| 1 | /* |
| 2 | * Copyright © 2018 Adobe Inc. |
| 3 | * |
| 4 | * This is part of HarfBuzz, a text shaping library. |
| 5 | * |
| 6 | * Permission is hereby granted, without written agreement and without |
| 7 | * license or royalty fees, to use, copy, modify, and distribute this |
| 8 | * software and its documentation for any purpose, provided that the |
| 9 | * above copyright notice and the following two paragraphs appear in |
| 10 | * all copies of this software. |
| 11 | * |
| 12 | * IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR |
| 13 | * DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES |
| 14 | * ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN |
| 15 | * IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH |
| 16 | * DAMAGE. |
| 17 | * |
| 18 | * THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING, |
| 19 | * BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND |
| 20 | * FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS |
| 21 | * ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO |
| 22 | * PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS. |
| 23 | * |
| 24 | * Adobe Author(s): Michiharu Ariza |
| 25 | */ |
| 26 | |
| 27 | #ifndef HB_SUBSET_CFF_COMMON_HH |
| 28 | #define HB_SUBSET_CFF_COMMON_HH |
| 29 | |
| 30 | #include "hb.hh" |
| 31 | |
| 32 | #include "hb-subset-plan.hh" |
| 33 | #include "hb-cff-interp-cs-common.hh" |
| 34 | |
| 35 | namespace CFF { |
| 36 | |
| 37 | /* Used for writing a temporary charstring */ |
| 38 | struct str_encoder_t |
| 39 | { |
| 40 | str_encoder_t (str_buff_t &buff_) |
| 41 | : buff (buff_), error (false) {} |
| 42 | |
| 43 | void reset () { buff.resize (0); } |
| 44 | |
| 45 | void encode_byte (unsigned char b) |
| 46 | { |
| 47 | if (unlikely (buff.push (b) == &Crap (unsigned char))) |
| 48 | set_error (); |
| 49 | } |
| 50 | |
| 51 | void encode_int (int v) |
| 52 | { |
| 53 | if ((-1131 <= v) && (v <= 1131)) |
| 54 | { |
| 55 | if ((-107 <= v) && (v <= 107)) |
| 56 | encode_byte (v + 139); |
| 57 | else if (v > 0) |
| 58 | { |
| 59 | v -= 108; |
| 60 | encode_byte ((v >> 8) + OpCode_TwoBytePosInt0); |
| 61 | encode_byte (v & 0xFF); |
| 62 | } |
| 63 | else |
| 64 | { |
| 65 | v = -v - 108; |
| 66 | encode_byte ((v >> 8) + OpCode_TwoByteNegInt0); |
| 67 | encode_byte (v & 0xFF); |
| 68 | } |
| 69 | } |
| 70 | else |
| 71 | { |
| 72 | if (unlikely (v < -32768)) |
| 73 | v = -32768; |
| 74 | else if (unlikely (v > 32767)) |
| 75 | v = 32767; |
| 76 | encode_byte (OpCode_shortint); |
| 77 | encode_byte ((v >> 8) & 0xFF); |
| 78 | encode_byte (v & 0xFF); |
| 79 | } |
| 80 | } |
| 81 | |
| 82 | void encode_num (const number_t& n) |
| 83 | { |
| 84 | if (n.in_int_range ()) |
| 85 | { |
| 86 | encode_int (n.to_int ()); |
| 87 | } |
| 88 | else |
| 89 | { |
| 90 | int32_t v = n.to_fixed (); |
| 91 | encode_byte (OpCode_fixedcs); |
| 92 | encode_byte ((v >> 24) & 0xFF); |
| 93 | encode_byte ((v >> 16) & 0xFF); |
| 94 | encode_byte ((v >> 8) & 0xFF); |
| 95 | encode_byte (v & 0xFF); |
| 96 | } |
| 97 | } |
| 98 | |
| 99 | void encode_op (op_code_t op) |
| 100 | { |
| 101 | if (Is_OpCode_ESC (op)) |
| 102 | { |
| 103 | encode_byte (OpCode_escape); |
| 104 | encode_byte (Unmake_OpCode_ESC (op)); |
| 105 | } |
| 106 | else |
| 107 | encode_byte (op); |
| 108 | } |
| 109 | |
| 110 | void copy_str (const byte_str_t &str) |
| 111 | { |
| 112 | unsigned int offset = buff.length; |
| 113 | buff.resize (offset + str.length); |
| 114 | if (unlikely (buff.length < offset + str.length)) |
| 115 | { |
| 116 | set_error (); |
| 117 | return; |
| 118 | } |
| 119 | memcpy (&buff[offset], &str[0], str.length); |
| 120 | } |
| 121 | |
| 122 | bool is_error () const { return error; } |
| 123 | |
| 124 | protected: |
| 125 | void set_error () { error = true; } |
| 126 | |
| 127 | str_buff_t &buff; |
| 128 | bool error; |
| 129 | }; |
| 130 | |
| 131 | struct cff_sub_table_info_t { |
| 132 | cff_sub_table_info_t () |
| 133 | : fd_array_link (0), |
| 134 | char_strings_link (0) |
| 135 | { |
