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#include "hb.hh"
28
29#ifndef HB_NO_SUBSET_CFF
30
31#include "hb-open-type.hh"
32#include "hb-ot-cff1-table.hh"
33#include "hb-set.h"
34#include "hb-bimap.hh"
35#include "hb-subset-cff1.hh"
36#include "hb-subset-plan.hh"
37#include "hb-subset-cff-common.hh"
38#include "hb-cff1-interp-cs.hh"
39
40using namespace CFF;
41
42struct remap_sid_t : hb_inc_bimap_t
43{
44 unsigned int add (unsigned int sid)
45 {
46 if ((sid != CFF_UNDEF_SID) && !is_std_std (sid))
47 return offset_sid (hb_inc_bimap_t::add (unoffset_sid (sid)));
48 else
49 return sid;
50 }
51
52 unsigned int operator[] (unsigned int sid) const
53 {
54 if (is_std_std (sid) || (sid == CFF_UNDEF_SID))
55 return sid;
56 else
57 return offset_sid (get (unoffset_sid (sid)));
58 }
59
60 static const unsigned int num_std_strings = 391;
61
62 static bool is_std_std (unsigned int sid) { return sid < num_std_strings; }
63 static unsigned int offset_sid (unsigned int sid) { return sid + num_std_strings; }
64 static unsigned int unoffset_sid (unsigned int sid) { return sid - num_std_strings; }
65};
66
67struct cff1_sub_table_offsets_t : cff_sub_table_offsets_t
68{
69 cff1_sub_table_offsets_t ()
70 : cff_sub_table_offsets_t (),
71 nameIndexOffset (0),
72 encodingOffset (0)
73 {
74 stringIndexInfo.init ();
75 charsetInfo.init ();
76 privateDictInfo.init ();
77 }
78
79 unsigned int nameIndexOffset;
80 table_info_t stringIndexInfo;
81 unsigned int encodingOffset;
82 table_info_t charsetInfo;
83 table_info_t privateDictInfo;
84};
85
86/* a copy of a parsed out cff1_top_dict_values_t augmented with additional operators */
87struct cff1_top_dict_values_mod_t : cff1_top_dict_values_t
88{
89 void init (const cff1_top_dict_values_t *base_= &Null(cff1_top_dict_values_t))
90 {
91 SUPER::init ();
92 base = base_;
93 }
94
95 void fini () { SUPER::fini (); }
96
97 unsigned get_count () const { return base->get_count () + SUPER::get_count (); }
98 const cff1_top_dict_val_t &get_value (unsigned int i) const
99 {
100 if (i < base->get_count ())
101 return (*base)[i];
102 else
103 return SUPER::values[i - base->get_count ()];
104 }
105 const cff1_top_dict_val_t &operator [] (unsigned int i) const { return get_value (i); }
106
107 void reassignSIDs (const remap_sid_t& sidmap)
108 {
109 for (unsigned int i = 0; i < name_dict_values_t::ValCount; i++)
110 nameSIDs[i] = sidmap[base->nameSIDs[i]];
111 }
112
113 protected:
114 typedef cff1_top_dict_values_t SUPER;
115 const cff1_top_dict_values_t *base;
116};
117
118struct top_dict_modifiers_t
119{
120 top_dict_modifiers_t (const cff1_sub_table_offsets_t &offsets_,
121 const unsigned int (&nameSIDs_)[name_dict_values_t::ValCount])
122 : offsets (offsets_),
123 nameSIDs (nameSIDs_)
124 {}
125
126 const cff1_sub_table_offsets_t &offsets;
127 const unsigned int (&nameSIDs)[name_dict_values_t::ValCount];
128};
129
130struct cff1_top_dict_op_serializer_t : cff_top_dict_op_serializer_t<cff1_top_dict_val_t>
131{
132 bool serialize (hb_serialize_context_t *c,
133 const cff1_top_dict_val_t &opstr,
134 const top_dict_modifiers_t &mod) const
135 {
136 TRACE_SERIALIZE (this);
137
138 op_code_t op = opstr.op;
139 switch (op)
140 {
141 case OpCode_charset:
142 return_trace (FontDict::serialize_offset4_op(c, op, mod.offsets.charsetInfo.offset));
143
144 case OpCode_Encoding:
145 return_trace (FontDict::serialize_offset4_op(c, op, mod.offsets.encodingOffset));
146
147 case OpCode_Private:
148 {
149 if (unlikely (!UnsizedByteStr::serialize_int2 (c, mod.offsets.privateDictInfo.size)))
150 return_trace (false);
151 if (unlikely (!UnsizedByteStr::serialize_int4 (c, mod.offsets.privateDictInfo.offset)))
152 return_trace (false);
153 HBUINT8 *p = c->allocate_size<HBUINT8> (1);
154 if (unlikely (p == nullptr)) return_trace (false);
155 *p = OpCode_Private;
156 }
157 break;
158
159 case OpCode_version:
160 case OpCode_Notice:
161 case OpCode_Copyright:
162 case OpCode_FullName:
163 case OpCode_FamilyName:
