Jeff Thompson | 86b6d64 | 2013-10-17 15:01:56 -0700 | [diff] [blame] | 1 | /* |
| 2 | * |
| 3 | * Copyright (c) 2004 |
| 4 | * John Maddock |
| 5 | * |
| 6 | * Use, modification and distribution are subject to the |
| 7 | * Boost Software License, Version 1.0. (See accompanying file |
| 8 | * LICENSE_1_0.txt or copy at http://www.boost.org/LICENSE_1_0.txt) |
| 9 | * |
| 10 | */ |
| 11 | |
| 12 | /* |
| 13 | * LOCATION: see http://www.boost.org for most recent version. |
| 14 | * FILE basic_regex_creator.cpp |
| 15 | * VERSION see <ndnboost/version.hpp> |
| 16 | * DESCRIPTION: Declares template class basic_regex_creator which fills in |
| 17 | * the data members of a regex_data object. |
| 18 | */ |
| 19 | |
| 20 | #ifndef NDNBOOST_REGEX_V4_BASIC_REGEX_CREATOR_HPP |
| 21 | #define NDNBOOST_REGEX_V4_BASIC_REGEX_CREATOR_HPP |
| 22 | |
| 23 | #ifdef NDNBOOST_MSVC |
| 24 | #pragma warning(push) |
| 25 | #pragma warning(disable: 4103) |
| 26 | #endif |
| 27 | #ifdef NDNBOOST_HAS_ABI_HEADERS |
| 28 | # include NDNBOOST_ABI_PREFIX |
| 29 | #endif |
| 30 | #ifdef NDNBOOST_MSVC |
| 31 | #pragma warning(pop) |
| 32 | #endif |
| 33 | |
| 34 | #ifdef NDNBOOST_MSVC |
| 35 | # pragma warning(push) |
| 36 | # pragma warning(disable: 4800) |
| 37 | #endif |
| 38 | |
| 39 | namespace ndnboost{ |
| 40 | |
| 41 | namespace re_detail{ |
| 42 | |
| 43 | template <class charT> |
| 44 | struct digraph : public std::pair<charT, charT> |
| 45 | { |
| 46 | digraph() : std::pair<charT, charT>(0, 0){} |
| 47 | digraph(charT c1) : std::pair<charT, charT>(c1, 0){} |
| 48 | digraph(charT c1, charT c2) : std::pair<charT, charT>(c1, c2) |
| 49 | {} |
| 50 | #if !NDNBOOST_WORKAROUND(NDNBOOST_MSVC, < 1300) |
| 51 | digraph(const digraph<charT>& d) : std::pair<charT, charT>(d.first, d.second){} |
| 52 | #endif |
| 53 | template <class Seq> |
| 54 | digraph(const Seq& s) : std::pair<charT, charT>() |
| 55 | { |
| 56 | NDNBOOST_ASSERT(s.size() <= 2); |
| 57 | NDNBOOST_ASSERT(s.size()); |
| 58 | this->first = s[0]; |
| 59 | this->second = (s.size() > 1) ? s[1] : 0; |
| 60 | } |
| 61 | }; |
| 62 | |
| 63 | template <class charT, class traits> |
| 64 | class basic_char_set |
| 65 | { |
| 66 | public: |
| 67 | typedef digraph<charT> digraph_type; |
| 68 | typedef typename traits::string_type string_type; |
| 69 | typedef typename traits::char_class_type m_type; |
| 70 | |
| 71 | basic_char_set() |
| 72 | { |
| 73 | m_negate = false; |
| 74 | m_has_digraphs = false; |
| 75 | m_classes = 0; |
| 76 | m_negated_classes = 0; |
| 77 | m_empty = true; |
| 78 | } |
| 79 | |
| 80 | void add_single(const digraph_type& s) |
| 81 | { |
| 82 | m_singles.insert(m_singles.end(), s); |
| 83 | if(s.second) |
| 84 | m_has_digraphs = true; |
| 85 | m_empty = false; |
| 86 | } |
| 87 | void add_range(const digraph_type& first, const digraph_type& end) |
| 88 | { |
| 89 | m_ranges.insert(m_ranges.end(), first); |
| 90 | m_ranges.insert(m_ranges.end(), end); |
| 91 | if(first.second) |
| 92 | { |
| 93 | m_has_digraphs = true; |
| 94 | add_single(first); |
| 95 | } |
| 96 | if(end.second) |
| 97 | { |
| 98 | m_has_digraphs = true; |
| 99 | add_single(end); |
| 100 | } |
| 101 | m_empty = false; |
| 102 | } |
| 103 | void add_class(m_type m) |
| 104 | { |
| 105 | m_classes |= m; |
| 106 | m_empty = false; |
| 107 | } |
| 108 | void add_negated_class(m_type m) |
| 109 | { |
| 110 | m_negated_classes |= m; |
| 111 | m_empty = false; |
| 112 | } |
| 113 | void add_equivalent(const digraph_type& s) |
| 114 | { |
| 115 | m_equivalents.insert(m_equivalents.end(), s); |
| 116 | if(s.second) |
| 117 | { |
| 118 | m_has_digraphs = true; |
| 119 | add_single(s); |
| 120 | } |
| 121 | m_empty = false; |
| 122 | } |
| 123 | void negate() |
| 124 | { |
| 125 | m_negate = true; |
| 126 | //m_empty = false; |
| 127 | } |
| 128 | |
| 129 | // |
| 130 | // accessor functions: |
| 131 | // |
| 132 | bool has_digraphs()const |
| 133 | { |
| 134 | return m_has_digraphs; |
| 135 | } |
| 136 | bool is_negated()const |
| 137 | { |
| 138 | return m_negate; |
| 139 | } |
| 140 | typedef typename std::vector<digraph_type>::const_iterator list_iterator; |
| 141 | list_iterator singles_begin()const |
| 142 | { |
| 143 | return m_singles.begin(); |
| 144 | } |
| 145 | list_iterator singles_end()const |
| 146 | { |
| 147 | return m_singles.end(); |
| 148 | } |
| 149 | list_iterator ranges_begin()const |
| 150 | { |
| 151 | return m_ranges.begin(); |
| 152 | } |
| 153 | list_iterator ranges_end()const |
| 154 | { |
| 155 | return m_ranges.end(); |
| 156 | } |
| 157 | list_iterator equivalents_begin()const |
| 158 | { |
| 159 | return m_equivalents.begin(); |
| 160 | } |
| 161 | list_iterator equivalents_end()const |
| 162 | { |
| 163 | return m_equivalents.end(); |
| 164 | } |
| 165 | m_type classes()const |
| 166 | { |
| 167 | return m_classes; |
| 168 | } |
| 169 | m_type negated_classes()const |
| 170 | { |
| 171 | return m_negated_classes; |
| 172 | } |
| 173 | bool empty()const |
| 174 | { |
| 175 | return m_empty; |
| 176 | } |
| 177 | private: |
| 178 | std::vector<digraph_type> m_singles; // a list of single characters to match |
| 179 | std::vector<digraph_type> m_ranges; // a list of end points of our ranges |
| 180 | bool m_negate; // true if the set is to be negated |
| 181 | bool m_has_digraphs; // true if we have digraphs present |
| 182 | m_type m_classes; // character classes to match |
| 183 | m_type m_negated_classes; // negated character classes to match |
| 184 | bool m_empty; // whether we've added anything yet |
| 185 | std::vector<digraph_type> m_equivalents; // a list of equivalence classes |
| 186 | }; |
| 187 | |
| 188 | template <class charT, class traits> |
| 189 | class basic_regex_creator |
| 190 | { |
| 191 | public: |
| 192 | basic_regex_creator(regex_data<charT, traits>* data); |
| 193 | std::ptrdiff_t getoffset(void* addr) |
| 194 | { |
| 195 | return getoffset(addr, m_pdata->m_data.data()); |
| 196 | } |
| 197 | std::ptrdiff_t getoffset(const void* addr, const void* base) |
| 198 | { |
| 199 | return static_cast<const char*>(addr) - static_cast<const char*>(base); |
| 200 | } |
| 201 | re_syntax_base* getaddress(std::ptrdiff_t off) |
| 202 | { |
| 203 | return getaddress(off, m_pdata->m_data.data()); |
| 204 | } |
| 205 | re_syntax_base* getaddress(std::ptrdiff_t off, void* base) |
| 206 | { |
| 207 | return static_cast<re_syntax_base*>(static_cast<void*>(static_cast<char*>(base) + off)); |
| 208 | } |
| 209 | void init(unsigned l_flags) |
| 210 | { |
| 211 | m_pdata->m_flags = l_flags; |
| 212 | m_icase = l_flags & regex_constants::icase; |
| 213 | } |
| 214 | regbase::flag_type flags() |
| 215 | { |
| 216 | return m_pdata->m_flags; |
| 217 | } |
| 218 | void flags(regbase::flag_type f) |
| 219 | { |
| 220 | m_pdata->m_flags = f; |
| 221 | if(m_icase != static_cast<bool>(f & regbase::icase)) |
| 222 | { |
| 223 | m_icase = static_cast<bool>(f & regbase::icase); |
| 224 | } |
| 225 | } |
| 226 | re_syntax_base* append_state(syntax_element_type t, std::size_t s = sizeof(re_syntax_base)); |
| 227 | re_syntax_base* insert_state(std::ptrdiff_t pos, syntax_element_type t, std::size_t s = sizeof(re_syntax_base)); |
| 228 | re_literal* append_literal(charT c); |