| 136 | fd_select.init (); |
| 137 | } |
| 138 | |
| 139 | table_info_t fd_select; |
| 140 | objidx_t fd_array_link; |
| 141 | objidx_t char_strings_link; |
| 142 | }; |
| 143 | |
| 144 | template <typename OPSTR=op_str_t> |
| 145 | struct cff_top_dict_op_serializer_t : op_serializer_t |
| 146 | { |
| 147 | bool serialize (hb_serialize_context_t *c, |
| 148 | const OPSTR &opstr, |
| 149 | const cff_sub_table_info_t &info) const |
| 150 | { |
| 151 | TRACE_SERIALIZE (this); |
| 152 | |
| 153 | switch (opstr.op) |
| 154 | { |
| 155 | case OpCode_CharStrings: |
| 156 | return_trace (FontDict::serialize_link4_op(c, opstr.op, info.char_strings_link, whence_t::Absolute)); |
| 157 | |
| 158 | case OpCode_FDArray: |
| 159 | return_trace (FontDict::serialize_link4_op(c, opstr.op, info.fd_array_link, whence_t::Absolute)); |
| 160 | |
| 161 | case OpCode_FDSelect: |
| 162 | return_trace (FontDict::serialize_link4_op(c, opstr.op, info.fd_select.link, whence_t::Absolute)); |
| 163 | |
| 164 | default: |
| 165 | return_trace (copy_opstr (c, opstr)); |
| 166 | } |
| 167 | return_trace (true); |
| 168 | } |
| 169 | }; |
| 170 | |
| 171 | struct cff_font_dict_op_serializer_t : op_serializer_t |
| 172 | { |
| 173 | bool serialize (hb_serialize_context_t *c, |
| 174 | const op_str_t &opstr, |
| 175 | const table_info_t &privateDictInfo) const |
| 176 | { |
| 177 | TRACE_SERIALIZE (this); |
| 178 | |
| 179 | if (opstr.op == OpCode_Private) |
| 180 | { |
| 181 | /* serialize the private dict size & offset as 2-byte & 4-byte integers */ |
| 182 | return_trace (UnsizedByteStr::serialize_int2 (c, privateDictInfo.size) && |
| 183 | Dict::serialize_link4_op (c, opstr.op, privateDictInfo.link, whence_t::Absolute)); |
| 184 | } |
| 185 | else |
| 186 | { |
| 187 | HBUINT8 *d = c->allocate_size<HBUINT8> (opstr.str.length); |
| 188 | if (unlikely (!d)) return_trace (false); |
| 189 | memcpy (d, &opstr.str[0], opstr.str.length); |
| 190 | } |
| 191 | return_trace (true); |
| 192 | } |
| 193 | }; |
| 194 | |
| 195 | struct cff_private_dict_op_serializer_t : op_serializer_t |
| 196 | { |
| 197 | cff_private_dict_op_serializer_t (bool desubroutinize_, bool drop_hints_) |
| 198 | : desubroutinize (desubroutinize_), drop_hints (drop_hints_) {} |
| 199 | |
| 200 | bool serialize (hb_serialize_context_t *c, |
| 201 | const op_str_t &opstr, |
| 202 | objidx_t subrs_link) const |
| 203 | { |
| 204 | TRACE_SERIALIZE (this); |
| 205 | |
| 206 | if (drop_hints && dict_opset_t::is_hint_op (opstr.op)) |
| 207 | return true; |
| 208 | if (opstr.op == OpCode_Subrs) |
| 209 | { |
| 210 | if (desubroutinize || !subrs_link) |
| 211 | return_trace (true); |
| 212 | else |
| 213 | return_trace (FontDict::serialize_link2_op (c, opstr.op, subrs_link)); |
| 214 | } |
| 215 | else |
| 216 | return_trace (copy_opstr (c, opstr)); |
| 217 | } |
| 218 | |
| 219 | protected: |
| 220 | const bool desubroutinize; |
| 221 | const bool drop_hints; |
| 222 | }; |
| 223 | |
| 224 | struct flatten_param_t |
| 225 | { |
| 226 | str_buff_t &flatStr; |
| 227 | bool drop_hints; |
| 228 | }; |
| 229 | |
| 230 | template <typename ACC, typename ENV, typename OPSET, op_code_t endchar_op=OpCode_Invalid> |
| 231 | struct subr_flattener_t |
| 232 | { |
| 233 | subr_flattener_t (const ACC &acc_, |
| 234 | const hb_subset_plan_t *plan_) |
| 235 | : acc (acc_), plan (plan_) {} |
| 236 | |
| 237 | bool flatten (str_buff_vec_t &flat_charstrings) |
| 238 | { |
| 239 | if (!flat_charstrings.resize (plan->num_output_glyphs ())) |
| 240 | return false; |
| 241 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 242 | flat_charstrings[i].init (); |
| 243 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 244 | { |
| 245 | hb_codepoint_t glyph; |
| 246 | if (!plan->old_gid_for_new_gid (i, &glyph)) |
| 247 | { |
| 248 | /* add an endchar only charstring for a missing glyph if CFF1 */ |
| 249 | if (endchar_op != OpCode_Invalid) flat_charstrings[i].push (endchar_op); |
| 250 | continue; |
| 251 | } |
| 252 | const byte_str_t str = (*acc.charStrings)[glyph]; |