164 case OpCode_Weight:
165 case OpCode_PostScript:
166 case OpCode_BaseFontName:
167 case OpCode_FontName:
168 return_trace (FontDict::serialize_offset2_op(c, op, mod.nameSIDs[name_dict_values_t::name_op_to_index (op)]));
169
170 case OpCode_ROS:
171 {
172 /* for registry & ordering, reassigned SIDs are serialized
173 * for supplement, the original byte string is copied along with the op code */
174 op_str_t supp_op;
175 supp_op.op = op;
176 if ( unlikely (!(opstr.str.length >= opstr.last_arg_offset + 3)))
177 return_trace (false);
178 supp_op.str = byte_str_t (&opstr.str + opstr.last_arg_offset, opstr.str.length - opstr.last_arg_offset);
179 return_trace (UnsizedByteStr::serialize_int2 (c, mod.nameSIDs[name_dict_values_t::registry]) &&
180 UnsizedByteStr::serialize_int2 (c, mod.nameSIDs[name_dict_values_t::ordering]) &&
181 copy_opstr (c, supp_op));
182 }
183 default:
184 return_trace (cff_top_dict_op_serializer_t<cff1_top_dict_val_t>::serialize (c, opstr, mod.offsets));
185 }
186 return_trace (true);
187 }
188
189 unsigned int calculate_serialized_size (const cff1_top_dict_val_t &opstr) const
190 {
191 op_code_t op = opstr.op;
192 switch (op)
193 {
194 case OpCode_charset:
195 case OpCode_Encoding:
196 return OpCode_Size (OpCode_longintdict) + 4 + OpCode_Size (op);
197
198 case OpCode_Private:
199 return OpCode_Size (OpCode_longintdict) + 4 + OpCode_Size (OpCode_shortint) + 2 + OpCode_Size (OpCode_Private);
200
201 case OpCode_version:
202 case OpCode_Notice:
203 case OpCode_Copyright:
204 case OpCode_FullName:
205 case OpCode_FamilyName:
206 case OpCode_Weight:
207 case OpCode_PostScript:
208 case OpCode_BaseFontName:
209 case OpCode_FontName:
210 return OpCode_Size (OpCode_shortint) + 2 + OpCode_Size (op);
211
212 case OpCode_ROS:
213 return ((OpCode_Size (OpCode_shortint) + 2) * 2) + (opstr.str.length - opstr.last_arg_offset)/* supplement + op */;
214
215 default:
216 return cff_top_dict_op_serializer_t<cff1_top_dict_val_t>::calculate_serialized_size (opstr);
217 }
218 }
219};
220
221struct font_dict_values_mod_t
222{
223 void init (const cff1_font_dict_values_t *base_,
224 unsigned int fontName_,
225 const table_info_t &privateDictInfo_)
226 {
227 base = base_;
228 fontName = fontName_;
229 privateDictInfo = privateDictInfo_;
230 }
231
232 unsigned get_count () const { return base->get_count (); }
233
234 const op_str_t &operator [] (unsigned int i) const { return (*base)[i]; }
235
236 const cff1_font_dict_values_t *base;
237 table_info_t privateDictInfo;
238 unsigned int fontName;
239};
240
241struct cff1_font_dict_op_serializer_t : cff_font_dict_op_serializer_t
242{
243 bool serialize (hb_serialize_context_t *c,
244 const op_str_t &opstr,
245 const font_dict_values_mod_t &mod) const
246 {
247 TRACE_SERIALIZE (this);
248
249 if (opstr.op == OpCode_FontName)
250 return_trace (FontDict::serialize_uint2_op (c, opstr.op, mod.fontName));
251 else
252 return_trace (SUPER::serialize (c, opstr, mod.privateDictInfo));
253 }
254
255 unsigned int calculate_serialized_size (const op_str_t &opstr) const
256 {
257 if (opstr.op == OpCode_FontName)
258 return OpCode_Size (OpCode_shortint) + 2 + OpCode_Size (OpCode_FontName);
259 else
260 return SUPER::calculate_serialized_size (opstr);
261 }
262
263 private:
264 typedef cff_font_dict_op_serializer_t SUPER;
265};
266
267struct cff1_cs_opset_flatten_t : cff1_cs_opset_t<cff1_cs_opset_flatten_t, flatten_param_t>
268{
269 static void flush_args_and_op (op_code_t op, cff1_cs_interp_env_t &env, flatten_param_t& param)
270 {
271 if (env.arg_start > 0)
272 flush_width (env, param);
273
274 switch (op)
275 {
276 case OpCode_hstem:
277 case OpCode_hstemhm:
278 case OpCode_vstem:
279 case OpCode_vstemhm:
280 case OpCode_hintmask:
281 case OpCode_cntrmask:
282 case OpCode_dotsection:
283 if (param.drop_hints)
284 {
285 env.clear_args ();
286 return;
287 }
288 HB_FALLTHROUGH;
289
290 default:
291 SUPER::flush_args_and_op (op, env, param);
292 break;
293 }
294 }