| 229 | re_syntax_base* append_set(const basic_char_set<charT, traits>& char_set); |
| 230 | re_syntax_base* append_set(const basic_char_set<charT, traits>& char_set, mpl::false_*); |
| 231 | re_syntax_base* append_set(const basic_char_set<charT, traits>& char_set, mpl::true_*); |
| 232 | void finalize(const charT* p1, const charT* p2); |
| 233 | protected: |
| 234 | regex_data<charT, traits>* m_pdata; // pointer to the basic_regex_data struct we are filling in |
| 235 | const ::ndnboost::regex_traits_wrapper<traits>& |
| 236 | m_traits; // convenience reference to traits class |
| 237 | re_syntax_base* m_last_state; // the last state we added |
| 238 | bool m_icase; // true for case insensitive matches |
| 239 | unsigned m_repeater_id; // the state_id of the next repeater |
| 240 | bool m_has_backrefs; // true if there are actually any backrefs |
| 241 | unsigned m_backrefs; // bitmask of permitted backrefs |
| 242 | ndnboost::uintmax_t m_bad_repeats; // bitmask of repeats we can't deduce a startmap for; |
| 243 | bool m_has_recursions; // set when we have recursive expresisons to fixup |
| 244 | std::vector<bool> m_recursion_checks; // notes which recursions we've followed while analysing this expression |
| 245 | typename traits::char_class_type m_word_mask; // mask used to determine if a character is a word character |
| 246 | typename traits::char_class_type m_mask_space; // mask used to determine if a character is a word character |
| 247 | typename traits::char_class_type m_lower_mask; // mask used to determine if a character is a lowercase character |
| 248 | typename traits::char_class_type m_upper_mask; // mask used to determine if a character is an uppercase character |
| 249 | typename traits::char_class_type m_alpha_mask; // mask used to determine if a character is an alphabetic character |
| 250 | private: |
| 251 | basic_regex_creator& operator=(const basic_regex_creator&); |
| 252 | basic_regex_creator(const basic_regex_creator&); |
| 253 | |
| 254 | void fixup_pointers(re_syntax_base* state); |
| 255 | void fixup_recursions(re_syntax_base* state); |
| 256 | void create_startmaps(re_syntax_base* state); |
| 257 | int calculate_backstep(re_syntax_base* state); |
| 258 | void create_startmap(re_syntax_base* state, unsigned char* l_map, unsigned int* pnull, unsigned char mask); |
| 259 | unsigned get_restart_type(re_syntax_base* state); |
| 260 | void set_all_masks(unsigned char* bits, unsigned char); |
| 261 | bool is_bad_repeat(re_syntax_base* pt); |
| 262 | void set_bad_repeat(re_syntax_base* pt); |
| 263 | syntax_element_type get_repeat_type(re_syntax_base* state); |
| 264 | void probe_leading_repeat(re_syntax_base* state); |
| 265 | }; |
| 266 | |
| 267 | template <class charT, class traits> |
| 268 | basic_regex_creator<charT, traits>::basic_regex_creator(regex_data<charT, traits>* data) |
| 269 | : m_pdata(data), m_traits(*(data->m_ptraits)), m_last_state(0), m_repeater_id(0), m_has_backrefs(false), m_backrefs(0), m_has_recursions(false) |
| 270 | { |
| 271 | m_pdata->m_data.clear(); |
| 272 | m_pdata->m_status = ::ndnboost::regex_constants::error_ok; |
| 273 | static const charT w = 'w'; |
| 274 | static const charT s = 's'; |
| 275 | static const charT l[5] = { 'l', 'o', 'w', 'e', 'r', }; |
| 276 | static const charT u[5] = { 'u', 'p', 'p', 'e', 'r', }; |
| 277 | static const charT a[5] = { 'a', 'l', 'p', 'h', 'a', }; |
| 278 | m_word_mask = m_traits.lookup_classname(&w, &w +1); |
| 279 | m_mask_space = m_traits.lookup_classname(&s, &s +1); |
| 280 | m_lower_mask = m_traits.lookup_classname(l, l + 5); |
| 281 | m_upper_mask = m_traits.lookup_classname(u, u + 5); |
| 282 | m_alpha_mask = m_traits.lookup_classname(a, a + 5); |
| 283 | m_pdata->m_word_mask = m_word_mask; |
| 284 | NDNBOOST_ASSERT(m_word_mask != 0); |
| 285 | NDNBOOST_ASSERT(m_mask_space != 0); |
| 286 | NDNBOOST_ASSERT(m_lower_mask != 0); |
| 287 | NDNBOOST_ASSERT(m_upper_mask != 0); |
| 288 | NDNBOOST_ASSERT(m_alpha_mask != 0); |
| 289 | } |
| 290 | |
| 291 | template <class charT, class traits> |
| 292 | re_syntax_base* basic_regex_creator<charT, traits>::append_state(syntax_element_type t, std::size_t s) |
| 293 | { |
| 294 | // if the state is a backref then make a note of it: |
| 295 | if(t == syntax_element_backref) |
| 296 | this->m_has_backrefs = true; |
| 297 | // append a new state, start by aligning our last one: |
| 298 | m_pdata->m_data.align(); |
| 299 | // set the offset to the next state in our last one: |
| 300 | if(m_last_state) |
| 301 | m_last_state->next.i = m_pdata->m_data.size() - getoffset(m_last_state); |
| 302 | // now actually extent our data: |
| 303 | m_last_state = static_cast<re_syntax_base*>(m_pdata->m_data.extend(s)); |
| 304 | // fill in boilerplate options in the new state: |
| 305 | m_last_state->next.i = 0; |
| 306 | m_last_state->type = t; |
| 307 | return m_last_state; |
| 308 | } |
| 309 | |
| 310 | template <class charT, class traits> |
| 311 | re_syntax_base* basic_regex_creator<charT, traits>::insert_state(std::ptrdiff_t pos, syntax_element_type t, std::size_t s) |
| 312 | { |
| 313 | // append a new state, start by aligning our last one: |
| 314 | m_pdata->m_data.align(); |
| 315 | // set the offset to the next state in our last one: |
| 316 | if(m_last_state) |
| 317 | m_last_state->next.i = m_pdata->m_data.size() - getoffset(m_last_state); |
| 318 | // remember the last state position: |
| 319 | std::ptrdiff_t off = getoffset(m_last_state) + s; |
| 320 | // now actually insert our data: |
| 321 | re_syntax_base* new_state = static_cast<re_syntax_base*>(m_pdata->m_data.insert(pos, s)); |
| 322 | // fill in boilerplate options in the new state: |
| 323 | new_state->next.i = s; |
| 324 | new_state->type = t; |
| 325 | m_last_state = getaddress(off); |
| 326 | return new_state; |
| 327 | } |
| 328 | |
| 329 | template <class charT, class traits> |
| 330 | re_literal* basic_regex_creator<charT, traits>::append_literal(charT c) |
| 331 | { |
| 332 | re_literal* result; |
| 333 | // start by seeing if we have an existing re_literal we can extend: |
| 334 | if((0 == m_last_state) || (m_last_state->type != syntax_element_literal)) |
| 335 | { |
| 336 | // no existing re_literal, create a new one: |
| 337 | result = static_cast<re_literal*>(append_state(syntax_element_literal, sizeof(re_literal) + sizeof(charT))); |
| 338 | result->length = 1; |
| 339 | *static_cast<charT*>(static_cast<void*>(result+1)) = m_traits.translate(c, m_icase); |
| 340 | } |
| 341 | else |
| 342 | { |
| 343 | // we have an existing re_literal, extend it: |
| 344 | std::ptrdiff_t off = getoffset(m_last_state); |
| 345 | m_pdata->m_data.extend(sizeof(charT)); |
| 346 | m_last_state = result = static_cast<re_literal*>(getaddress(off)); |
| 347 | charT* characters = static_cast<charT*>(static_cast<void*>(result+1)); |
| 348 | characters[result->length] = m_traits.translate(c, m_icase); |
| 349 | ++(result->length); |
| 350 | } |
| 351 | return result; |
| 352 | } |
| 353 | |
| 354 | template <class charT, class traits> |
| 355 | inline re_syntax_base* basic_regex_creator<charT, traits>::append_set( |
| 356 | const basic_char_set<charT, traits>& char_set) |
| 357 | { |