| 253 | unsigned int fd = acc.fdSelect->get_fd (glyph); |
| 254 | if (unlikely (fd >= acc.fdCount)) |
| 255 | return false; |
| 256 | cs_interpreter_t<ENV, OPSET, flatten_param_t> interp; |
| 257 | interp.env.init (str, acc, fd); |
| 258 | flatten_param_t param = { flat_charstrings[i], plan->drop_hints }; |
| 259 | if (unlikely (!interp.interpret (param))) |
| 260 | return false; |
| 261 | } |
| 262 | return true; |
| 263 | } |
| 264 | |
| 265 | const ACC &acc; |
| 266 | const hb_subset_plan_t *plan; |
| 267 | }; |
| 268 | |
| 269 | struct subr_closures_t |
| 270 | { |
| 271 | subr_closures_t () : valid (false), global_closure (nullptr) |
| 272 | { local_closures.init (); } |
| 273 | |
| 274 | void init (unsigned int fd_count) |
| 275 | { |
| 276 | valid = true; |
| 277 | global_closure = hb_set_create (); |
| 278 | if (global_closure == hb_set_get_empty ()) |
| 279 | valid = false; |
| 280 | if (!local_closures.resize (fd_count)) |
| 281 | valid = false; |
| 282 | |
| 283 | for (unsigned int i = 0; i < local_closures.length; i++) |
| 284 | { |
| 285 | local_closures[i] = hb_set_create (); |
| 286 | if (local_closures[i] == hb_set_get_empty ()) |
| 287 | valid = false; |
| 288 | } |
| 289 | } |
| 290 | |
| 291 | void fini () |
| 292 | { |
| 293 | hb_set_destroy (global_closure); |
| 294 | for (unsigned int i = 0; i < local_closures.length; i++) |
| 295 | hb_set_destroy (local_closures[i]); |
| 296 | local_closures.fini (); |
| 297 | } |
| 298 | |
| 299 | void reset () |
| 300 | { |
| 301 | hb_set_clear (global_closure); |
| 302 | for (unsigned int i = 0; i < local_closures.length; i++) |
| 303 | hb_set_clear (local_closures[i]); |
| 304 | } |
| 305 | |
| 306 | bool is_valid () const { return valid; } |
| 307 | bool valid; |
| 308 | hb_set_t *global_closure; |
| 309 | hb_vector_t<hb_set_t *> local_closures; |
| 310 | }; |
| 311 | |
| 312 | struct parsed_cs_op_t : op_str_t |
| 313 | { |
| 314 | void init (unsigned int subr_num_ = 0) |
| 315 | { |
| 316 | op_str_t::init (); |
| 317 | subr_num = subr_num_; |
| 318 | drop_flag = false; |
| 319 | keep_flag = false; |
| 320 | skip_flag = false; |
| 321 | } |
| 322 | |
| 323 | void fini () { op_str_t::fini (); } |
| 324 | |
| 325 | bool for_drop () const { return drop_flag; } |
| 326 | void set_drop () { if (!for_keep ()) drop_flag = true; } |
| 327 | |
| 328 | bool for_keep () const { return keep_flag; } |
| 329 | void set_keep () { keep_flag = true; } |
| 330 | |
| 331 | bool for_skip () const { return skip_flag; } |
| 332 | void set_skip () { skip_flag = true; } |
| 333 | |
| 334 | unsigned int subr_num; |
| 335 | |
| 336 | protected: |
| 337 | bool drop_flag : 1; |
| 338 | bool keep_flag : 1; |
| 339 | bool skip_flag : 1; |
| 340 | }; |
| 341 | |
| 342 | struct parsed_cs_str_t : parsed_values_t<parsed_cs_op_t> |
| 343 | { |
| 344 | void init () |
| 345 | { |
| 346 | SUPER::init (); |
| 347 | parsed = false; |
| 348 | hint_dropped = false; |
| 349 | has_prefix_ = false; |
| 350 | } |
| 351 | |
| 352 | void add_op (op_code_t op, const byte_str_ref_t& str_ref) |
| 353 | { |
| 354 | if (!is_parsed ()) |
| 355 | SUPER::add_op (op, str_ref); |
| 356 | } |
| 357 | |
| 358 | void add_call_op (op_code_t op, const byte_str_ref_t& str_ref, unsigned int subr_num) |
| 359 | { |
| 360 | if (!is_parsed ()) |
| 361 | { |
| 362 | unsigned int parsed_len = get_count (); |
| 363 | if (likely (parsed_len > 0)) |
| 364 | values[parsed_len-1].set_skip (); |
| 365 | |
| 366 | parsed_cs_op_t val; |
| 367 | val.init (subr_num); |
| 368 | SUPER::add_op (op, str_ref, val); |
| 369 | } |
| 370 | } |
| 371 | |
| 372 | void set_prefix (const number_t &num, op_code_t op = OpCode_Invalid) |
| 373 | { |
| 374 | has_prefix_ = true; |
| 375 | prefix_op_ = op; |
| 376 | prefix_num_ = num; |
| 377 | } |
| 378 | |
| 379 | bool at_end (unsigned int pos) const |
| 380 | { |
| 381 | return ((pos + 1 >= values.length) /* CFF2 */ |
| 382 | || (values[pos + 1].op == OpCode_return)); |
| 383 | } |
| 384 | |
| 385 | bool is_parsed () const { return parsed; } |