295 static void flush_args (cff1_cs_interp_env_t &env, flatten_param_t& param)
296 {
297 str_encoder_t encoder (param.flatStr);
298 for (unsigned int i = env.arg_start; i < env.argStack.get_count (); i++)
299 encoder.encode_num (env.eval_arg (i));
300 SUPER::flush_args (env, param);
301 }
302
303 static void flush_op (op_code_t op, cff1_cs_interp_env_t &env, flatten_param_t& param)
304 {
305 str_encoder_t encoder (param.flatStr);
306 encoder.encode_op (op);
307 }
308
309 static void flush_width (cff1_cs_interp_env_t &env, flatten_param_t& param)
310 {
311 assert (env.has_width);
312 str_encoder_t encoder (param.flatStr);
313 encoder.encode_num (env.width);
314 }
315
316 static void flush_hintmask (op_code_t op, cff1_cs_interp_env_t &env, flatten_param_t& param)
317 {
318 SUPER::flush_hintmask (op, env, param);
319 if (!param.drop_hints)
320 {
321 str_encoder_t encoder (param.flatStr);
322 for (unsigned int i = 0; i < env.hintmask_size; i++)
323 encoder.encode_byte (env.str_ref[i]);
324 }
325 }
326
327 private:
328 typedef cff1_cs_opset_t<cff1_cs_opset_flatten_t, flatten_param_t> SUPER;
329};
330
331struct range_list_t : hb_vector_t<code_pair_t>
332{
333 /* replace the first glyph ID in the "glyph" field each range with a nLeft value */
334 bool finalize (unsigned int last_glyph)
335 {
336 bool two_byte = false;
337 for (unsigned int i = (*this).length; i > 0; i--)
338 {
339 code_pair_t &pair = (*this)[i - 1];
340 unsigned int nLeft = last_glyph - pair.glyph - 1;
341 if (nLeft >= 0x100)
342 two_byte = true;
343 last_glyph = pair.glyph;
344 pair.glyph = nLeft;
345 }
346 return two_byte;
347 }
348};
349
350struct cff1_cs_opset_subr_subset_t : cff1_cs_opset_t<cff1_cs_opset_subr_subset_t, subr_subset_param_t>
351{
352 static void process_op (op_code_t op, cff1_cs_interp_env_t &env, subr_subset_param_t& param)
353 {
354 switch (op) {
355
356 case OpCode_return:
357 param.current_parsed_str->add_op (op, env.str_ref);
358 param.current_parsed_str->set_parsed ();
359 env.return_from_subr ();
360 param.set_current_str (env, false);
361 break;
362
363 case OpCode_endchar:
364 param.current_parsed_str->add_op (op, env.str_ref);
365 param.current_parsed_str->set_parsed ();
366 SUPER::process_op (op, env, param);
367 break;
368
369 case OpCode_callsubr:
370 process_call_subr (op, CSType_LocalSubr, env, param, env.localSubrs, param.local_closure);
371 break;
372
373 case OpCode_callgsubr:
374 process_call_subr (op, CSType_GlobalSubr, env, param, env.globalSubrs, param.global_closure);
375 break;
376
377 default:
378 SUPER::process_op (op, env, param);
379 param.current_parsed_str->add_op (op, env.str_ref);
380 break;
381 }
382 }
383
384 protected:
385 static void process_call_subr (op_code_t op, cs_type_t type,
386 cff1_cs_interp_env_t &env, subr_subset_param_t& param,
387 cff1_biased_subrs_t& subrs, hb_set_t *closure)
388 {
389 byte_str_ref_t str_ref = env.str_ref;
390 env.call_subr (subrs, type);
391 param.current_parsed_str->add_call_op (op, str_ref, env.context.subr_num);
392 closure->add (env.context.subr_num);
393 param.set_current_str (env, true);
394 }
395
396 private:
397 typedef cff1_cs_opset_t<cff1_cs_opset_subr_subset_t, subr_subset_param_t> SUPER;
398};
399
400struct cff1_subr_subsetter_t : subr_subsetter_t<cff1_subr_subsetter_t, CFF1Subrs, const OT::cff1::accelerator_subset_t, cff1_cs_interp_env_t, cff1_cs_opset_subr_subset_t, OpCode_endchar>
401{
402 cff1_subr_subsetter_t (const OT::cff1::accelerator_subset_t &acc_, const hb_subset_plan_t *plan_)
403 : subr_subsetter_t (acc_, plan_) {}
404
405 static void finalize_parsed_str (cff1_cs_interp_env_t &env, subr_subset_param_t& param, parsed_cs_str_t &charstring)
406 {
407 /* insert width at the beginning of the charstring as necessary */
408 if (env.has_width)
409 charstring.set_prefix (env.width);
410
411 /* subroutines/charstring left on the call stack are legally left unmarked
412 * unmarked when a subroutine terminates with endchar. mark them.