| 358 | typedef mpl::bool_< (sizeof(charT) == 1) > truth_type; |
| 359 | return char_set.has_digraphs() |
| 360 | ? append_set(char_set, static_cast<mpl::false_*>(0)) |
| 361 | : append_set(char_set, static_cast<truth_type*>(0)); |
| 362 | } |
| 363 | |
| 364 | template <class charT, class traits> |
| 365 | re_syntax_base* basic_regex_creator<charT, traits>::append_set( |
| 366 | const basic_char_set<charT, traits>& char_set, mpl::false_*) |
| 367 | { |
| 368 | typedef typename traits::string_type string_type; |
| 369 | typedef typename basic_char_set<charT, traits>::list_iterator item_iterator; |
| 370 | typedef typename traits::char_class_type m_type; |
| 371 | |
| 372 | re_set_long<m_type>* result = static_cast<re_set_long<m_type>*>(append_state(syntax_element_long_set, sizeof(re_set_long<m_type>))); |
| 373 | // |
| 374 | // fill in the basics: |
| 375 | // |
| 376 | result->csingles = static_cast<unsigned int>(::ndnboost::re_detail::distance(char_set.singles_begin(), char_set.singles_end())); |
| 377 | result->cranges = static_cast<unsigned int>(::ndnboost::re_detail::distance(char_set.ranges_begin(), char_set.ranges_end())) / 2; |
| 378 | result->cequivalents = static_cast<unsigned int>(::ndnboost::re_detail::distance(char_set.equivalents_begin(), char_set.equivalents_end())); |
| 379 | result->cclasses = char_set.classes(); |
| 380 | result->cnclasses = char_set.negated_classes(); |
| 381 | if(flags() & regbase::icase) |
| 382 | { |
| 383 | // adjust classes as needed: |
| 384 | if(((result->cclasses & m_lower_mask) == m_lower_mask) || ((result->cclasses & m_upper_mask) == m_upper_mask)) |
| 385 | result->cclasses |= m_alpha_mask; |
| 386 | if(((result->cnclasses & m_lower_mask) == m_lower_mask) || ((result->cnclasses & m_upper_mask) == m_upper_mask)) |
| 387 | result->cnclasses |= m_alpha_mask; |
| 388 | } |
| 389 | |
| 390 | result->isnot = char_set.is_negated(); |
| 391 | result->singleton = !char_set.has_digraphs(); |
| 392 | // |
| 393 | // remember where the state is for later: |
| 394 | // |
| 395 | std::ptrdiff_t offset = getoffset(result); |
| 396 | // |
| 397 | // now extend with all the singles: |
| 398 | // |
| 399 | item_iterator first, last; |
| 400 | first = char_set.singles_begin(); |
| 401 | last = char_set.singles_end(); |
| 402 | while(first != last) |
| 403 | { |
| 404 | charT* p = static_cast<charT*>(this->m_pdata->m_data.extend(sizeof(charT) * (first->second ? 3 : 2))); |
| 405 | p[0] = m_traits.translate(first->first, m_icase); |
| 406 | if(first->second) |
| 407 | { |
| 408 | p[1] = m_traits.translate(first->second, m_icase); |
| 409 | p[2] = 0; |
| 410 | } |
| 411 | else |
| 412 | p[1] = 0; |
| 413 | ++first; |
| 414 | } |
| 415 | // |
| 416 | // now extend with all the ranges: |
| 417 | // |
| 418 | first = char_set.ranges_begin(); |
| 419 | last = char_set.ranges_end(); |
| 420 | while(first != last) |
| 421 | { |
| 422 | // first grab the endpoints of the range: |
| 423 | digraph<charT> c1 = *first; |
| 424 | c1.first = this->m_traits.translate(c1.first, this->m_icase); |
| 425 | c1.second = this->m_traits.translate(c1.second, this->m_icase); |
| 426 | ++first; |
| 427 | digraph<charT> c2 = *first; |
| 428 | c2.first = this->m_traits.translate(c2.first, this->m_icase); |
| 429 | c2.second = this->m_traits.translate(c2.second, this->m_icase); |
| 430 | ++first; |
| 431 | string_type s1, s2; |
| 432 | // different actions now depending upon whether collation is turned on: |
| 433 | if(flags() & regex_constants::collate) |
| 434 | { |
| 435 | // we need to transform our range into sort keys: |
| 436 | #if NDNBOOST_WORKAROUND(__GNUC__, < 3) |
| 437 | string_type in(3, charT(0)); |
| 438 | in[0] = c1.first; |
| 439 | in[1] = c1.second; |
| 440 | s1 = this->m_traits.transform(in.c_str(), (in[1] ? in.c_str()+2 : in.c_str()+1)); |
| 441 | in[0] = c2.first; |
| 442 | in[1] = c2.second; |
| 443 | s2 = this->m_traits.transform(in.c_str(), (in[1] ? in.c_str()+2 : in.c_str()+1)); |
| 444 | #else |
| 445 | charT a1[3] = { c1.first, c1.second, charT(0), }; |
| 446 | charT a2[3] = { c2.first, c2.second, charT(0), }; |
| 447 | s1 = this->m_traits.transform(a1, (a1[1] ? a1+2 : a1+1)); |
| 448 | s2 = this->m_traits.transform(a2, (a2[1] ? a2+2 : a2+1)); |
| 449 | #endif |
| 450 | if(s1.size() == 0) |
| 451 | s1 = string_type(1, charT(0)); |
| 452 | if(s2.size() == 0) |
| 453 | s2 = string_type(1, charT(0)); |
| 454 | } |
| 455 | else |
| 456 | { |
| 457 | if(c1.second) |
| 458 | { |
| 459 | s1.insert(s1.end(), c1.first); |
| 460 | s1.insert(s1.end(), c1.second); |
| 461 | } |
| 462 | else |
| 463 | s1 = string_type(1, c1.first); |
| 464 | if(c2.second) |
| 465 | { |
| 466 | s2.insert(s2.end(), c2.first); |
| 467 | s2.insert(s2.end(), c2.second); |
| 468 | } |
| 469 | else |
| 470 | s2.insert(s2.end(), c2.first); |
| 471 | } |
| 472 | if(s1 > s2) |
| 473 | { |
| 474 | // Oops error: |
| 475 | return 0; |
| 476 | } |
| 477 | charT* p = static_cast<charT*>(this->m_pdata->m_data.extend(sizeof(charT) * (s1.size() + s2.size() + 2) ) ); |
| 478 | re_detail::copy(s1.begin(), s1.end(), p); |
| 479 | p[s1.size()] = charT(0); |
| 480 | p += s1.size() + 1; |
| 481 | re_detail::copy(s2.begin(), s2.end(), p); |
| 482 | p[s2.size()] = charT(0); |
| 483 | } |
| 484 | // |
| 485 | // now process the equivalence classes: |
| 486 | // |
| 487 | first = char_set.equivalents_begin(); |
| 488 | last = char_set.equivalents_end(); |
| 489 | while(first != last) |
| 490 | { |
| 491 | string_type s; |
| 492 | if(first->second) |
| 493 | { |
| 494 | #if NDNBOOST_WORKAROUND(__GNUC__, < 3) |
| 495 | string_type in(3, charT(0)); |
| 496 | in[0] = first->first; |
| 497 | in[1] = first->second; |
| 498 | s = m_traits.transform_primary(in.c_str(), in.c_str()+2); |
| 499 | #else |
| 500 | charT cs[3] = { first->first, first->second, charT(0), }; |
| 501 | s = m_traits.transform_primary(cs, cs+2); |
| 502 | #endif |
| 503 | } |
| 504 | else |
| 505 | s = m_traits.transform_primary(&first->first, &first->first+1); |
| 506 | if(s.empty()) |
| 507 | return 0; // invalid or unsupported equivalence class |
| 508 | charT* p = static_cast<charT*>(this->m_pdata->m_data.extend(sizeof(charT) * (s.size()+1) ) ); |
| 509 | re_detail::copy(s.begin(), s.end(), p); |
| 510 | p[s.size()] = charT(0); |
| 511 | ++first; |
| 512 | } |
| 513 | // |
| 514 | // finally reset the address of our last state: |
| 515 | // |
| 516 | m_last_state = result = static_cast<re_set_long<m_type>*>(getaddress(offset)); |
| 517 | return result; |
| 518 | } |
| 519 | |
| 520 | template<class T> |
| 521 | inline bool char_less(T t1, T t2) |
| 522 | { |
| 523 | return t1 < t2; |
| 524 | } |
| 525 | inline bool char_less(char t1, char t2) |
| 526 | { |
| 527 | return static_cast<unsigned char>(t1) < static_cast<unsigned char>(t2); |
| 528 | } |
| 529 | inline bool char_less(signed char t1, signed char t2) |
| 530 | { |
| 531 | return static_cast<unsigned char>(t1) < static_cast<unsigned char>(t2); |
| 532 | } |
| 533 | |
| 534 | template <class charT, class traits> |
| 535 | re_syntax_base* basic_regex_creator<charT, traits>::append_set( |
| 536 | const basic_char_set<charT, traits>& char_set, mpl::true_*) |
| 537 | { |
| 538 | typedef typename traits::string_type string_type; |