| 386 | void set_parsed () { parsed = true; } |
| 387 | |
| 388 | bool is_hint_dropped () const { return hint_dropped; } |
| 389 | void set_hint_dropped () { hint_dropped = true; } |
| 390 | |
| 391 | bool is_vsindex_dropped () const { return vsindex_dropped; } |
| 392 | void set_vsindex_dropped () { vsindex_dropped = true; } |
| 393 | |
| 394 | bool has_prefix () const { return has_prefix_; } |
| 395 | op_code_t prefix_op () const { return prefix_op_; } |
| 396 | const number_t &prefix_num () const { return prefix_num_; } |
| 397 | |
| 398 | protected: |
| 399 | bool parsed; |
| 400 | bool hint_dropped; |
| 401 | bool vsindex_dropped; |
| 402 | bool has_prefix_; |
| 403 | op_code_t prefix_op_; |
| 404 | number_t prefix_num_; |
| 405 | |
| 406 | private: |
| 407 | typedef parsed_values_t<parsed_cs_op_t> SUPER; |
| 408 | }; |
| 409 | |
| 410 | struct parsed_cs_str_vec_t : hb_vector_t<parsed_cs_str_t> |
| 411 | { |
| 412 | void init (unsigned int len_ = 0) |
| 413 | { |
| 414 | SUPER::init (); |
| 415 | resize (len_); |
| 416 | for (unsigned int i = 0; i < length; i++) |
| 417 | (*this)[i].init (); |
| 418 | } |
| 419 | void fini () { SUPER::fini_deep (); } |
| 420 | |
| 421 | private: |
| 422 | typedef hb_vector_t<parsed_cs_str_t> SUPER; |
| 423 | }; |
| 424 | |
| 425 | struct subr_subset_param_t |
| 426 | { |
| 427 | void init (parsed_cs_str_t *parsed_charstring_, |
| 428 | parsed_cs_str_vec_t *parsed_global_subrs_, parsed_cs_str_vec_t *parsed_local_subrs_, |
| 429 | hb_set_t *global_closure_, hb_set_t *local_closure_, |
| 430 | bool drop_hints_) |
| 431 | { |
| 432 | parsed_charstring = parsed_charstring_; |
| 433 | current_parsed_str = parsed_charstring; |
| 434 | parsed_global_subrs = parsed_global_subrs_; |
| 435 | parsed_local_subrs = parsed_local_subrs_; |
| 436 | global_closure = global_closure_; |
| 437 | local_closure = local_closure_; |
| 438 | drop_hints = drop_hints_; |
| 439 | } |
| 440 | |
| 441 | parsed_cs_str_t *get_parsed_str_for_context (call_context_t &context) |
| 442 | { |
| 443 | switch (context.type) |
| 444 | { |
| 445 | case CSType_CharString: |
| 446 | return parsed_charstring; |
| 447 | |
| 448 | case CSType_LocalSubr: |
| 449 | if (likely (context.subr_num < parsed_local_subrs->length)) |
| 450 | return &(*parsed_local_subrs)[context.subr_num]; |
| 451 | break; |
| 452 | |
| 453 | case CSType_GlobalSubr: |
| 454 | if (likely (context.subr_num < parsed_global_subrs->length)) |
| 455 | return &(*parsed_global_subrs)[context.subr_num]; |
| 456 | break; |
| 457 | } |
| 458 | return nullptr; |
| 459 | } |
| 460 | |
| 461 | template <typename ENV> |
| 462 | void set_current_str (ENV &env, bool calling) |
| 463 | { |
| 464 | parsed_cs_str_t *parsed_str = get_parsed_str_for_context (env.context); |
| 465 | if (unlikely (!parsed_str)) |
| 466 | { |
| 467 | env.set_error (); |
| 468 | return; |
| 469 | } |
| 470 | /* If the called subroutine is parsed partially but not completely yet, |
| 471 | * it must be because we are calling it recursively. |
| 472 | * Handle it as an error. */ |
| 473 | if (unlikely (calling && !parsed_str->is_parsed () && (parsed_str->values.length > 0))) |
| 474 | env.set_error (); |
| 475 | else |
| 476 | current_parsed_str = parsed_str; |
| 477 | } |
| 478 | |
| 479 | parsed_cs_str_t *current_parsed_str; |
| 480 | |
| 481 | parsed_cs_str_t *parsed_charstring; |
| 482 | parsed_cs_str_vec_t *parsed_global_subrs; |
| 483 | parsed_cs_str_vec_t *parsed_local_subrs; |
| 484 | hb_set_t *global_closure; |
| 485 | hb_set_t *local_closure; |
| 486 | bool drop_hints; |
| 487 | }; |
| 488 | |
| 489 | struct subr_remap_t : hb_inc_bimap_t |
| 490 | { |
| 491 | void create (hb_set_t *closure) |
| 492 | { |
| 493 | /* create a remapping of subroutine numbers from old to new. |
| 494 | * no optimization based on usage counts. fonttools doesn't appear doing that either. |
| 495 | */ |
| 496 | |
| 497 | hb_codepoint_t old_num = HB_SET_VALUE_INVALID; |
| 498 | while (hb_set_next (closure, &old_num)) |
| 499 | add (old_num); |