413 */
414 param.current_parsed_str->set_parsed ();
415 for (unsigned int i = 0; i < env.callStack.get_count (); i++)
416 {
417 parsed_cs_str_t *parsed_str = param.get_parsed_str_for_context (env.callStack[i]);
418 if (likely (parsed_str != nullptr))
419 parsed_str->set_parsed ();
420 else
421 env.set_error ();
422 }
423 }
424};
425
426struct cff_subset_plan {
427 cff_subset_plan ()
428 : final_size (0),
429 offsets (),
430 orig_fdcount (0),
431 subset_fdcount (1),
432 subset_fdselect_format (0),
433 drop_hints (false),
434 desubroutinize(false)
435 {
436 topdict_sizes.init ();
437 topdict_sizes.resize (1);
438 topdict_mod.init ();
439 subset_fdselect_ranges.init ();
440 fdmap.init ();
441 subset_charstrings.init ();
442 subset_globalsubrs.init ();
443 subset_localsubrs.init ();
444 fontdicts_mod.init ();
445 subset_enc_code_ranges.init ();
446 subset_enc_supp_codes.init ();
447 subset_charset_ranges.init ();
448 sidmap.init ();
449 for (unsigned int i = 0; i < name_dict_values_t::ValCount; i++)
450 topDictModSIDs[i] = CFF_UNDEF_SID;
451 }
452
453 ~cff_subset_plan ()
454 {
455 topdict_sizes.fini ();
456 topdict_mod.fini ();
457 subset_fdselect_ranges.fini ();
458 fdmap.fini ();
459 subset_charstrings.fini_deep ();
460 subset_globalsubrs.fini_deep ();
461 subset_localsubrs.fini_deep ();
462 fontdicts_mod.fini ();
463 subset_enc_code_ranges.fini ();
464 subset_enc_supp_codes.fini ();
465 subset_charset_ranges.fini ();
466 sidmap.fini ();
467 }
468
469 unsigned int plan_subset_encoding (const OT::cff1::accelerator_subset_t &acc, hb_subset_plan_t *plan)
470 {
471 const Encoding *encoding = acc.encoding;
472 unsigned int size0, size1, supp_size;
473 hb_codepoint_t code, last_code = CFF_UNDEF_CODE;
474 hb_vector_t<hb_codepoint_t> supp_codes;
475
476 subset_enc_code_ranges.resize (0);
477 supp_size = 0;
478 supp_codes.init ();
479
480 subset_enc_num_codes = plan->num_output_glyphs () - 1;
481 unsigned int glyph;
482 for (glyph = 1; glyph < plan->num_output_glyphs (); glyph++)
483 {
484 hb_codepoint_t old_glyph;
485 if (!plan->old_gid_for_new_gid (glyph, &old_glyph))
486 {
487 /* Retain the code for the old missing glyph ID */
488 old_glyph = glyph;
489 }
490 code = acc.glyph_to_code (old_glyph);
491 if (code == CFF_UNDEF_CODE)
492 {
493 subset_enc_num_codes = glyph - 1;
494 break;
495 }
496
497 if ((last_code == CFF_UNDEF_CODE) || (code != last_code + 1))
498 {
499 code_pair_t pair = { code, glyph };
500 subset_enc_code_ranges.push (pair);
501 }
502 last_code = code;
503
504 if (encoding != &Null(Encoding))
505 {
506 hb_codepoint_t sid = acc.glyph_to_sid (old_glyph);
507 encoding->get_supplement_codes (sid, supp_codes);
508 for (unsigned int i = 0; i < supp_codes.length; i++)
509 {
510 code_pair_t pair = { supp_codes[i], sid };
511 subset_enc_supp_codes.push (pair);
512 }
513 supp_size += SuppEncoding::static_size * supp_codes.length;
514 }
515 }
516 supp_codes.fini ();
517
518 subset_enc_code_ranges.finalize (glyph);
519
520 assert (subset_enc_num_codes <= 0xFF);
521 size0 = Encoding0::min_size + HBUINT8::static_size * subset_enc_num_codes;
522 size1 = Encoding1::min_size + Encoding1_Range::static_size * subset_enc_code_ranges.length;
523
524 if (size0 < size1)
525 subset_enc_format = 0;
526 else
527 subset_enc_format = 1;
528
529 return Encoding::calculate_serialized_size (
530 subset_enc_format,
531 subset_enc_format? subset_enc_code_ranges.length: subset_enc_num_codes,
532 subset_enc_supp_codes.length);
533 }
534
535 unsigned int plan_subset_charset (const OT::cff1::accelerator_subset_t &acc, hb_subset_plan_t *plan)
536 {
537 unsigned int size0, size_ranges;
538 hb_codepoint_t sid, last_sid = CFF_UNDEF_CODE;
539
540 subset_charset_ranges.resize (0);
541 unsigned int glyph;
542 for (glyph = 1; glyph < plan->num_output_glyphs (); glyph++)
543 {
544 hb_codepoint_t old_glyph;
545 if (!plan->old_gid_for_new_gid (glyph, &old_glyph))
546 {
547 /* Retain the SID for the old missing glyph ID */
548 old_glyph = glyph;
549 }
550 sid = acc.glyph_to_sid (old_glyph);
551
552 if (!acc.is_CID ())
553 sid = sidmap.add (sid);
554
555 if ((last_sid == CFF_UNDEF_CODE) || (sid != last_sid + 1))
556 {
557 code_pair_t pair = { sid, glyph };
558 subset_charset_ranges.push (pair);
559 }
560 last_sid = sid;
561 }
562
563 bool two_byte = subset_charset_ranges.finalize (glyph);
564
565 size0 = Charset0::min_size + HBUINT16::static_size * (plan->num_output_glyphs () - 1);
566 if (!two_byte)
567 size_ranges = Charset1::min_size + Charset1_Range::static_size * subset_charset_ranges.length;
568 else
569 size_ranges = Charset2::min_size + Charset2_Range::static_size * subset_charset_ranges.length;
570
571 if (size0 < size_ranges)
572 subset_charset_format = 0;
573 else if (!two_byte)
574 subset_charset_format = 1;
575 else