| 539 | typedef typename basic_char_set<charT, traits>::list_iterator item_iterator; |
| 540 | |
| 541 | re_set* result = static_cast<re_set*>(append_state(syntax_element_set, sizeof(re_set))); |
| 542 | bool negate = char_set.is_negated(); |
| 543 | std::memset(result->_map, 0, sizeof(result->_map)); |
| 544 | // |
| 545 | // handle singles first: |
| 546 | // |
| 547 | item_iterator first, last; |
| 548 | first = char_set.singles_begin(); |
| 549 | last = char_set.singles_end(); |
| 550 | while(first != last) |
| 551 | { |
| 552 | for(unsigned int i = 0; i < (1 << CHAR_BIT); ++i) |
| 553 | { |
| 554 | if(this->m_traits.translate(static_cast<charT>(i), this->m_icase) |
| 555 | == this->m_traits.translate(first->first, this->m_icase)) |
| 556 | result->_map[i] = true; |
| 557 | } |
| 558 | ++first; |
| 559 | } |
| 560 | // |
| 561 | // OK now handle ranges: |
| 562 | // |
| 563 | first = char_set.ranges_begin(); |
| 564 | last = char_set.ranges_end(); |
| 565 | while(first != last) |
| 566 | { |
| 567 | // first grab the endpoints of the range: |
| 568 | charT c1 = this->m_traits.translate(first->first, this->m_icase); |
| 569 | ++first; |
| 570 | charT c2 = this->m_traits.translate(first->first, this->m_icase); |
| 571 | ++first; |
| 572 | // different actions now depending upon whether collation is turned on: |
| 573 | if(flags() & regex_constants::collate) |
| 574 | { |
| 575 | // we need to transform our range into sort keys: |
| 576 | charT c3[2] = { c1, charT(0), }; |
| 577 | string_type s1 = this->m_traits.transform(c3, c3+1); |
| 578 | c3[0] = c2; |
| 579 | string_type s2 = this->m_traits.transform(c3, c3+1); |
| 580 | if(s1 > s2) |
| 581 | { |
| 582 | // Oops error: |
| 583 | return 0; |
| 584 | } |
| 585 | NDNBOOST_ASSERT(c3[1] == charT(0)); |
| 586 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 587 | { |
| 588 | c3[0] = static_cast<charT>(i); |
| 589 | string_type s3 = this->m_traits.transform(c3, c3 +1); |
| 590 | if((s1 <= s3) && (s3 <= s2)) |
| 591 | result->_map[i] = true; |
| 592 | } |
| 593 | } |
| 594 | else |
| 595 | { |
| 596 | if(char_less(c2, c1)) |
| 597 | { |
| 598 | // Oops error: |
| 599 | return 0; |
| 600 | } |
| 601 | // everything in range matches: |
| 602 | std::memset(result->_map + static_cast<unsigned char>(c1), true, 1 + static_cast<unsigned char>(c2) - static_cast<unsigned char>(c1)); |
| 603 | } |
| 604 | } |
| 605 | // |
| 606 | // and now the classes: |
| 607 | // |
| 608 | typedef typename traits::char_class_type m_type; |
| 609 | m_type m = char_set.classes(); |
| 610 | if(flags() & regbase::icase) |
| 611 | { |
| 612 | // adjust m as needed: |
| 613 | if(((m & m_lower_mask) == m_lower_mask) || ((m & m_upper_mask) == m_upper_mask)) |
| 614 | m |= m_alpha_mask; |
| 615 | } |
| 616 | if(m != 0) |
| 617 | { |
| 618 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 619 | { |
| 620 | if(this->m_traits.isctype(static_cast<charT>(i), m)) |
| 621 | result->_map[i] = true; |
| 622 | } |
| 623 | } |
| 624 | // |
| 625 | // and now the negated classes: |
| 626 | // |
| 627 | m = char_set.negated_classes(); |
| 628 | if(flags() & regbase::icase) |
| 629 | { |
| 630 | // adjust m as needed: |
| 631 | if(((m & m_lower_mask) == m_lower_mask) || ((m & m_upper_mask) == m_upper_mask)) |
| 632 | m |= m_alpha_mask; |
| 633 | } |
| 634 | if(m != 0) |
| 635 | { |
| 636 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 637 | { |
| 638 | if(0 == this->m_traits.isctype(static_cast<charT>(i), m)) |
| 639 | result->_map[i] = true; |
| 640 | } |
| 641 | } |
| 642 | // |
| 643 | // now process the equivalence classes: |
| 644 | // |
| 645 | first = char_set.equivalents_begin(); |
| 646 | last = char_set.equivalents_end(); |
| 647 | while(first != last) |
| 648 | { |
| 649 | string_type s; |
| 650 | NDNBOOST_ASSERT(static_cast<charT>(0) == first->second); |
| 651 | s = m_traits.transform_primary(&first->first, &first->first+1); |
| 652 | if(s.empty()) |
| 653 | return 0; // invalid or unsupported equivalence class |
| 654 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 655 | { |
| 656 | charT c[2] = { (static_cast<charT>(i)), charT(0), }; |
| 657 | string_type s2 = this->m_traits.transform_primary(c, c+1); |
| 658 | if(s == s2) |
| 659 | result->_map[i] = true; |
| 660 | } |
| 661 | ++first; |
| 662 | } |
| 663 | if(negate) |
| 664 | { |
| 665 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 666 | { |
| 667 | result->_map[i] = !(result->_map[i]); |
| 668 | } |
| 669 | } |
| 670 | return result; |
| 671 | } |
| 672 | |
| 673 | template <class charT, class traits> |
| 674 | void basic_regex_creator<charT, traits>::finalize(const charT* p1, const charT* p2) |
| 675 | { |
| 676 | if(this->m_pdata->m_status) |
| 677 | return; |
| 678 | // we've added all the states we need, now finish things off. |
| 679 | // start by adding a terminating state: |
| 680 | append_state(syntax_element_match); |
| 681 | // extend storage to store original expression: |
| 682 | std::ptrdiff_t len = p2 - p1; |
| 683 | m_pdata->m_expression_len = len; |
| 684 | charT* ps = static_cast<charT*>(m_pdata->m_data.extend(sizeof(charT) * (1 + (p2 - p1)))); |
| 685 | m_pdata->m_expression = ps; |
| 686 | re_detail::copy(p1, p2, ps); |
| 687 | ps[p2 - p1] = 0; |
| 688 | // fill in our other data... |
| 689 | // successful parsing implies a zero status: |
| 690 | m_pdata->m_status = 0; |
| 691 | // get the first state of the machine: |
| 692 | m_pdata->m_first_state = static_cast<re_syntax_base*>(m_pdata->m_data.data()); |
| 693 | // fixup pointers in the machine: |
| 694 | fixup_pointers(m_pdata->m_first_state); |
| 695 | if(m_has_recursions) |
| 696 | { |
| 697 | m_pdata->m_has_recursions = true; |
| 698 | fixup_recursions(m_pdata->m_first_state); |
| 699 | if(this->m_pdata->m_status) |
| 700 | return; |
| 701 | } |
| 702 | else |
| 703 | m_pdata->m_has_recursions = false; |
| 704 | // create nested startmaps: |
| 705 | create_startmaps(m_pdata->m_first_state); |
| 706 | // create main startmap: |
| 707 | std::memset(m_pdata->m_startmap, 0, sizeof(m_pdata->m_startmap)); |
| 708 | m_pdata->m_can_be_null = 0; |
| 709 | |
| 710 | m_bad_repeats = 0; |
| 711 | if(m_has_recursions) |
| 712 | m_recursion_checks.assign(1 + m_pdata->m_mark_count, false); |
| 713 | create_startmap(m_pdata->m_first_state, m_pdata->m_startmap, &(m_pdata->m_can_be_null), mask_all); |
| 714 | // get the restart type: |
| 715 | m_pdata->m_restart_type = get_restart_type(m_pdata->m_first_state); |
| 716 | // optimise a leading repeat if there is one: |
| 717 | probe_leading_repeat(m_pdata->m_first_state); |
| 718 | } |
| 719 | |
| 720 | template <class charT, class traits> |
| 721 | void basic_regex_creator<charT, traits>::fixup_pointers(re_syntax_base* state) |
| 722 | { |
| 723 | while(state) |
| 724 | { |
| 725 | switch(state->type) |
| 726 | { |
| 727 | case syntax_element_recurse: |
| 728 | m_has_recursions = true; |
| 729 | if(state->next.i) |
| 730 | state->next.p = getaddress(state->next.i, state); |
| 731 | else |
| 732 | state->next.p = 0; |
| 733 | break; |
| 734 | case syntax_element_rep: |
| 735 | case syntax_element_dot_rep: |
| 736 | case syntax_element_char_rep: |
| 737 | case syntax_element_short_set_rep: |
