""" This module contains a tokenizer for Excel formulae. The tokenizer is based on the Javascript tokenizer found at http://ewbi.blogs.com/develops/2004/12/excel_formula_p.html written by Eric Bachtal """ import re class TokenizerError(Exception): """Base class for all Tokenizer errors.""" class Tokenizer(object): """ A tokenizer for Excel worksheet formulae. Converts a unicode string representing an Excel formula (in A1 notation) into a sequence of `Token` objects. `formula`: The unicode string to tokenize Tokenizer defines a method `._parse()` to parse the formula into tokens, which can then be accessed through the `.items` attribute. """ SN_RE = re.compile("^[1-9](\\.[0-9]+)?[Ee]$") # Scientific notation WSPACE_RE = re.compile(r"[ \n]+") STRING_REGEXES = { # Inside a string, all characters are treated as literals, except for # the quote character used to start the string. That character, when # doubled is treated as a single character in the string. If an # unmatched quote appears, the string is terminated. '"': re.compile('"(?:[^"]*"")*[^"]*"(?!")'), "'": re.compile("'(?:[^']*'')*[^']*'(?!')"), } ERROR_CODES = ("#NULL!", "#DIV/0!", "#VALUE!", "#REF!", "#NAME?", "#NUM!", "#N/A", "#GETTING_DATA") TOKEN_ENDERS = ',;}) +-*/^&=><%' # Each of these characters, marks the # end of an operand token def __init__(self, formula): self.formula = formula self.items = [] self.token_stack = [] # Used to keep track of arrays, functions, and # parentheses self.offset = 0 # How many chars have we read self.token = [] # Used to build up token values char by char self._parse() def _parse(self): """Populate self.items with the tokens from the formula.""" if self.offset: return # Already parsed! if not self.formula: return elif self.formula[0] == '=': self.offset += 1 else: self.items.append(Token(self.formula, Token.LITERAL)) return consumers = ( ('"\'', self._parse_string), ('[', self._parse_brackets), ('#', self._parse_error), (' ', self._parse_whitespace), ('\n', self._parse_whitespace), ('+-*/^&=><%', self._parse_operator), ('{(', self._parse_opener), (')}', self._parse_closer), (';,', self._parse_separator), ) dispatcher = {} # maps chars to the specific parsing function for chars, consumer in consumers: dispatcher.update(dict.fromkeys(chars, consumer)) while self.offset < len(self.formula): if self.check_scientific_notation(): # May consume one character continue curr_char = self.formula[self.offset] if curr_char in self.TOKEN_ENDERS: self.save_token() if curr_char in dispatcher: self.offset += dispatcher[curr_char]() else: # TODO: this can probably be sped up using a regex to get to # the next interesting character self.token.append(curr_char) self.offset += 1 self.save_token() def _parse_string(self): """ Parse a "-delimited string or '-delimited link. The offset must be pointing to either a single quote ("'") or double quote ('"') character. The strings are parsed according to Excel rules where to escape the delimiter you just double it up. E.g., "abc""def" in Excel is parsed as 'abc"def' in Python. Returns the number of characters matched. (Does not update self.offset) """ self.assert_empty_token(can_follow=':') delim = self.formula[self.offset] assert delim in ('"', "'") regex = self.STRING_REGEXES[delim] match = regex.match(self.formula[self.offset:]) if match is None: subtype = "string" if delim == '"' else 'link' raise TokenizerError( "Reached end of formula while parsing %s in %s" % (subtype, self.formula)) match = match.group(0) if delim == '"': self.items.append(Token.make_operand(match)) else: self.token.append(match) return len(match) def _parse_brackets(self): """ Consume all the text between square brackets []. Returns the number of characters matched. (Does not update self.offset) """ assert self.formula[self.offset] == '[' lefts = [(t.start(), 1) for t in re.finditer(r"\[", self.formula[self.offset:])] rights = [(t.start(), -1) for t in re.finditer(r"\]", self.formula[self.offset:])] open_count = 0 for idx, open_close in sorted(lefts + rights): open_count += open_close if open_count == 0: outer_right = idx + 1 self.token.append( self.formula[self.offset:self.offset + outer_right]) return outer_right raise TokenizerError("Encountered unmatched '[' in %s" % self.formula) def _parse_error(self): """ Consume the text following a '#' as an error. Looks for a match in self.ERROR_CODES and returns the number of characters matched. (Does not update self.offset) """ self.assert_empty_token(can_follow='!') assert self.formula[self.offset] == '#' subformula = self.formula[self.offset:] for err in self.ERROR_CODES: if subformula.startswith(err): self.items.append(Token.make_operand(''.join(self.token) + err)) del self.token[:] return len(err) raise TokenizerError( "Invalid error code at position %d in '%s'" % (self.offset, self.formula)) def _parse_whitespace(self): """ Consume a string of consecutive spaces. Returns the number of spaces found. (Does not update self.offset). """ assert self.formula[self.offset] in (' ', '\n') self.items.append(Token(self.formula[self.offset], Token.WSPACE)) return self.WSPACE_RE.match(self.formula[self.offset:]).end() def _parse_operator(self): """ Consume the characters constituting an operator. Returns the number of characters consumed. (Does not update self.offset) """ if self.formula[self.offset:self.offset + 2] in ('>=', '<=', '<>'): self.items.append(Token( self.formula[self.offset:self.offset + 2], Token.OP_IN )) return 2 curr_char = self.formula[self.offset] # guaranteed to be 1 char assert curr_char in '%*/^&=><+-' if curr_char == '%': token = Token('%', Token.OP_POST) elif curr_char in "*/^&=><": token = Token(curr_char, Token.OP_IN) # From here on, curr_char is guaranteed to be in '+-' elif not self.items: token = Token(curr_char, Token.OP_PRE) else: prev = next((i for i in reversed(self.items) if i.type != Token.WSPACE), None) is_infix = prev and ( prev.subtype == Token.CLOSE or prev.type == Token.OP_POST or prev.type == Token.OPERAND ) if is_infix: token = Token(curr_char, Token.OP_IN) else: token = Token(curr_char, Token.OP_PRE) self.items.append(token) return 1 def _parse_opener(self): """ Consumes a ( or { character. Returns the number of characters consumed. (Does not update self.offset) """ assert self.formula[self.offset] in ('(', '{') if self.formula[self.offset] == '{': self.assert_empty_token() token = Token.make_subexp("{") elif self.token: token_value = "".join(self.token) + '(' del self.token[:] token = Token.make_subexp(token_value) else: token = Token.make_subexp("(") self.items.append(token) self.token_stack.append(token) return 1 def _parse_closer(self): """ Consumes a } or ) character. Returns the number of characters consumed. (Does not update self.offset) """ assert self.formula[self.offset] in (')', '}') token = self.token_stack.pop().get_closer() if token.value != self.formula[self.offset]: raise TokenizerError( "Mismatched ( and { pair in '%s'" % self.formula) self.items.append(token) return 1 def _parse_separator(self): """ Consumes a ; or , character. Returns the number of characters consumed. (Does not update self.offset) """ curr_char = self.formula[self.offset] assert curr_char in (';', ',') if curr_char == ';': token = Token.make_separator(";") else: try: top_type = self.token_stack[-1].type except IndexError: token = Token(",", Token.OP_IN) # Range Union operator else: if top_type == Token.PAREN: token = Token(",", Token.OP_IN) # Range Union operator else: token = Token.make_separator(",") self.items.append(token) return 1 def check_scientific_notation(self): """ Consumes a + or - character if part of a number in sci. notation. Returns True if the character was consumed and self.offset was updated, False otherwise. """ curr_char = self.formula[self.offset] if (curr_char in '+-' and len(self.token) >= 1 and self.SN_RE.match("".join(self.token))): self.token.append(curr_char) self.offset += 1 return True return False def assert_empty_token(self, can_follow=()): """ Ensure that there's no token currently being parsed. Or if there is a token being parsed, it must end with a character in can_follow. If there are unconsumed token contents, it means we hit an unexpected token transition. In this case, we raise a TokenizerError """ if self.token and self.token[-1] not in can_follow: raise TokenizerError( "Unexpected character at position %d in '%s'" % (self.offset, self.formula)) def save_token(self): """If there's a token being parsed, add it to the item list.""" if self.token: self.items.append(Token.make_operand("".join(self.token))) del self.token[:] def render(self): """Convert the parsed tokens back to a string.""" if not self.items: return "" elif self.items[0].type == Token.LITERAL: return self.items[0].value return "=" + "".join(token.value for token in self.items) class Token(object): """ A token in an Excel formula. Tokens have three attributes: * `value`: The string value parsed that led to this token * `type`: A string identifying the type of token * `subtype`: A string identifying subtype of the token (optional, and defaults to "") """ __slots__ = ['value', 'type', 'subtype'] LITERAL = "LITERAL" OPERAND = "OPERAND" FUNC = "FUNC" ARRAY = "ARRAY" PAREN = "PAREN" SEP = "SEP" OP_PRE = "OPERATOR-PREFIX" OP_IN = "OPERATOR-INFIX" OP_POST = "OPERATOR-POSTFIX" WSPACE = "WHITE-SPACE" def __init__(self, value, type_, subtype=""): self.value = value self.type = type_ self.subtype = subtype # Literal operands: # # Literal operands are always of type 'OPERAND' and can be of subtype # 'TEXT' (for text strings), 'NUMBER' (for all numeric types), 'LOGICAL' # (for TRUE and FALSE), 'ERROR' (for literal error values), or 'RANGE' # (for all range references). TEXT = 'TEXT' NUMBER = 'NUMBER' LOGICAL = 'LOGICAL' ERROR = 'ERROR' RANGE = 'RANGE' def __repr__(self): return u"{0} {1} {2}:".format(self.type, self.subtype, self.value) @classmethod def make_operand(cls, value): """Create an operand token.""" if value.startswith('"'): subtype = cls.TEXT elif value.startswith('#'): subtype = cls.ERROR elif value in ('TRUE', 'FALSE'): subtype = cls.LOGICAL else: try: float(value) subtype = cls.NUMBER except ValueError: subtype = cls.RANGE return cls(value, cls.OPERAND, subtype) # Subexpresssions # # There are 3 types of `Subexpressions`: functions, array literals, and # parentheticals. Subexpressions have 'OPEN' and 'CLOSE' tokens. 'OPEN' # is used when parsing the initial expression token (i.e., '(' or '{') # and 'CLOSE' is used when parsing the closing expression token ('}' or # ')'). OPEN = "OPEN" CLOSE = "CLOSE" @classmethod def make_subexp(cls, value, func=False): """ Create a subexpression token. `value`: The value of the token `func`: If True, force the token to be of type FUNC """ assert value[-1] in ('{', '}', '(', ')') if func: assert re.match('.+\\(|\\)', value) type_ = Token.FUNC elif value in '{}': type_ = Token.ARRAY elif value in '()': type_ = Token.PAREN else: type_ = Token.FUNC subtype = cls.CLOSE if value in ')}' else cls.OPEN return cls(value, type_, subtype) def get_closer(self): """Return a closing token that matches this token's type.""" assert self.type in (self.FUNC, self.ARRAY, self.PAREN) assert self.subtype == self.OPEN value = "}" if self.type == self.ARRAY else ")" return self.make_subexp(value, func=self.type == self.FUNC) # Separator tokens # # Argument separators always have type 'SEP' and can have one of two # subtypes: 'ARG', 'ROW'. 'ARG' is used for the ',' token, when used to # delimit either function arguments or array elements. 'ROW' is used for # the ';' token, which is always used to delimit rows in an array # literal. ARG = "ARG" ROW = "ROW" @classmethod def make_separator(cls, value): """Create a separator token""" assert value in (',', ';') subtype = cls.ARG if value == ',' else cls.ROW return cls(value, cls.SEP, subtype)