| 500 | |
| 501 | if (get_population () < 1240) |
| 502 | bias = 107; |
| 503 | else if (get_population () < 33900) |
| 504 | bias = 1131; |
| 505 | else |
| 506 | bias = 32768; |
| 507 | } |
| 508 | |
| 509 | int biased_num (unsigned int old_num) const |
| 510 | { |
| 511 | hb_codepoint_t new_num = get (old_num); |
| 512 | return (int)new_num - bias; |
| 513 | } |
| 514 | |
| 515 | protected: |
| 516 | int bias; |
| 517 | }; |
| 518 | |
| 519 | struct subr_remaps_t |
| 520 | { |
| 521 | subr_remaps_t () |
| 522 | { |
| 523 | global_remap.init (); |
| 524 | local_remaps.init (); |
| 525 | } |
| 526 | |
| 527 | ~subr_remaps_t () { fini (); } |
| 528 | |
| 529 | void init (unsigned int fdCount) |
| 530 | { |
| 531 | local_remaps.resize (fdCount); |
| 532 | for (unsigned int i = 0; i < fdCount; i++) |
| 533 | local_remaps[i].init (); |
| 534 | } |
| 535 | |
| 536 | void create (subr_closures_t& closures) |
| 537 | { |
| 538 | global_remap.create (closures.global_closure); |
| 539 | for (unsigned int i = 0; i < local_remaps.length; i++) |
| 540 | local_remaps[i].create (closures.local_closures[i]); |
| 541 | } |
| 542 | |
| 543 | void fini () |
| 544 | { |
| 545 | global_remap.fini (); |
| 546 | local_remaps.fini_deep (); |
| 547 | } |
| 548 | |
| 549 | subr_remap_t global_remap; |
| 550 | hb_vector_t<subr_remap_t> local_remaps; |
| 551 | }; |
| 552 | |
| 553 | template <typename SUBSETTER, typename SUBRS, typename ACC, typename ENV, typename OPSET, op_code_t endchar_op=OpCode_Invalid> |
| 554 | struct subr_subsetter_t |
| 555 | { |
| 556 | subr_subsetter_t (ACC &acc_, const hb_subset_plan_t *plan_) |
| 557 | : acc (acc_), plan (plan_) |
| 558 | { |
| 559 | parsed_charstrings.init (); |
| 560 | parsed_global_subrs.init (); |
| 561 | parsed_local_subrs.init (); |
| 562 | } |
| 563 | |
| 564 | ~subr_subsetter_t () |
| 565 | { |
| 566 | closures.fini (); |
| 567 | remaps.fini (); |
| 568 | parsed_charstrings.fini_deep (); |
| 569 | parsed_global_subrs.fini_deep (); |
| 570 | parsed_local_subrs.fini_deep (); |
| 571 | } |
| 572 | |
| 573 | /* Subroutine subsetting with --no-desubroutinize runs in phases: |
| 574 | * |
| 575 | * 1. execute charstrings/subroutines to determine subroutine closures |
| 576 | * 2. parse out all operators and numbers |
| 577 | * 3. mark hint operators and operands for removal if --no-hinting |
| 578 | * 4. re-encode all charstrings and subroutines with new subroutine numbers |
| 579 | * |
| 580 | * Phases #1 and #2 are done at the same time in collect_subrs (). |
| 581 | * Phase #3 walks charstrings/subroutines forward then backward (hence parsing required), |
| 582 | * because we can't tell if a number belongs to a hint op until we see the first moveto. |
| 583 | * |
| 584 | * Assumption: a callsubr/callgsubr operator must immediately follow a (biased) subroutine number |
| 585 | * within the same charstring/subroutine, e.g., not split across a charstring and a subroutine. |
| 586 | */ |
| 587 | bool subset (void) |
| 588 | { |
| 589 | closures.init (acc.fdCount); |
| 590 | remaps.init (acc.fdCount); |
| 591 | |
| 592 | parsed_charstrings.init (plan->num_output_glyphs ()); |
| 593 | parsed_global_subrs.init (acc.globalSubrs->count); |
| 594 | parsed_local_subrs.resize (acc.fdCount); |
| 595 | for (unsigned int i = 0; i < acc.fdCount; i++) |
| 596 | { |
| 597 | parsed_local_subrs[i].init (acc.privateDicts[i].localSubrs->count); |
| 598 | } |
| 599 | if (unlikely (!closures.valid)) |
| 600 | return false; |
| 601 | |
| 602 | /* phase 1 & 2 */ |
| 603 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 604 | { |
| 605 | hb_codepoint_t glyph; |
| 606 | if (!plan->old_gid_for_new_gid (i, &glyph)) |
| 607 | continue; |
| 608 | const byte_str_t str = (*acc.charStrings)[glyph]; |
| 609 | unsigned int fd = acc.fdSelect->get_fd (glyph); |
| 610 | if (unlikely (fd >= acc.fdCount)) |
| 611 | return false; |
| 612 | |
| 613 | cs_interpreter_t<ENV, OPSET, subr_subset_param_t> interp; |
| 614 | interp.env.init (str, acc, fd); |
| 615 | |