576 subset_charset_format = 2;
577
578 return Charset::calculate_serialized_size (
579 subset_charset_format,
580 subset_charset_format? subset_charset_ranges.length: plan->num_output_glyphs ());
581 }
582
583 bool collect_sids_in_dicts (const OT::cff1::accelerator_subset_t &acc)
584 {
585 sidmap.reset ();
586
587 for (unsigned int i = 0; i < name_dict_values_t::ValCount; i++)
588 {
589 unsigned int sid = acc.topDict.nameSIDs[i];
590 if (sid != CFF_UNDEF_SID)
591 {
592 (void)sidmap.add (sid);
593 topDictModSIDs[i] = sidmap[sid];
594 }
595 }
596
597 if (acc.fdArray != &Null(CFF1FDArray))
598 for (unsigned int i = 0; i < orig_fdcount; i++)
599 if (fdmap.has (i))
600 (void)sidmap.add (acc.fontDicts[i].fontName);
601
602 return true;
603 }
604
605 bool create (const OT::cff1::accelerator_subset_t &acc,
606 hb_subset_plan_t *plan)
607 {
608 /* make sure notdef is first */
609 hb_codepoint_t old_glyph;
610 if (!plan->old_gid_for_new_gid (0, &old_glyph) || (old_glyph != 0)) return false;
611
612 final_size = 0;
613 num_glyphs = plan->num_output_glyphs ();
614 orig_fdcount = acc.fdCount;
615 drop_hints = plan->drop_hints;
616 desubroutinize = plan->desubroutinize;
617
618 /* check whether the subset renumbers any glyph IDs */
619 gid_renum = false;
620 for (hb_codepoint_t new_glyph = 0; new_glyph < plan->num_output_glyphs (); new_glyph++)
621 {
622 if (!plan->old_gid_for_new_gid(new_glyph, &old_glyph))
623 continue;
624 if (new_glyph != old_glyph) {
625 gid_renum = true;
626 break;
627 }
628 }
629
630 subset_charset = gid_renum || !acc.is_predef_charset ();
631 subset_encoding = !acc.is_CID() && !acc.is_predef_encoding ();
632
633 /* CFF header */
634 final_size += OT::cff1::static_size;
635
636 /* Name INDEX */
637 offsets.nameIndexOffset = final_size;
638 final_size += acc.nameIndex->get_size ();
639
640 /* top dict INDEX */
641 {
642 /* Add encoding/charset to a (copy of) top dict as necessary */
643 topdict_mod.init (&acc.topDict);
644 bool need_to_add_enc = (subset_encoding && !acc.topDict.has_op (OpCode_Encoding));
645 bool need_to_add_set = (subset_charset && !acc.topDict.has_op (OpCode_charset));
646 if (need_to_add_enc || need_to_add_set)
647 {
648 if (need_to_add_enc)
649 topdict_mod.add_op (OpCode_Encoding);
650 if (need_to_add_set)
651 topdict_mod.add_op (OpCode_charset);
652 }
653 offsets.topDictInfo.offset = final_size;
654 cff1_top_dict_op_serializer_t topSzr;
655 unsigned int topDictSize = TopDict::calculate_serialized_size (topdict_mod, topSzr);
656 offsets.topDictInfo.offSize = calcOffSize(topDictSize);
657 if (unlikely (offsets.topDictInfo.offSize > 4))
658 return false;
659 final_size += CFF1IndexOf<TopDict>::calculate_serialized_size<cff1_top_dict_values_mod_t>
660 (offsets.topDictInfo.offSize,
661 &topdict_mod, 1, topdict_sizes, topSzr);
662 }
663
664 /* Determine re-mapping of font index as fdmap among other info */
665 if (acc.fdSelect != &Null(CFF1FDSelect))
666 {
667 if (unlikely (!hb_plan_subset_cff_fdselect (plan,
668 orig_fdcount,
669 *acc.fdSelect,
670 subset_fdcount,
671 offsets.FDSelectInfo.size,
672 subset_fdselect_format,
673 subset_fdselect_ranges,
674 fdmap)))
675 return false;
676 }
677 else
678 fdmap.identity (1);
679
680 /* remove unused SIDs & reassign SIDs */
681 {
682 /* SIDs for name strings in dicts are added before glyph names so they fit in 16-bit int range */
683 if (unlikely (!collect_sids_in_dicts (acc)))
684 return false;
685 if (unlikely (sidmap.get_population () > 0x8000)) /* assumption: a dict won't reference that many strings */
686 return false;
687 if (subset_charset)
688 offsets.charsetInfo.size = plan_subset_charset (acc, plan);
689
690 topdict_mod.reassignSIDs (sidmap);
691 }
692
693 /* String INDEX */
694 {
695 offsets.stringIndexInfo.offset = final_size;
696 offsets.stringIndexInfo.size = acc.stringIndex->calculate_serialized_size (offsets.stringIndexInfo.offSize, sidmap);
697 final_size += offsets.stringIndexInfo.size;
698 }
699
700 if (desubroutinize)
701 {
702 /* Flatten global & local subrs */
703 subr_flattener_t<const OT::cff1::accelerator_subset_t, cff1_cs_interp_env_t, cff1_cs_opset_flatten_t, OpCode_endchar>
704 flattener(acc, plan);
705 if (!flattener.flatten (subset_charstrings))
706 return false;
707
708 /* no global/local subroutines */
709 offsets.globalSubrsInfo.size = CFF1Subrs::calculate_serialized_size (1, 0, 0);
710 }
711 else
712 {
713 cff1_subr_subsetter_t subr_subsetter (acc, plan);
714
715 /* Subset subrs: collect used subroutines, leaving all unused ones behind */
716 if (!subr_subsetter.subset ())
717 return false;
718
719 /* encode charstrings, global subrs, local subrs with new subroutine numbers */
720 if (!subr_subsetter.encode_charstrings (subset_charstrings))
721 return false;
722
723 if (!subr_subsetter.encode_globalsubrs (subset_globalsubrs))
