tok, Dedent)):
            raise ValueError('Empty token')
        tok.pre = self.pre_whitespace
        tok.line = self.line
        self.tok.append(tok)
        self.i += len(tok.string)
        self.pre_whitespace = ''
    
    def add_special_token(self, tok, line, skip):
        """Like add_token, but caller sets the number of chars to skip."""
        if (tok.string == '' and not isinstance(tok, Eof)
                and not isinstance(tok, Break)
                and not isinstance(tok, LexError)
                and not isinstance(tok, Dedent)):
            raise ValueError('Empty token')
        tok.pre = self.pre_whitespace
        tok.line = line
        self.tok.append(tok)
        self.i += skip
        self.pre_whitespace = ''
    
    def ignore_break(self):
        """If the next token is a break, can we ignore it?"""
        if len(self.open_brackets) > 0 or len(self.tok) == 0:
            # Ignore break after open ( [ or { or at the beginning of file.
            return True
        else:
            # Ignore break after another break or dedent.
            t = self.tok[-1]
            return isinstance(t, Break) or isinstance(t, Dedent)
    
    def verify_encoding(self, string, context):
        """Verify that a token (represented by a string) is encoded correctly
        according to the file encoding.
        """
        codec = None
        if self.enc == ASCII_ENCODING:
            codec = 'ascii'
        elif self.enc in [UTF8_ENCODING, DEFAULT_ENCODING]:
            codec = 'utf8'
        if codec is not None:
            try:
                pass # FIX string.decode(codec)
            except UnicodeDecodeError:
                type = INVALID_UTF8_SEQUENCE
                if self.enc == ASCII_ENCODING:
                    if context == STR_CONTEXT:
                        type = NON_ASCII_CHARACTER_IN_STRING
                    else:
                        type = NON_ASCII_CHARACTER_IN_COMMENT
                self.add_token(LexError('', type))


import sys

if __name__ == '__main__':
    if len(sys.argv) != 2:
        print('Usage: lex.py FILE')
        sys.exit(2)
    fnam = sys.argv[1]
    s = open(fnam).read()
    for t in lex(s):
        print(t)