| 616 | subr_subset_param_t param; |
| 617 | param.init (&parsed_charstrings[i], |
| 618 | &parsed_global_subrs, &parsed_local_subrs[fd], |
| 619 | closures.global_closure, closures.local_closures[fd], |
| 620 | plan->drop_hints); |
| 621 | |
| 622 | if (unlikely (!interp.interpret (param))) |
| 623 | return false; |
| 624 | |
| 625 | /* complete parsed string esp. copy CFF1 width or CFF2 vsindex to the parsed charstring for encoding */ |
| 626 | SUBSETTER::complete_parsed_str (interp.env, param, parsed_charstrings[i]); |
| 627 | } |
| 628 | |
| 629 | if (plan->drop_hints) |
| 630 | { |
| 631 | /* mark hint ops and arguments for drop */ |
| 632 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 633 | { |
| 634 | hb_codepoint_t glyph; |
| 635 | if (!plan->old_gid_for_new_gid (i, &glyph)) |
| 636 | continue; |
| 637 | unsigned int fd = acc.fdSelect->get_fd (glyph); |
| 638 | if (unlikely (fd >= acc.fdCount)) |
| 639 | return false; |
| 640 | subr_subset_param_t param; |
| 641 | param.init (&parsed_charstrings[i], |
| 642 | &parsed_global_subrs, &parsed_local_subrs[fd], |
| 643 | closures.global_closure, closures.local_closures[fd], |
| 644 | plan->drop_hints); |
| 645 | |
| 646 | drop_hints_param_t drop; |
| 647 | if (drop_hints_in_str (parsed_charstrings[i], param, drop)) |
| 648 | { |
| 649 | parsed_charstrings[i].set_hint_dropped (); |
| 650 | if (drop.vsindex_dropped) |
| 651 | parsed_charstrings[i].set_vsindex_dropped (); |
| 652 | } |
| 653 | } |
| 654 | |
| 655 | /* after dropping hints recreate closures of actually used subrs */ |
| 656 | closures.reset (); |
| 657 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 658 | { |
| 659 | hb_codepoint_t glyph; |
| 660 | if (!plan->old_gid_for_new_gid (i, &glyph)) |
| 661 | continue; |
| 662 | unsigned int fd = acc.fdSelect->get_fd (glyph); |
| 663 | if (unlikely (fd >= acc.fdCount)) |
| 664 | return false; |
| 665 | subr_subset_param_t param; |
| 666 | param.init (&parsed_charstrings[i], |
| 667 | &parsed_global_subrs, &parsed_local_subrs[fd], |
| 668 | closures.global_closure, closures.local_closures[fd], |
| 669 | plan->drop_hints); |
| 670 | collect_subr_refs_in_str (parsed_charstrings[i], param); |
| 671 | } |
| 672 | } |
| 673 | |
| 674 | remaps.create (closures); |
| 675 | |
| 676 | return true; |
| 677 | } |
| 678 | |
| 679 | bool encode_charstrings (str_buff_vec_t &buffArray) const |
| 680 | { |
| 681 | if (unlikely (!buffArray.resize (plan->num_output_glyphs ()))) |
| 682 | return false; |
| 683 | for (unsigned int i = 0; i < plan->num_output_glyphs (); i++) |
| 684 | { |
| 685 | hb_codepoint_t glyph; |
| 686 | if (!plan->old_gid_for_new_gid (i, &glyph)) |
| 687 | { |
| 688 | /* add an endchar only charstring for a missing glyph if CFF1 */ |
| 689 | if (endchar_op != OpCode_Invalid) buffArray[i].push (endchar_op); |
| 690 | continue; |
| 691 | } |
| 692 | unsigned int fd = acc.fdSelect->get_fd (glyph); |
| 693 | if (unlikely (fd >= acc.fdCount)) |
| 694 | return false; |
| 695 | if (unlikely (!encode_str (parsed_charstrings[i], fd, buffArray[i]))) |
| 696 | return false; |
| 697 | } |
| 698 | return true; |
| 699 | } |
| 700 | |
| 701 | bool encode_subrs (const parsed_cs_str_vec_t &subrs, const subr_remap_t& remap, unsigned int fd, str_buff_vec_t &buffArray) const |
| 702 | { |
| 703 | unsigned int count = remap.get_population (); |
| 704 | |
| 705 | if (unlikely (!buffArray.resize (count))) |
| 706 | return false; |
| 707 | for (unsigned int old_num = 0; old_num < subrs.length; old_num++) |
| 708 | { |
| 709 | hb_codepoint_t new_num = remap[old_num]; |
| 710 | if (new_num != CFF_UNDEF_CODE) |
| 711 | { |
| 712 | if (unlikely (!encode_str (subrs[old_num], fd, buffArray[new_num]))) |
| 713 | return false; |
| 714 | } |
| 715 | } |
| 716 | return true; |
| 717 | } |
| 718 | |
| 719 | bool encode_globalsubrs (str_buff_vec_t &buffArray) |
| 720 | { |
| 721 | return encode_subrs (parsed_global_subrs, remaps.global_remap, 0, buffArray); |
| 722 | } |
| 723 | |
| 724 | bool encode_localsubrs (unsigned int fd, str_buff_vec_t &buffArray) const |