724 return false;
725
726 /* global subrs */
727 unsigned int dataSize = subset_globalsubrs.total_size ();
728 offsets.globalSubrsInfo.offSize = calcOffSize (dataSize);
729 if (unlikely (offsets.globalSubrsInfo.offSize > 4))
730 return false;
731 offsets.globalSubrsInfo.size = CFF1Subrs::calculate_serialized_size (offsets.globalSubrsInfo.offSize, subset_globalsubrs.length, dataSize);
732
733 /* local subrs */
734 if (!offsets.localSubrsInfos.resize (orig_fdcount))
735 return false;
736 if (!subset_localsubrs.resize (orig_fdcount))
737 return false;
738 for (unsigned int fd = 0; fd < orig_fdcount; fd++)
739 {
740 subset_localsubrs[fd].init ();
741 offsets.localSubrsInfos[fd].init ();
742 if (fdmap.has (fd))
743 {
744 if (!subr_subsetter.encode_localsubrs (fd, subset_localsubrs[fd]))
745 return false;
746
747 unsigned int dataSize = subset_localsubrs[fd].total_size ();
748 if (dataSize > 0)
749 {
750 offsets.localSubrsInfos[fd].offset = final_size;
751 offsets.localSubrsInfos[fd].offSize = calcOffSize (dataSize);
752 if (unlikely (offsets.localSubrsInfos[fd].offSize > 4))
753 return false;
754 offsets.localSubrsInfos[fd].size = CFF1Subrs::calculate_serialized_size (offsets.localSubrsInfos[fd].offSize, subset_localsubrs[fd].length, dataSize);
755 }
756 }
757 }
758 }
759
760 /* global subrs */
761 offsets.globalSubrsInfo.offset = final_size;
762 final_size += offsets.globalSubrsInfo.size;
763
764 /* Encoding */
765 if (!subset_encoding)
766 offsets.encodingOffset = acc.topDict.EncodingOffset;
767 else
768 {
769 offsets.encodingOffset = final_size;
770 final_size += plan_subset_encoding (acc, plan);
771 }
772
773 /* Charset */
774 if (!subset_charset && acc.is_predef_charset ())
775 offsets.charsetInfo.offset = acc.topDict.CharsetOffset;
776 else
777 offsets.charsetInfo.offset = final_size;
778 final_size += offsets.charsetInfo.size;
779
780 /* FDSelect */
781 if (acc.fdSelect != &Null(CFF1FDSelect))
782 {
783 offsets.FDSelectInfo.offset = final_size;
784 final_size += offsets.FDSelectInfo.size;
785 }
786
787 /* FDArray (FDIndex) */
788 if (acc.fdArray != &Null(CFF1FDArray)) {
789 offsets.FDArrayInfo.offset = final_size;
790 cff1_font_dict_op_serializer_t fontSzr;
791 unsigned int dictsSize = 0;
792 for (unsigned int i = 0; i < acc.fontDicts.length; i++)
793 if (fdmap.has (i))
794 dictsSize += FontDict::calculate_serialized_size (acc.fontDicts[i], fontSzr);
795
796 offsets.FDArrayInfo.offSize = calcOffSize (dictsSize);
797 if (unlikely (offsets.FDArrayInfo.offSize > 4))
798 return false;
799 final_size += CFF1Index::calculate_serialized_size (offsets.FDArrayInfo.offSize, subset_fdcount, dictsSize);
800 }
801
802 /* CharStrings */
803 {
804 offsets.charStringsInfo.offset = final_size;
805 unsigned int dataSize = subset_charstrings.total_size ();
806 offsets.charStringsInfo.offSize = calcOffSize (dataSize);
807 if (unlikely (offsets.charStringsInfo.offSize > 4))
808 return false;
809 final_size += CFF1CharStrings::calculate_serialized_size (offsets.charStringsInfo.offSize, plan->num_output_glyphs (), dataSize);
810 }
811
812 /* private dicts & local subrs */
813 offsets.privateDictInfo.offset = final_size;
814 for (unsigned int i = 0; i < orig_fdcount; i++)
815 {
816 if (fdmap.has (i))
817 {
818 bool has_localsubrs = offsets.localSubrsInfos[i].size > 0;
819 cff_private_dict_op_serializer_t privSzr (desubroutinize, plan->drop_hints);
820 unsigned int priv_size = PrivateDict::calculate_serialized_size (acc.privateDicts[i], privSzr, has_localsubrs);
821 table_info_t privInfo = { final_size, priv_size, 0 };
822 font_dict_values_mod_t fontdict_mod;
823 if (!acc.is_CID ())
824 fontdict_mod.init ( &Null(cff1_font_dict_values_t), CFF_UNDEF_SID, privInfo );
825 else
826 fontdict_mod.init ( &acc.fontDicts[i], sidmap[acc.fontDicts[i].fontName], privInfo );
827 fontdicts_mod.push (fontdict_mod);
828 final_size += privInfo.size;
829
830 if (!plan->desubroutinize && has_localsubrs)
831 {
832 offsets.localSubrsInfos[i].offset = final_size;
833 final_size += offsets.localSubrsInfos[i].size;
834 }
835 }
836 }
837
838 if (!acc.is_CID ())
839 offsets.privateDictInfo = fontdicts_mod[0].privateDictInfo;
840
841 return ((subset_charstrings.length == plan->num_output_glyphs ())
842 && (fontdicts_mod.length == subset_fdcount));
843 }
844
845 unsigned int get_final_size () const { return final_size; }
846
847 unsigned int final_size;
848 hb_vector_t<unsigned int> topdict_sizes;
849 cff1_top_dict_values_mod_t topdict_mod;
850 cff1_sub_table_offsets_t offsets;
851
852 unsigned int num_glyphs;
853 unsigned int orig_fdcount;
854 unsigned int subset_fdcount;
855 unsigned int subset_fdselect_format;
856 hb_vector_t<code_pair_t> subset_fdselect_ranges;
857
858 /* font dict index remap table from fullset FDArray to subset FDArray.