| 738 | case syntax_element_long_set_rep: |
| 739 | // set the state_id of this repeat: |
| 740 | static_cast<re_repeat*>(state)->state_id = m_repeater_id++; |
| 741 | NDNBOOST_FALLTHROUGH; |
| 742 | case syntax_element_alt: |
| 743 | std::memset(static_cast<re_alt*>(state)->_map, 0, sizeof(static_cast<re_alt*>(state)->_map)); |
| 744 | static_cast<re_alt*>(state)->can_be_null = 0; |
| 745 | NDNBOOST_FALLTHROUGH; |
| 746 | case syntax_element_jump: |
| 747 | static_cast<re_jump*>(state)->alt.p = getaddress(static_cast<re_jump*>(state)->alt.i, state); |
| 748 | NDNBOOST_FALLTHROUGH; |
| 749 | default: |
| 750 | if(state->next.i) |
| 751 | state->next.p = getaddress(state->next.i, state); |
| 752 | else |
| 753 | state->next.p = 0; |
| 754 | } |
| 755 | state = state->next.p; |
| 756 | } |
| 757 | } |
| 758 | |
| 759 | template <class charT, class traits> |
| 760 | void basic_regex_creator<charT, traits>::fixup_recursions(re_syntax_base* state) |
| 761 | { |
| 762 | re_syntax_base* base = state; |
| 763 | while(state) |
| 764 | { |
| 765 | switch(state->type) |
| 766 | { |
| 767 | case syntax_element_assert_backref: |
| 768 | { |
| 769 | // just check that the index is valid: |
| 770 | int idx = static_cast<const re_brace*>(state)->index; |
| 771 | if(idx < 0) |
| 772 | { |
| 773 | idx = -idx-1; |
| 774 | if(idx >= 10000) |
| 775 | { |
| 776 | idx = m_pdata->get_id(idx); |
| 777 | if(idx <= 0) |
| 778 | { |
| 779 | // check of sub-expression that doesn't exist: |
| 780 | if(0 == this->m_pdata->m_status) // update the error code if not already set |
| 781 | this->m_pdata->m_status = ndnboost::regex_constants::error_bad_pattern; |
| 782 | // |
| 783 | // clear the expression, we should be empty: |
| 784 | // |
| 785 | this->m_pdata->m_expression = 0; |
| 786 | this->m_pdata->m_expression_len = 0; |
| 787 | // |
| 788 | // and throw if required: |
| 789 | // |
| 790 | if(0 == (this->flags() & regex_constants::no_except)) |
| 791 | { |
| 792 | std::string message = "Encountered a forward reference to a marked sub-expression that does not exist."; |
| 793 | ndnboost::regex_error e(message, ndnboost::regex_constants::error_bad_pattern, 0); |
| 794 | e.raise(); |
| 795 | } |
| 796 | } |
| 797 | } |
| 798 | } |
| 799 | } |
| 800 | break; |
| 801 | case syntax_element_recurse: |
| 802 | { |
| 803 | bool ok = false; |
| 804 | re_syntax_base* p = base; |
| 805 | std::ptrdiff_t idx = static_cast<re_jump*>(state)->alt.i; |
| 806 | if(idx > 10000) |
| 807 | { |
| 808 | // |
| 809 | // There may be more than one capture group with this hash, just do what Perl |
| 810 | // does and recurse to the leftmost: |
| 811 | // |
| 812 | idx = m_pdata->get_id(static_cast<int>(idx)); |
| 813 | } |
| 814 | while(p) |
| 815 | { |
| 816 | if((p->type == syntax_element_startmark) && (static_cast<re_brace*>(p)->index == idx)) |
| 817 | { |
| 818 | // |
| 819 | // We've found the target of the recursion, set the jump target: |
| 820 | // |
| 821 | static_cast<re_jump*>(state)->alt.p = p; |
| 822 | ok = true; |
| 823 | // |
| 824 | // Now scan the target for nested repeats: |
| 825 | // |
| 826 | p = p->next.p; |
| 827 | int next_rep_id = 0; |
| 828 | while(p) |
| 829 | { |
| 830 | switch(p->type) |
| 831 | { |
| 832 | case syntax_element_rep: |
| 833 | case syntax_element_dot_rep: |
| 834 | case syntax_element_char_rep: |
| 835 | case syntax_element_short_set_rep: |
| 836 | case syntax_element_long_set_rep: |
| 837 | next_rep_id = static_cast<re_repeat*>(p)->state_id; |
| 838 | break; |
| 839 | case syntax_element_endmark: |
| 840 | if(static_cast<const re_brace*>(p)->index == idx) |
| 841 | next_rep_id = -1; |
| 842 | break; |
| 843 | default: |
| 844 | break; |
| 845 | } |
| 846 | if(next_rep_id) |
| 847 | break; |
| 848 | p = p->next.p; |
| 849 | } |
| 850 | if(next_rep_id > 0) |
| 851 | { |
| 852 | static_cast<re_recurse*>(state)->state_id = next_rep_id - 1; |
| 853 | } |
| 854 | |
| 855 | break; |
| 856 | } |
| 857 | p = p->next.p; |
| 858 | } |
| 859 | if(!ok) |
| 860 | { |
| 861 | // recursion to sub-expression that doesn't exist: |
| 862 | if(0 == this->m_pdata->m_status) // update the error code if not already set |
| 863 | this->m_pdata->m_status = ndnboost::regex_constants::error_bad_pattern; |
| 864 | // |
| 865 | // clear the expression, we should be empty: |
| 866 | // |
| 867 | this->m_pdata->m_expression = 0; |
| 868 | this->m_pdata->m_expression_len = 0; |
| 869 | // |
| 870 | // and throw if required: |
| 871 | // |
| 872 | if(0 == (this->flags() & regex_constants::no_except)) |
| 873 | { |
| 874 | std::string message = "Encountered a forward reference to a recursive sub-expression that does not exist."; |
| 875 | ndnboost::regex_error e(message, ndnboost::regex_constants::error_bad_pattern, 0); |
| 876 | e.raise(); |
| 877 | } |
| 878 | } |
| 879 | } |
| 880 | break; |
| 881 | default: |
| 882 | break; |
| 883 | } |
| 884 | state = state->next.p; |
| 885 | } |
| 886 | } |
| 887 | |
| 888 | template <class charT, class traits> |
| 889 | void basic_regex_creator<charT, traits>::create_startmaps(re_syntax_base* state) |
| 890 | { |
| 891 | // non-recursive implementation: |
| 892 | // create the last map in the machine first, so that earlier maps |
| 893 | // can make use of the result... |
| 894 | // |
| 895 | // This was originally a recursive implementation, but that caused stack |
| 896 | // overflows with complex expressions on small stacks (think COM+). |
| 897 | |
| 898 | // start by saving the case setting: |
| 899 | bool l_icase = m_icase; |
| 900 | std::vector<std::pair<bool, re_syntax_base*> > v; |
| 901 | |
| 902 | while(state) |
| 903 | { |
| 904 | switch(state->type) |
| 905 | { |
| 906 | case syntax_element_toggle_case: |
| 907 | // we need to track case changes here: |
| 908 | m_icase = static_cast<re_case*>(state)->icase; |
| 909 | state = state->next.p; |
| 910 | continue; |
| 911 | case syntax_element_alt: |
| 912 | case syntax_element_rep: |
| 913 | case syntax_element_dot_rep: |
| 914 | case syntax_element_char_rep: |
| 915 | case syntax_element_short_set_rep: |
| 916 | case syntax_element_long_set_rep: |
| 917 | // just push the state onto our stack for now: |
| 918 | v.push_back(std::pair<bool, re_syntax_base*>(m_icase, state)); |
| 919 | state = state->next.p; |
| 920 | break; |
| 921 | case syntax_element_backstep: |
| 922 | // we need to calculate how big the backstep is: |
| 923 | static_cast<re_brace*>(state)->index |
| 924 | = this->calculate_backstep(state->next.p); |
| 925 | if(static_cast<re_brace*>(state)->index < 0) |
| 926 | { |
| 927 | // Oops error: |
| 928 | if(0 == this->m_pdata->m_status) // update the error code if not already set |
| 929 | this->m_pdata->m_status = ndnboost::regex_constants::error_bad_pattern; |
| 930 | // |
| 931 | // clear the expression, we should be empty: |
| 932 | // |
| 933 | this->m_pdata->m_expression = 0; |
| 934 | this->m_pdata->m_expression_len = 0; |
| 935 | // |
| 936 | // and throw if required: |
| 937 | // |
| 938 | if(0 == (this->flags() & regex_constants::no_except)) |
| 939 | { |
| 940 | std::string message = "Invalid lookbehind assertion encountered in the regular expression."; |