| 725 | { |
| 726 | return encode_subrs (parsed_local_subrs[fd], remaps.local_remaps[fd], fd, buffArray); |
| 727 | } |
| 728 | |
| 729 | protected: |
| 730 | struct drop_hints_param_t |
| 731 | { |
| 732 | drop_hints_param_t () |
| 733 | : seen_moveto (false), |
| 734 | ends_in_hint (false), |
| 735 | all_dropped (false), |
| 736 | vsindex_dropped (false) {} |
| 737 | |
| 738 | bool seen_moveto; |
| 739 | bool ends_in_hint; |
| 740 | bool all_dropped; |
| 741 | bool vsindex_dropped; |
| 742 | }; |
| 743 | |
| 744 | bool drop_hints_in_subr (parsed_cs_str_t &str, unsigned int pos, |
| 745 | parsed_cs_str_vec_t &subrs, unsigned int subr_num, |
| 746 | const subr_subset_param_t ¶m, drop_hints_param_t &drop) |
| 747 | { |
| 748 | drop.ends_in_hint = false; |
| 749 | bool has_hint = drop_hints_in_str (subrs[subr_num], param, drop); |
| 750 | |
| 751 | /* if this subr ends with a stem hint (i.e., not a number; potential argument for moveto), |
| 752 | * then this entire subroutine must be a hint. drop its call. */ |
| 753 | if (drop.ends_in_hint) |
| 754 | { |
| 755 | str.values[pos].set_drop (); |
| 756 | /* if this subr call is at the end of the parent subr, propagate the flag |
| 757 | * otherwise reset the flag */ |
| 758 | if (!str.at_end (pos)) |
| 759 | drop.ends_in_hint = false; |
| 760 | } |
| 761 | else if (drop.all_dropped) |
| 762 | { |
| 763 | str.values[pos].set_drop (); |
| 764 | } |
| 765 | |
| 766 | return has_hint; |
| 767 | } |
| 768 | |
| 769 | /* returns true if it sees a hint op before the first moveto */ |
| 770 | bool drop_hints_in_str (parsed_cs_str_t &str, const subr_subset_param_t ¶m, drop_hints_param_t &drop) |
| 771 | { |
| 772 | bool seen_hint = false; |
| 773 | |
| 774 | for (unsigned int pos = 0; pos < str.values.length; pos++) |
| 775 | { |
| 776 | bool has_hint = false; |
| 777 | switch (str.values[pos].op) |
| 778 | { |
| 779 | case OpCode_callsubr: |
| 780 | has_hint = drop_hints_in_subr (str, pos, |
| 781 | *param.parsed_local_subrs, str.values[pos].subr_num, |
| 782 | param, drop); |
| 783 | break; |
| 784 | |
| 785 | case OpCode_callgsubr: |
| 786 | has_hint = drop_hints_in_subr (str, pos, |
| 787 | *param.parsed_global_subrs, str.values[pos].subr_num, |
| 788 | param, drop); |
| 789 | break; |
| 790 | |
| 791 | case OpCode_rmoveto: |
| 792 | case OpCode_hmoveto: |
| 793 | case OpCode_vmoveto: |
| 794 | drop.seen_moveto = true; |
| 795 | break; |
| 796 | |
| 797 | case OpCode_hintmask: |
| 798 | case OpCode_cntrmask: |
| 799 | if (drop.seen_moveto) |
| 800 | { |
| 801 | str.values[pos].set_drop (); |
| 802 | break; |
| 803 | } |
| 804 | HB_FALLTHROUGH; |
| 805 | |
| 806 | case OpCode_hstemhm: |
| 807 | case OpCode_vstemhm: |
| 808 | case OpCode_hstem: |
| 809 | case OpCode_vstem: |
| 810 | has_hint = true; |
| 811 | str.values[pos].set_drop (); |
| 812 | if (str.at_end (pos)) |
| 813 | drop.ends_in_hint = true; |
| 814 | break; |
| 815 | |
| 816 | case OpCode_dotsection: |
| 817 | str.values[pos].set_drop (); |
| 818 | break; |
| 819 | |
| 820 | default: |
| 821 | /* NONE */ |
| 822 | break; |
| 823 | } |
| 824 | if (has_hint) |
| 825 | { |
| 826 | for (int i = pos - 1; i >= 0; i--) |
| 827 | { |
| 828 | parsed_cs_op_t &csop = str.values[(unsigned)i]; |
| 829 | if (csop.for_drop ()) |
| 830 | break; |
| 831 | csop.set_drop (); |
| 832 | if (csop.op == OpCode_vsindexcs) |
| 833 | drop.vsindex_dropped = true; |
| 834 | } |
| 835 | seen_hint |= has_hint; |
| 836 | } |
| 837 | } |
| 838 | |
| 839 | /* Raise all_dropped flag if all operators except return are dropped from a subr. |
| 840 | * It may happen even after seeing the first moveto if a subr contains |
| 841 | * only (usually one) hintmask operator, then calls to this subr can be dropped. |
| 842 | */ |
| 843 | drop.all_dropped = true; |
| 844 | for (unsigned int pos = 0; pos < str.values.length; pos++) |
| 845 | { |
| 846 | parsed_cs_op_t &csop = str.values[pos]; |
| 847 | if (csop.op == OpCode_return) |
| 848 | break; |
| 849 | if (!csop.for_drop ()) |