859 * set to CFF_UNDEF_CODE if excluded from subset */
860 hb_inc_bimap_t fdmap;
861
862 str_buff_vec_t subset_charstrings;
863 str_buff_vec_t subset_globalsubrs;
864 hb_vector_t<str_buff_vec_t> subset_localsubrs;
865 hb_vector_t<font_dict_values_mod_t> fontdicts_mod;
866
867 bool drop_hints;
868
869 bool gid_renum;
870 bool subset_encoding;
871 uint8_t subset_enc_format;
872 unsigned int subset_enc_num_codes;
873 range_list_t subset_enc_code_ranges;
874 hb_vector_t<code_pair_t> subset_enc_supp_codes;
875
876 uint8_t subset_charset_format;
877 range_list_t subset_charset_ranges;
878 bool subset_charset;
879
880 remap_sid_t sidmap;
881 unsigned int topDictModSIDs[name_dict_values_t::ValCount];
882
883 bool desubroutinize;
884};
885
886static inline bool _write_cff1 (const cff_subset_plan &plan,
887 const OT::cff1::accelerator_subset_t &acc,
888 unsigned int num_glyphs,
889 unsigned int dest_sz,
890 void *dest)
891{
892 hb_serialize_context_t c (dest, dest_sz);
893
894 OT::cff1 *cff = c.start_serialize<OT::cff1> ();
895 if (unlikely (!c.extend_min (*cff)))
896 return false;
897
898 /* header */
899 cff->version.major = 0x01;
900 cff->version.minor = 0x00;
901 cff->nameIndex = cff->min_size;
902 cff->offSize = 4; /* unused? */
903
904 /* name INDEX */
905 {
906 assert (cff->nameIndex == (unsigned) (c.head - c.start));
907 CFF1NameIndex *dest = c.start_embed<CFF1NameIndex> ();
908 if (unlikely (dest == nullptr)) return false;
909 if (unlikely (!dest->serialize (&c, *acc.nameIndex)))
910 {
911 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF name INDEX");
912 return false;
913 }
914 }
915
916 /* top dict INDEX */
917 {
918 assert (plan.offsets.topDictInfo.offset == (unsigned) (c.head - c.start));
919 CFF1IndexOf<TopDict> *dest = c.start_embed< CFF1IndexOf<TopDict>> ();
920 if (dest == nullptr) return false;
921 cff1_top_dict_op_serializer_t topSzr;
922 top_dict_modifiers_t modifier (plan.offsets, plan.topDictModSIDs);
923 if (unlikely (!dest->serialize (&c, plan.offsets.topDictInfo.offSize,
924 &plan.topdict_mod, 1,
925 plan.topdict_sizes, topSzr, modifier)))
926 {
927 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF top dict");
928 return false;
929 }
930 }
931
932 /* String INDEX */
933 {
934 assert (plan.offsets.stringIndexInfo.offset == (unsigned) (c.head - c.start));
935 CFF1StringIndex *dest = c.start_embed<CFF1StringIndex> ();
936 if (unlikely (dest == nullptr)) return false;
937 if (unlikely (!dest->serialize (&c, *acc.stringIndex, plan.offsets.stringIndexInfo.offSize, plan.sidmap)))
938 {
939 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF string INDEX");
940 return false;
941 }
942 }
943
944 /* global subrs */
945 {
946 assert (plan.offsets.globalSubrsInfo.offset != 0);
947 assert (plan.offsets.globalSubrsInfo.offset == (unsigned) (c.head - c.start));
948
949 CFF1Subrs *dest = c.start_embed <CFF1Subrs> ();
950 if (unlikely (dest == nullptr)) return false;
951 if (unlikely (!dest->serialize (&c, plan.offsets.globalSubrsInfo.offSize, plan.subset_globalsubrs)))
952 {
953 DEBUG_MSG (SUBSET, nullptr, "failed to serialize global subroutines");
954 return false;
955 }
956 }
957
958 /* Encoding */
959 if (plan.subset_encoding)
960 {
961 assert (plan.offsets.encodingOffset == (unsigned) (c.head - c.start));
962 Encoding *dest = c.start_embed<Encoding> ();
963 if (unlikely (dest == nullptr)) return false;
964 if (unlikely (!dest->serialize (&c,
965 plan.subset_enc_format,
966 plan.subset_enc_num_codes,
967 plan.subset_enc_code_ranges,
968 plan.subset_enc_supp_codes)))
969 {
970 DEBUG_MSG (SUBSET, nullptr, "failed to serialize Encoding");
971 return false;
972 }
973 }
974
975 /* Charset */
976 if (plan.subset_charset)
977 {
978 assert (plan.offsets.charsetInfo.offset == (unsigned) (c.head - c.start));
979 Charset *dest = c.start_embed<Charset> ();
980 if (unlikely (dest == nullptr)) return false;
981 if (unlikely (!dest->serialize (&c,
982 plan.subset_charset_format,
983 plan.num_glyphs,
984 plan.subset_charset_ranges)))
985 {
986 DEBUG_MSG (SUBSET, nullptr, "failed to serialize Charset");
987 return false;
988 }
989 }
990
991 /* FDSelect */
992 if (acc.fdSelect != &Null(CFF1FDSelect))
993 {
994 assert (plan.offsets.FDSelectInfo.offset == (unsigned) (c.head - c.start));
995