| 941 | ndnboost::regex_error e(message, ndnboost::regex_constants::error_bad_pattern, 0); |
| 942 | e.raise(); |
| 943 | } |
| 944 | } |
| 945 | NDNBOOST_FALLTHROUGH; |
| 946 | default: |
| 947 | state = state->next.p; |
| 948 | } |
| 949 | } |
| 950 | |
| 951 | // now work through our list, building all the maps as we go: |
| 952 | while(v.size()) |
| 953 | { |
| 954 | // Initialize m_recursion_checks if we need it: |
| 955 | if(m_has_recursions) |
| 956 | m_recursion_checks.assign(1 + m_pdata->m_mark_count, false); |
| 957 | |
| 958 | const std::pair<bool, re_syntax_base*>& p = v.back(); |
| 959 | m_icase = p.first; |
| 960 | state = p.second; |
| 961 | v.pop_back(); |
| 962 | |
| 963 | // Build maps: |
| 964 | m_bad_repeats = 0; |
| 965 | create_startmap(state->next.p, static_cast<re_alt*>(state)->_map, &static_cast<re_alt*>(state)->can_be_null, mask_take); |
| 966 | m_bad_repeats = 0; |
| 967 | |
| 968 | if(m_has_recursions) |
| 969 | m_recursion_checks.assign(1 + m_pdata->m_mark_count, false); |
| 970 | create_startmap(static_cast<re_alt*>(state)->alt.p, static_cast<re_alt*>(state)->_map, &static_cast<re_alt*>(state)->can_be_null, mask_skip); |
| 971 | // adjust the type of the state to allow for faster matching: |
| 972 | state->type = this->get_repeat_type(state); |
| 973 | } |
| 974 | // restore case sensitivity: |
| 975 | m_icase = l_icase; |
| 976 | } |
| 977 | |
| 978 | template <class charT, class traits> |
| 979 | int basic_regex_creator<charT, traits>::calculate_backstep(re_syntax_base* state) |
| 980 | { |
| 981 | typedef typename traits::char_class_type m_type; |
| 982 | int result = 0; |
| 983 | while(state) |
| 984 | { |
| 985 | switch(state->type) |
| 986 | { |
| 987 | case syntax_element_startmark: |
| 988 | if((static_cast<re_brace*>(state)->index == -1) |
| 989 | || (static_cast<re_brace*>(state)->index == -2)) |
| 990 | { |
| 991 | state = static_cast<re_jump*>(state->next.p)->alt.p->next.p; |
| 992 | continue; |
| 993 | } |
| 994 | else if(static_cast<re_brace*>(state)->index == -3) |
| 995 | { |
| 996 | state = state->next.p->next.p; |
| 997 | continue; |
| 998 | } |
| 999 | break; |
| 1000 | case syntax_element_endmark: |
| 1001 | if((static_cast<re_brace*>(state)->index == -1) |
| 1002 | || (static_cast<re_brace*>(state)->index == -2)) |
| 1003 | return result; |
| 1004 | break; |
| 1005 | case syntax_element_literal: |
| 1006 | result += static_cast<re_literal*>(state)->length; |
| 1007 | break; |
| 1008 | case syntax_element_wild: |
| 1009 | case syntax_element_set: |
| 1010 | result += 1; |
| 1011 | break; |
| 1012 | case syntax_element_dot_rep: |
| 1013 | case syntax_element_char_rep: |
| 1014 | case syntax_element_short_set_rep: |
| 1015 | case syntax_element_backref: |
| 1016 | case syntax_element_rep: |
| 1017 | case syntax_element_combining: |
| 1018 | case syntax_element_long_set_rep: |
| 1019 | case syntax_element_backstep: |
| 1020 | { |
| 1021 | re_repeat* rep = static_cast<re_repeat *>(state); |
| 1022 | // adjust the type of the state to allow for faster matching: |
| 1023 | state->type = this->get_repeat_type(state); |
| 1024 | if((state->type == syntax_element_dot_rep) |
| 1025 | || (state->type == syntax_element_char_rep) |
| 1026 | || (state->type == syntax_element_short_set_rep)) |
| 1027 | { |
| 1028 | if(rep->max != rep->min) |
| 1029 | return -1; |
| 1030 | result += static_cast<int>(rep->min); |
| 1031 | state = rep->alt.p; |
| 1032 | continue; |
| 1033 | } |
| 1034 | else if(state->type == syntax_element_long_set_rep) |
| 1035 | { |
| 1036 | NDNBOOST_ASSERT(rep->next.p->type == syntax_element_long_set); |
| 1037 | if(static_cast<re_set_long<m_type>*>(rep->next.p)->singleton == 0) |
| 1038 | return -1; |
| 1039 | if(rep->max != rep->min) |
| 1040 | return -1; |
| 1041 | result += static_cast<int>(rep->min); |
| 1042 | state = rep->alt.p; |
| 1043 | continue; |
| 1044 | } |
| 1045 | } |
| 1046 | return -1; |
| 1047 | case syntax_element_long_set: |
| 1048 | if(static_cast<re_set_long<m_type>*>(state)->singleton == 0) |
| 1049 | return -1; |
| 1050 | result += 1; |
| 1051 | break; |
| 1052 | case syntax_element_jump: |
| 1053 | state = static_cast<re_jump*>(state)->alt.p; |
| 1054 | continue; |
| 1055 | case syntax_element_alt: |
| 1056 | { |
| 1057 | int r1 = calculate_backstep(state->next.p); |
| 1058 | int r2 = calculate_backstep(static_cast<re_alt*>(state)->alt.p); |
| 1059 | if((r1 < 0) || (r1 != r2)) |
| 1060 | return -1; |
| 1061 | return result + r1; |
| 1062 | } |
| 1063 | default: |
| 1064 | break; |
| 1065 | } |
| 1066 | state = state->next.p; |
| 1067 | } |
| 1068 | return -1; |
| 1069 | } |
| 1070 | |
| 1071 | template <class charT, class traits> |
| 1072 | void basic_regex_creator<charT, traits>::create_startmap(re_syntax_base* state, unsigned char* l_map, unsigned int* pnull, unsigned char mask) |
| 1073 | { |
| 1074 | int not_last_jump = 1; |
| 1075 | re_syntax_base* recursion_start = 0; |
| 1076 | int recursion_sub = 0; |
| 1077 | re_syntax_base* recursion_restart = 0; |
| 1078 | |
| 1079 | // track case sensitivity: |
| 1080 | bool l_icase = m_icase; |
| 1081 | |
| 1082 | while(state) |
| 1083 | { |
| 1084 | switch(state->type) |
| 1085 | { |
| 1086 | case syntax_element_toggle_case: |
| 1087 | l_icase = static_cast<re_case*>(state)->icase; |
| 1088 | state = state->next.p; |
| 1089 | break; |
| 1090 | case syntax_element_literal: |
| 1091 | { |
| 1092 | // don't set anything in *pnull, set each element in l_map |
| 1093 | // that could match the first character in the literal: |
| 1094 | if(l_map) |
| 1095 | { |
| 1096 | l_map[0] |= mask_init; |
| 1097 | charT first_char = *static_cast<charT*>(static_cast<void*>(static_cast<re_literal*>(state) + 1)); |
| 1098 | for(unsigned int i = 0; i < (1u << CHAR_BIT); ++i) |
| 1099 | { |
| 1100 | if(m_traits.translate(static_cast<charT>(i), l_icase) == first_char) |
| 1101 | l_map[i] |= mask; |
| 1102 | } |
| 1103 | } |
| 1104 | return; |
| 1105 | } |
| 1106 | case syntax_element_end_line: |
| 1107 | { |
| 1108 | // next character must be a line separator (if there is one): |
| 1109 | if(l_map) |
| 1110 | { |
| 1111 | l_map[0] |= mask_init; |
| 1112 | l_map[static_cast<unsigned>('\n')] |= mask; |
| 1113 | l_map[static_cast<unsigned>('\r')] |= mask; |
| 1114 | l_map[static_cast<unsigned>('\f')] |= mask; |
| 1115 | l_map[0x85] |= mask; |
| 1116 | } |
| 1117 | // now figure out if we can match a NULL string at this point: |
| 1118 | if(pnull) |
| 1119 | create_startmap(state->next.p, 0, pnull, mask); |
| 1120 | return; |
| 1121 | } |
| 1122 | case syntax_element_recurse: |
| 1123 | { |
| 1124 | if(state->type == syntax_element_startmark) |
| 1125 | recursion_sub = static_cast<re_brace*>(state)->index; |
| 1126 | else |
| 1127 | recursion_sub = 0; |
| 1128 | if(m_recursion_checks[recursion_sub]) |
| 1129 | { |
| 1130 | // Infinite recursion!! |
| 1131 | if(0 == this->m_pdata->m_status) // update the error code if not already set |
| 1132 | this->m_pdata->m_status = ndnboost::regex_constants::error_bad_pattern; |
| 1133 | // |
| 1134 | // clear the expression, we should be empty: |
| 1135 | // |
| 1136 | this->m_pdata->m_expression = 0; |
| 1137 | this->m_pdata->m_expression_len = 0; |
| 1138 | // |
| 1139 | // and throw if required: |
