3 This file is part of libmspstrings
4 Copyright © 2007 Mikko Rasa
5 Distributed under the LGPL
7 #ifndef MSP_STRINGS_REGEX_H_
8 #define MSP_STRINGS_REGEX_H_
16 This class provides regular expression matching. It supports a subset of
17 POSIX.2 extended regex syntax. Character classes, equivalence classes and
18 collating elements are not supported. Refer to the regex(7) manpage for more
21 A description of the internal workings of the class follows. You may skip this
22 if you are not a developer or otherwise interested.
24 The core of this class is a virtual machine running a non-deterministic finite
25 automaton (NFA). The state of the automaton is represented by an iterator into
26 the code. Transitions between states are represented by instructions.
27 Instructions may take multiple bytes, so a valid code iterator may not be a
30 The virtual machine may have any number of execution contexts at any one time.
31 On every cycle, each context is advanced until an input-consuming instruction
32 is encountered, keeping the input positions in sync with each other. Execution
33 continues as long as there's at least one context remaining.
35 The GROUP_BEGIN and GROUP_END instructions record the begin and end offset of a
36 match group, respectively. The GROUP_END instruction also marks the group as
37 successfully matched. If the target group was already matched, these
38 instructions do nothing.
40 The JUMP instruction causes the execution iterator to be modified by an offset.
42 The ND_JUMP instruction causes the execution context to be split in two. One
43 continues directly after the instruction and the other continues at an offset.
45 The NEGATE instruction causes the result of the next match instruction to be
48 Match instructions compare the input against a condition. If the match
49 succeeds, execution continues at the next instruction. If the match fails,
50 execution of that context is terminated.
52 The MATCH_BEGIN and MATCH_END instructions match the beginning and end of
53 the input string, respectively. They do not consume input.
55 The MATCH_CHAR instruction consumes the input character and matches it against
56 a single character. Since regexes often match sequences of printable character,
57 a match for a non-opcode character may be encoded as the character itself.
59 The MATCH_RANGE instruction consumes the input character and matches it against
60 an inclusive character range.
62 The MATCH_MASK instruction consumes the input character and matches it against
65 The MATCH_ANY instruction consumes the input character and always succeeds.
71 Constructs a new Regex object.
73 Regex(const std::string &expr);
76 Matches the regex against a string. Refer to RegMatch documentation for
77 more information on the resulting object.
79 RegMatch match(const std::string &str) const;
82 Returns a disassembled representation of the NFA bytecode. For debugging
85 std::string disassemble() const;
87 typedef std::string Code;
88 typedef unsigned short Count;
90 typedef unsigned short Index;
116 Code::const_iterator citer;
117 RegMatch::GroupArray groups;
124 Compiles a regular expression into NFA bytecode. When compiling a whole
125 regex, \a group should be set to 0. When the function returns, \a group will
126 be the index of the last subexpression and \a iter will point to the first
127 unused character in the expression.
129 \param expr Expression to be compiled
130 \param begin Iterator into the expression
131 \param group Group counter, gets incremented for each subregex
132 \param branch Whether we are compiling a branch
134 \return Compiled NFA bytecode
136 Code compile(const std::string &expr, std::string::const_iterator &iter, unsigned &group, bool branch);
137 Code parse_atom(const std::string &, std::string::const_iterator &i, unsigned &);
138 Code parse_brackets(const std::string &, std::string::const_iterator &);
139 bool parse_repeat(std::string::const_iterator &, Count &, Count &);
140 bool run(const std::string &, const std::string::const_iterator &, RegMatch::GroupArray &) const;
141 bool group_compare(const RegMatch::Group &, const RegMatch::Group &) const;
142 std::string disassemble_instruction(Code::const_iterator &) const;