1 #ifndef MSP_STRINGS_REGEX_H_
2 #define MSP_STRINGS_REGEX_H_
10 class bad_regex: public std::logic_error
13 bad_regex(const std::string &, const std::string &, const std::string::const_iterator &);
14 virtual ~bad_regex() throw() { }
17 std::string make_where(const std::string &, const std::string::const_iterator &);
22 This class provides regular expression matching. It supports a subset of
23 POSIX.2 extended regex syntax. Character classes, equivalence classes and
24 collating elements are not supported. Refer to the regex(7) manpage for more
27 A description of the internal workings of the class follows. You may skip this
28 if you are not a developer or otherwise interested.
30 The core of this class is a virtual machine running a non-deterministic finite
31 automaton (NFA). The state of the automaton is represented by an iterator into
32 the code. Transitions between states are represented by instructions.
33 Instructions may take multiple bytes, so a valid code iterator may not be a
36 The virtual machine may have any number of execution contexts at any one time.
37 On every cycle, each context is advanced until an input-consuming instruction
38 is encountered, keeping the input positions in sync with each other. Execution
39 continues as long as there's at least one context remaining.
41 The GROUP_BEGIN and GROUP_END instructions record the begin and end offset of a
42 match group, respectively. The GROUP_END instruction also marks the group as
43 successfully matched. If the target group was already matched, these
44 instructions do nothing.
46 The JUMP instruction causes the execution iterator to be modified by an offset.
48 The ND_JUMP instruction causes the execution context to be split in two. One
49 continues directly after the instruction and the other continues at an offset.
51 The NEGATE instruction causes the result of the next match instruction to be
54 Match instructions compare the input against a condition. If the match
55 succeeds, execution continues at the next instruction. If the match fails,
56 execution of that context is terminated.
58 The MATCH_BEGIN and MATCH_END instructions match the beginning and end of
59 the input string, respectively. They do not consume input.
61 The MATCH_CHAR instruction consumes the input character and matches it against
62 a single character. Since regexes often match sequences of printable character,
63 a match for a non-opcode character may be encoded as the character itself.
65 The MATCH_RANGE instruction consumes the input character and matches it against
66 an inclusive character range.
68 The MATCH_MASK instruction consumes the input character and matches it against
71 The MATCH_ANY instruction consumes the input character and always succeeds.
76 typedef std::basic_string<unsigned char> Code;
77 typedef unsigned short Count;
79 typedef unsigned short Index;
83 FIRST_INSTRUCTION_ = 0,
100 LAST_INSTRUCTION_ = 31
105 Code::const_iterator citer;
106 RegMatch::GroupArray groups;
113 /** Constructs a new Regex object from a string representation. */
114 Regex(const std::string &expr);
117 /** Compiles a regular expression into NFA bytecode. , 2011The iterator will be
118 advanced to the first unused character in the string. */
119 Code compile(const std::string &expr, std::string::const_iterator &iter, unsigned &group, bool branch);
121 Code parse_atom(const std::string &, std::string::const_iterator &i, unsigned &);
122 Code parse_brackets(const std::string &, std::string::const_iterator &);
123 bool parse_repeat(const std::string &, std::string::const_iterator &, Count &, Count &);
126 /** Matches the regex against a string. Refer to RegMatch documentation for
127 more information on the resulting object. */
128 RegMatch match(const std::string &str) const;
131 bool run(const std::string &, const std::string::const_iterator &, RegMatch::GroupArray &) const;
132 bool group_compare(const RegMatch::Group &, const RegMatch::Group &) const;
135 /** Returns a disassembled representation of the NFA bytecode. For debugging
137 std::string disassemble() const;
139 std::string disassemble_instruction(Code::const_iterator &) const;