| 1140 | // |
| 1141 | if(0 == (this->flags() & regex_constants::no_except)) |
| 1142 | { |
| 1143 | std::string message = "Encountered an infinite recursion."; |
| 1144 | ndnboost::regex_error e(message, ndnboost::regex_constants::error_bad_pattern, 0); |
| 1145 | e.raise(); |
| 1146 | } |
| 1147 | } |
| 1148 | else if(recursion_start == 0) |
| 1149 | { |
| 1150 | recursion_start = state; |
| 1151 | recursion_restart = state->next.p; |
| 1152 | state = static_cast<re_jump*>(state)->alt.p; |
| 1153 | m_recursion_checks[recursion_sub] = true; |
| 1154 | break; |
| 1155 | } |
| 1156 | m_recursion_checks[recursion_sub] = true; |
| 1157 | // can't handle nested recursion here... |
| 1158 | NDNBOOST_FALLTHROUGH; |
| 1159 | } |
| 1160 | case syntax_element_backref: |
| 1161 | // can be null, and any character can match: |
| 1162 | if(pnull) |
| 1163 | *pnull |= mask; |
| 1164 | NDNBOOST_FALLTHROUGH; |
| 1165 | case syntax_element_wild: |
| 1166 | { |
| 1167 | // can't be null, any character can match: |
| 1168 | set_all_masks(l_map, mask); |
| 1169 | return; |
| 1170 | } |
| 1171 | case syntax_element_match: |
| 1172 | { |
| 1173 | // must be null, any character can match: |
| 1174 | set_all_masks(l_map, mask); |
| 1175 | if(pnull) |
| 1176 | *pnull |= mask; |
| 1177 | return; |
| 1178 | } |
| 1179 | case syntax_element_word_start: |
| 1180 | { |
| 1181 | // recurse, then AND with all the word characters: |
| 1182 | create_startmap(state->next.p, l_map, pnull, mask); |
| 1183 | if(l_map) |
| 1184 | { |
| 1185 | l_map[0] |= mask_init; |
| 1186 | for(unsigned int i = 0; i < (1u << CHAR_BIT); ++i) |
| 1187 | { |
| 1188 | if(!m_traits.isctype(static_cast<charT>(i), m_word_mask)) |
| 1189 | l_map[i] &= static_cast<unsigned char>(~mask); |
| 1190 | } |
| 1191 | } |
| 1192 | return; |
| 1193 | } |
| 1194 | case syntax_element_word_end: |
| 1195 | { |
| 1196 | // recurse, then AND with all the word characters: |
| 1197 | create_startmap(state->next.p, l_map, pnull, mask); |
| 1198 | if(l_map) |
| 1199 | { |
| 1200 | l_map[0] |= mask_init; |
| 1201 | for(unsigned int i = 0; i < (1u << CHAR_BIT); ++i) |
| 1202 | { |
| 1203 | if(m_traits.isctype(static_cast<charT>(i), m_word_mask)) |
| 1204 | l_map[i] &= static_cast<unsigned char>(~mask); |
| 1205 | } |
| 1206 | } |
| 1207 | return; |
| 1208 | } |
| 1209 | case syntax_element_buffer_end: |
| 1210 | { |
| 1211 | // we *must be null* : |
| 1212 | if(pnull) |
| 1213 | *pnull |= mask; |
| 1214 | return; |
| 1215 | } |
| 1216 | case syntax_element_long_set: |
| 1217 | if(l_map) |
| 1218 | { |
| 1219 | typedef typename traits::char_class_type m_type; |
| 1220 | if(static_cast<re_set_long<m_type>*>(state)->singleton) |
| 1221 | { |
| 1222 | l_map[0] |= mask_init; |
| 1223 | for(unsigned int i = 0; i < (1u << CHAR_BIT); ++i) |
| 1224 | { |
| 1225 | charT c = static_cast<charT>(i); |
| 1226 | if(&c != re_is_set_member(&c, &c + 1, static_cast<re_set_long<m_type>*>(state), *m_pdata, l_icase)) |
| 1227 | l_map[i] |= mask; |
| 1228 | } |
| 1229 | } |
| 1230 | else |
| 1231 | set_all_masks(l_map, mask); |
| 1232 | } |
| 1233 | return; |
| 1234 | case syntax_element_set: |
| 1235 | if(l_map) |
| 1236 | { |
| 1237 | l_map[0] |= mask_init; |
| 1238 | for(unsigned int i = 0; i < (1u << CHAR_BIT); ++i) |
| 1239 | { |
| 1240 | if(static_cast<re_set*>(state)->_map[ |
| 1241 | static_cast<unsigned char>(m_traits.translate(static_cast<charT>(i), l_icase))]) |
| 1242 | l_map[i] |= mask; |
| 1243 | } |
| 1244 | } |
| 1245 | return; |
| 1246 | case syntax_element_jump: |
| 1247 | // take the jump: |
| 1248 | state = static_cast<re_alt*>(state)->alt.p; |
| 1249 | not_last_jump = -1; |
| 1250 | break; |
| 1251 | case syntax_element_alt: |
| 1252 | case syntax_element_rep: |
| 1253 | case syntax_element_dot_rep: |
| 1254 | case syntax_element_char_rep: |
| 1255 | case syntax_element_short_set_rep: |
| 1256 | case syntax_element_long_set_rep: |
| 1257 | { |
| 1258 | re_alt* rep = static_cast<re_alt*>(state); |
| 1259 | if(rep->_map[0] & mask_init) |
| 1260 | { |
| 1261 | if(l_map) |
| 1262 | { |
| 1263 | // copy previous results: |
| 1264 | l_map[0] |= mask_init; |
| 1265 | for(unsigned int i = 0; i <= UCHAR_MAX; ++i) |
| 1266 | { |
| 1267 | if(rep->_map[i] & mask_any) |
| 1268 | l_map[i] |= mask; |
| 1269 | } |
| 1270 | } |
| 1271 | if(pnull) |
| 1272 | { |
| 1273 | if(rep->can_be_null & mask_any) |
| 1274 | *pnull |= mask; |
| 1275 | } |
| 1276 | } |
| 1277 | else |
| 1278 | { |
| 1279 | // we haven't created a startmap for this alternative yet |
| 1280 | // so take the union of the two options: |
| 1281 | if(is_bad_repeat(state)) |
| 1282 | { |
| 1283 | set_all_masks(l_map, mask); |
| 1284 | if(pnull) |
| 1285 | *pnull |= mask; |
| 1286 | return; |
| 1287 | } |
| 1288 | set_bad_repeat(state); |
| 1289 | create_startmap(state->next.p, l_map, pnull, mask); |
| 1290 | if((state->type == syntax_element_alt) |
| 1291 | || (static_cast<re_repeat*>(state)->min == 0) |
| 1292 | || (not_last_jump == 0)) |
| 1293 | create_startmap(rep->alt.p, l_map, pnull, mask); |
| 1294 | } |
| 1295 | } |
| 1296 | return; |
| 1297 | case syntax_element_soft_buffer_end: |
| 1298 | // match newline or null: |
| 1299 | if(l_map) |
| 1300 | { |
| 1301 | l_map[0] |= mask_init; |
| 1302 | l_map[static_cast<unsigned>('\n')] |= mask; |
| 1303 | l_map[static_cast<unsigned>('\r')] |= mask; |
| 1304 | } |
| 1305 | if(pnull) |
| 1306 | *pnull |= mask; |
| 1307 | return; |
| 1308 | case syntax_element_endmark: |
| 1309 | // need to handle independent subs as a special case: |
| 1310 | if(static_cast<re_brace*>(state)->index < 0) |
| 1311 | { |
| 1312 | // can be null, any character can match: |
| 1313 | set_all_masks(l_map, mask); |
| 1314 | if(pnull) |
| 1315 | *pnull |= mask; |
| 1316 | return; |
| 1317 | } |
| 1318 | else if(recursion_start && (recursion_sub != 0) && (recursion_sub == static_cast<re_brace*>(state)->index)) |
| 1319 | { |
| 1320 | // recursion termination: |
| 1321 | recursion_start = 0; |
| 1322 | state = recursion_restart; |
| 1323 | break; |
| 1324 | } |
| 1325 | |
| 1326 | // |
| 1327 | // Normally we just go to the next state... but if this sub-expression is |
| 1328 | // the target of a recursion, then we might be ending a recursion, in which |
| 1329 | // case we should check whatever follows that recursion, as well as whatever |
| 1330 | // follows this state: |
| 1331 | // |
| 1332 | if(m_pdata->m_has_recursions && static_cast<re_brace*>(state)->index) |
| 1333 | { |
| 1334 | bool ok = false; |
| 1335 | re_syntax_base* p = m_pdata->m_first_state; |
| 1336 | while(p) |
| 1337 | { |
| 1338 | if(p->type == syntax_element_recurse) |
| 1339 | { |
| 1340 | re_brace* p2 = static_cast<re_brace*>(static_cast<re_jump*>(p)->alt.p); |
| 1341 | if((p2->type == syntax_element_startmark) && (p2->index == static_cast<re_brace*>(state)->index)) |
| 1342 | { |
| 1343 | ok = true; |
| 1344 | break; |
| 1345 | } |
| 1346 | } |
| 1347 | p = p->next.p; |
| 1348 | } |
| 1349 | if(ok) |
| 1350 | { |
| 1351 | create_startmap(p->next.p, l_map, pnull, mask); |
| 1352 | } |
| 1353 | } |
| 1354 | state = state->next.p; |
| 1355 | break; |
| 1356 | |
| 1357 | case syntax_element_startmark: |