| 850 | { |
| 851 | drop.all_dropped = false; |
| 852 | break; |
| 853 | } |
| 854 | } |
| 855 | |
| 856 | return seen_hint; |
| 857 | } |
| 858 | |
| 859 | void collect_subr_refs_in_subr (parsed_cs_str_t &str, unsigned int pos, |
| 860 | unsigned int subr_num, parsed_cs_str_vec_t &subrs, |
| 861 | hb_set_t *closure, |
| 862 | const subr_subset_param_t ¶m) |
| 863 | { |
| 864 | closure->add (subr_num); |
| 865 | collect_subr_refs_in_str (subrs[subr_num], param); |
| 866 | } |
| 867 | |
| 868 | void collect_subr_refs_in_str (parsed_cs_str_t &str, const subr_subset_param_t ¶m) |
| 869 | { |
| 870 | for (unsigned int pos = 0; pos < str.values.length; pos++) |
| 871 | { |
| 872 | if (!str.values[pos].for_drop ()) |
| 873 | { |
| 874 | switch (str.values[pos].op) |
| 875 | { |
| 876 | case OpCode_callsubr: |
| 877 | collect_subr_refs_in_subr (str, pos, |
| 878 | str.values[pos].subr_num, *param.parsed_local_subrs, |
| 879 | param.local_closure, param); |
| 880 | break; |
| 881 | |
| 882 | case OpCode_callgsubr: |
| 883 | collect_subr_refs_in_subr (str, pos, |
| 884 | str.values[pos].subr_num, *param.parsed_global_subrs, |
| 885 | param.global_closure, param); |
| 886 | break; |
| 887 | |
| 888 | default: break; |
| 889 | } |
| 890 | } |
| 891 | } |
| 892 | } |
| 893 | |
| 894 | bool encode_str (const parsed_cs_str_t &str, const unsigned int fd, str_buff_t &buff) const |
| 895 | { |
| 896 | buff.init (); |
| 897 | str_encoder_t encoder (buff); |
| 898 | encoder.reset (); |
| 899 | /* if a prefix (CFF1 width or CFF2 vsindex) has been removed along with hints, |
| 900 | * re-insert it at the beginning of charstreing */ |
| 901 | if (str.has_prefix () && str.is_hint_dropped ()) |
| 902 | { |
| 903 | encoder.encode_num (str.prefix_num ()); |
| 904 | if (str.prefix_op () != OpCode_Invalid) |
| 905 | encoder.encode_op (str.prefix_op ()); |
| 906 | } |
| 907 | for (unsigned int i = 0; i < str.get_count(); i++) |
| 908 | { |
| 909 | const parsed_cs_op_t &opstr = str.values[i]; |
| 910 | if (!opstr.for_drop () && !opstr.for_skip ()) |
| 911 | { |
| 912 | switch (opstr.op) |
| 913 | { |
| 914 | case OpCode_callsubr: |
| 915 | encoder.encode_int (remaps.local_remaps[fd].biased_num (opstr.subr_num)); |
| 916 | encoder.encode_op (OpCode_callsubr); |
| 917 | break; |
| 918 | |
| 919 | case OpCode_callgsubr: |
| 920 | encoder.encode_int (remaps.global_remap.biased_num (opstr.subr_num)); |
| 921 | encoder.encode_op (OpCode_callgsubr); |
| 922 | break; |
| 923 | |
| 924 | default: |
| 925 | encoder.copy_str (opstr.str); |
| 926 | break; |
| 927 | } |
| 928 | } |
| 929 | } |
| 930 | return !encoder.is_error (); |
| 931 | } |
| 932 | |
| 933 | protected: |
| 934 | const ACC &acc; |
| 935 | const hb_subset_plan_t *plan; |
| 936 | |
| 937 | subr_closures_t closures; |
| 938 | |
| 939 | parsed_cs_str_vec_t parsed_charstrings; |
| 940 | parsed_cs_str_vec_t parsed_global_subrs; |
| 941 | hb_vector_t<parsed_cs_str_vec_t> parsed_local_subrs; |
| 942 | |
| 943 | subr_remaps_t remaps; |
| 944 | |
| 945 | private: |
| 946 | typedef typename SUBRS::count_type subr_count_type; |
| 947 | }; |
| 948 | |
| 949 | } /* namespace CFF */ |
| 950 | |
| 951 | HB_INTERNAL bool |
| 952 | hb_plan_subset_cff_fdselect (const hb_subset_plan_t *plan, |
| 953 | unsigned int fdCount, |
| 954 | const CFF::FDSelect &src, /* IN */ |
| 955 | unsigned int &subset_fd_count /* OUT */, |
| 956 | unsigned int &subset_fdselect_size /* OUT */, |
| 957 | unsigned int &subset_fdselect_format /* OUT */, |
| 958 | hb_vector_t<CFF::code_pair_t> &fdselect_ranges /* OUT */, |
| 959 | hb_inc_bimap_t &fdmap /* OUT */); |
| 960 | |
| 961 | HB_INTERNAL bool |
| 962 | hb_serialize_cff_fdselect (hb_serialize_context_t *c, |
| 963 | unsigned int num_glyphs, |
| 964 | const CFF::FDSelect &src, |
| 965 | unsigned int fd_count, |
| 966 | unsigned int fdselect_format, |
| 967 | unsigned int size, |
| 968 | const hb_vector_t<CFF::code_pair_t> &fdselect_ranges); |
| 969 | |
| 970 | #endif /* HB_SUBSET_CFF_COMMON_HH */ |
| 971 | |