996 if (unlikely (!hb_serialize_cff_fdselect (&c, num_glyphs, *acc.fdSelect, acc.fdCount,
997 plan.subset_fdselect_format, plan.offsets.FDSelectInfo.size,
998 plan.subset_fdselect_ranges)))
999 {
1000 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF subset FDSelect");
1001 return false;
1002 }
1003 }
1004
1005 /* FDArray (FD Index) */
1006 if (acc.fdArray != &Null(CFF1FDArray))
1007 {
1008 assert (plan.offsets.FDArrayInfo.offset == (unsigned) (c.head - c.start));
1009 CFF1FDArray *fda = c.start_embed<CFF1FDArray> ();
1010 if (unlikely (fda == nullptr)) return false;
1011 cff1_font_dict_op_serializer_t fontSzr;
1012 if (unlikely (!fda->serialize (&c, plan.offsets.FDArrayInfo.offSize,
1013 plan.fontdicts_mod,
1014 fontSzr)))
1015 {
1016 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF FDArray");
1017 return false;
1018 }
1019 }
1020
1021 /* CharStrings */
1022 {
1023 assert (plan.offsets.charStringsInfo.offset == (unsigned) (c.head - c.start));
1024 CFF1CharStrings *cs = c.start_embed<CFF1CharStrings> ();
1025 if (unlikely (cs == nullptr)) return false;
1026 if (unlikely (!cs->serialize (&c, plan.offsets.charStringsInfo.offSize, plan.subset_charstrings)))
1027 {
1028 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF CharStrings");
1029 return false;
1030 }
1031 }
1032
1033 /* private dicts & local subrs */
1034 assert (plan.offsets.privateDictInfo.offset == (unsigned) (c.head - c.start));
1035 for (unsigned int i = 0; i < acc.privateDicts.length; i++)
1036 {
1037 if (plan.fdmap.has (i))
1038 {
1039 PrivateDict *pd = c.start_embed<PrivateDict> ();
1040 if (unlikely (pd == nullptr)) return false;
1041 unsigned int priv_size = plan.fontdicts_mod[plan.fdmap[i]].privateDictInfo.size;
1042 bool result;
1043 cff_private_dict_op_serializer_t privSzr (plan.desubroutinize, plan.drop_hints);
1044 /* N.B. local subrs immediately follows its corresponding private dict. i.e., subr offset == private dict size */
1045 unsigned int subroffset = (plan.offsets.localSubrsInfos[i].size > 0) ? priv_size : 0;
1046 result = pd->serialize (&c, acc.privateDicts[i], privSzr, subroffset);
1047 if (unlikely (!result))
1048 {
1049 DEBUG_MSG (SUBSET, nullptr, "failed to serialize CFF Private Dict[%d]", i);
1050 return false;
1051 }
1052 if (plan.offsets.localSubrsInfos[i].size > 0)
1053 {
1054 CFF1Subrs *dest = c.start_embed <CFF1Subrs> ();
1055 if (unlikely (dest == nullptr)) return false;
1056 if (unlikely (!dest->serialize (&c, plan.offsets.localSubrsInfos[i].offSize, plan.subset_localsubrs[i])))
1057 {
1058 DEBUG_MSG (SUBSET, nullptr, "failed to serialize local subroutines");
1059 return false;
1060 }
1061 }
1062 }
1063 }
1064
1065 assert (c.head == c.end);
1066 c.end_serialize ();
1067
1068 return true;
1069}
1070
1071static inline bool
1072_hb_subset_cff1 (const OT::cff1::accelerator_subset_t &acc,
1073 const char *data,
1074 hb_subset_plan_t *plan,
1075 hb_blob_t **prime /* OUT */)
1076{
1077 cff_subset_plan cff_plan;
1078
1079 if (unlikely (!cff_plan.create (acc, plan)))
1080 {
1081 DEBUG_MSG(SUBSET, nullptr, "Failed to generate a cff subsetting plan.");
1082 return false;
1083 }
1084
1085 unsigned int cff_prime_size = cff_plan.get_final_size ();
1086 char *cff_prime_data = (char *) calloc (1, cff_prime_size);
1087
1088 if (unlikely (!_write_cff1 (cff_plan, acc, plan->num_output_glyphs (),
1089 cff_prime_size, cff_prime_data))) {
1090 DEBUG_MSG(SUBSET, nullptr, "Failed to write a subset cff.");
1091 free (cff_prime_data);
1092 return false;
1093 }
1094
1095 *prime = hb_blob_create (cff_prime_data,
1096 cff_prime_size,
1097 HB_MEMORY_MODE_READONLY,
1098 cff_prime_data,
1099 free);
1100 return true;
1101}
1102
1103/**
1104 * hb_subset_cff1:
1105 * Subsets the CFF table according to a provided plan.
1106 *
1107 * Return value: subsetted cff table.
1108 **/
1109bool
1110hb_subset_cff1 (hb_subset_plan_t *plan,
1111 hb_blob_t **prime /* OUT */)
1112{
1113 hb_blob_t *cff_blob = hb_sanitize_context_t().reference_table<CFF::cff1> (plan->source);
1114 const char *data = hb_blob_get_data(cff_blob, nullptr);
1115
1116 OT::cff1::accelerator_subset_t acc;
1117 acc.init(plan->source);
1118 bool result = likely (acc.is_valid ()) &&
1119 _hb_subset_cff1 (acc, data, plan, prime);
1120 hb_blob_destroy (cff_blob);
1121 acc.fini ();
1122
1123 return result;
1124}
1125
1126
1127#endif
1128