| 1358 | // need to handle independent subs as a special case: |
| 1359 | if(static_cast<re_brace*>(state)->index == -3) |
| 1360 | { |
| 1361 | state = state->next.p->next.p; |
| 1362 | break; |
| 1363 | } |
| 1364 | NDNBOOST_FALLTHROUGH; |
| 1365 | default: |
| 1366 | state = state->next.p; |
| 1367 | } |
| 1368 | ++not_last_jump; |
| 1369 | } |
| 1370 | } |
| 1371 | |
| 1372 | template <class charT, class traits> |
| 1373 | unsigned basic_regex_creator<charT, traits>::get_restart_type(re_syntax_base* state) |
| 1374 | { |
| 1375 | // |
| 1376 | // find out how the machine starts, so we can optimise the search: |
| 1377 | // |
| 1378 | while(state) |
| 1379 | { |
| 1380 | switch(state->type) |
| 1381 | { |
| 1382 | case syntax_element_startmark: |
| 1383 | case syntax_element_endmark: |
| 1384 | state = state->next.p; |
| 1385 | continue; |
| 1386 | case syntax_element_start_line: |
| 1387 | return regbase::restart_line; |
| 1388 | case syntax_element_word_start: |
| 1389 | return regbase::restart_word; |
| 1390 | case syntax_element_buffer_start: |
| 1391 | return regbase::restart_buf; |
| 1392 | case syntax_element_restart_continue: |
| 1393 | return regbase::restart_continue; |
| 1394 | default: |
| 1395 | state = 0; |
| 1396 | continue; |
| 1397 | } |
| 1398 | } |
| 1399 | return regbase::restart_any; |
| 1400 | } |
| 1401 | |
| 1402 | template <class charT, class traits> |
| 1403 | void basic_regex_creator<charT, traits>::set_all_masks(unsigned char* bits, unsigned char mask) |
| 1404 | { |
| 1405 | // |
| 1406 | // set mask in all of bits elements, |
| 1407 | // if bits[0] has mask_init not set then we can |
| 1408 | // optimise this to a call to memset: |
| 1409 | // |
| 1410 | if(bits) |
| 1411 | { |
| 1412 | if(bits[0] == 0) |
| 1413 | (std::memset)(bits, mask, 1u << CHAR_BIT); |
| 1414 | else |
| 1415 | { |
| 1416 | for(unsigned i = 0; i < (1u << CHAR_BIT); ++i) |
| 1417 | bits[i] |= mask; |
| 1418 | } |
| 1419 | bits[0] |= mask_init; |
| 1420 | } |
| 1421 | } |
| 1422 | |
| 1423 | template <class charT, class traits> |
| 1424 | bool basic_regex_creator<charT, traits>::is_bad_repeat(re_syntax_base* pt) |
| 1425 | { |
| 1426 | switch(pt->type) |
| 1427 | { |
| 1428 | case syntax_element_rep: |
| 1429 | case syntax_element_dot_rep: |
| 1430 | case syntax_element_char_rep: |
| 1431 | case syntax_element_short_set_rep: |
| 1432 | case syntax_element_long_set_rep: |
| 1433 | { |
| 1434 | unsigned state_id = static_cast<re_repeat*>(pt)->state_id; |
| 1435 | if(state_id > sizeof(m_bad_repeats) * CHAR_BIT) |
| 1436 | return true; // run out of bits, assume we can't traverse this one. |
| 1437 | static const ndnboost::uintmax_t one = 1uL; |
| 1438 | return m_bad_repeats & (one << state_id); |
| 1439 | } |
| 1440 | default: |
| 1441 | return false; |
| 1442 | } |
| 1443 | } |
| 1444 | |
| 1445 | template <class charT, class traits> |
| 1446 | void basic_regex_creator<charT, traits>::set_bad_repeat(re_syntax_base* pt) |
| 1447 | { |
| 1448 | switch(pt->type) |
| 1449 | { |
| 1450 | case syntax_element_rep: |
| 1451 | case syntax_element_dot_rep: |
| 1452 | case syntax_element_char_rep: |
| 1453 | case syntax_element_short_set_rep: |
| 1454 | case syntax_element_long_set_rep: |
| 1455 | { |
| 1456 | unsigned state_id = static_cast<re_repeat*>(pt)->state_id; |
| 1457 | static const ndnboost::uintmax_t one = 1uL; |
| 1458 | if(state_id <= sizeof(m_bad_repeats) * CHAR_BIT) |
| 1459 | m_bad_repeats |= (one << state_id); |
| 1460 | } |
| 1461 | break; |
| 1462 | default: |
| 1463 | break; |
| 1464 | } |
| 1465 | } |
| 1466 | |
| 1467 | template <class charT, class traits> |
| 1468 | syntax_element_type basic_regex_creator<charT, traits>::get_repeat_type(re_syntax_base* state) |
| 1469 | { |
| 1470 | typedef typename traits::char_class_type m_type; |
| 1471 | if(state->type == syntax_element_rep) |
| 1472 | { |
| 1473 | // check to see if we are repeating a single state: |
| 1474 | if(state->next.p->next.p->next.p == static_cast<re_alt*>(state)->alt.p) |
| 1475 | { |
| 1476 | switch(state->next.p->type) |
| 1477 | { |
| 1478 | case re_detail::syntax_element_wild: |
| 1479 | return re_detail::syntax_element_dot_rep; |
| 1480 | case re_detail::syntax_element_literal: |
| 1481 | return re_detail::syntax_element_char_rep; |
| 1482 | case re_detail::syntax_element_set: |
| 1483 | return re_detail::syntax_element_short_set_rep; |
| 1484 | case re_detail::syntax_element_long_set: |
| 1485 | if(static_cast<re_detail::re_set_long<m_type>*>(state->next.p)->singleton) |
| 1486 | return re_detail::syntax_element_long_set_rep; |
| 1487 | break; |
| 1488 | default: |
| 1489 | break; |
| 1490 | } |
| 1491 | } |
| 1492 | } |
| 1493 | return state->type; |
| 1494 | } |
| 1495 | |
| 1496 | template <class charT, class traits> |
| 1497 | void basic_regex_creator<charT, traits>::probe_leading_repeat(re_syntax_base* state) |
| 1498 | { |
| 1499 | // enumerate our states, and see if we have a leading repeat |
| 1500 | // for which failed search restarts can be optimised; |
| 1501 | do |
| 1502 | { |
| 1503 | switch(state->type) |
| 1504 | { |
| 1505 | case syntax_element_startmark: |
| 1506 | if(static_cast<re_brace*>(state)->index >= 0) |
| 1507 | { |
| 1508 | state = state->next.p; |
| 1509 | continue; |
| 1510 | } |
| 1511 | if((static_cast<re_brace*>(state)->index == -1) |
| 1512 | || (static_cast<re_brace*>(state)->index == -2)) |
| 1513 | { |
| 1514 | // skip past the zero width assertion: |
| 1515 | state = static_cast<const re_jump*>(state->next.p)->alt.p->next.p; |
| 1516 | continue; |
| 1517 | } |
| 1518 | if(static_cast<re_brace*>(state)->index == -3) |
| 1519 | { |
| 1520 | // Have to skip the leading jump state: |
| 1521 | state = state->next.p->next.p; |
| 1522 | continue; |
| 1523 | } |
| 1524 | return; |
| 1525 | case syntax_element_endmark: |
| 1526 | case syntax_element_start_line: |
| 1527 | case syntax_element_end_line: |
| 1528 | case syntax_element_word_boundary: |
| 1529 | case syntax_element_within_word: |
| 1530 | case syntax_element_word_start: |
| 1531 | case syntax_element_word_end: |
| 1532 | case syntax_element_buffer_start: |
| 1533 | case syntax_element_buffer_end: |
| 1534 | case syntax_element_restart_continue: |
| 1535 | state = state->next.p; |
| 1536 | break; |
| 1537 | case syntax_element_dot_rep: |
| 1538 | case syntax_element_char_rep: |
| 1539 | case syntax_element_short_set_rep: |
| 1540 | case syntax_element_long_set_rep: |
| 1541 | if(this->m_has_backrefs == 0) |
| 1542 | static_cast<re_repeat*>(state)->leading = true; |
| 1543 | NDNBOOST_FALLTHROUGH; |
| 1544 | default: |
| 1545 | return; |
| 1546 | } |
| 1547 | }while(state); |
| 1548 | } |
| 1549 | |
| 1550 | |
| 1551 | } // namespace re_detail |
| 1552 | |
| 1553 | } // namespace ndnboost |
| 1554 | |
| 1555 | #ifdef NDNBOOST_MSVC |
| 1556 | # pragma warning(pop) |
| 1557 | #endif |
| 1558 | |
| 1559 | #ifdef NDNBOOST_MSVC |
| 1560 | #pragma warning(push) |
| 1561 | #pragma warning(disable: 4103) |
| 1562 | #endif |
| 1563 | #ifdef NDNBOOST_HAS_ABI_HEADERS |
| 1564 | # include NDNBOOST_ABI_SUFFIX |
| 1565 | #endif |
| 1566 | #ifdef NDNBOOST_MSVC |
| 1567 | #pragma warning(pop) |
| 1568 | #endif |
| 1569 | |
| 1570 | #endif |
| 1571 | |