2010-10-06 00:36:58 +02:00
|
|
|
/*
|
2012-08-12 00:20:28 +02:00
|
|
|
* Copyright © 2010,2012 Google, Inc.
|
2010-10-06 00:36:58 +02:00
|
|
|
*
|
|
|
|
* This is part of HarfBuzz, a text shaping library.
|
|
|
|
*
|
|
|
|
* Permission is hereby granted, without written agreement and without
|
|
|
|
* license or royalty fees, to use, copy, modify, and distribute this
|
|
|
|
* software and its documentation for any purpose, provided that the
|
|
|
|
* above copyright notice and the following two paragraphs appear in
|
|
|
|
* all copies of this software.
|
|
|
|
*
|
|
|
|
* IN NO EVENT SHALL THE COPYRIGHT HOLDER BE LIABLE TO ANY PARTY FOR
|
|
|
|
* DIRECT, INDIRECT, SPECIAL, INCIDENTAL, OR CONSEQUENTIAL DAMAGES
|
|
|
|
* ARISING OUT OF THE USE OF THIS SOFTWARE AND ITS DOCUMENTATION, EVEN
|
|
|
|
* IF THE COPYRIGHT HOLDER HAS BEEN ADVISED OF THE POSSIBILITY OF SUCH
|
|
|
|
* DAMAGE.
|
|
|
|
*
|
|
|
|
* THE COPYRIGHT HOLDER SPECIFICALLY DISCLAIMS ANY WARRANTIES, INCLUDING,
|
|
|
|
* BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE. THE SOFTWARE PROVIDED HEREUNDER IS
|
|
|
|
* ON AN "AS IS" BASIS, AND THE COPYRIGHT HOLDER HAS NO OBLIGATION TO
|
|
|
|
* PROVIDE MAINTENANCE, SUPPORT, UPDATES, ENHANCEMENTS, OR MODIFICATIONS.
|
|
|
|
*
|
|
|
|
* Google Author(s): Behdad Esfahbod
|
|
|
|
*/
|
|
|
|
|
2018-08-26 07:36:36 +02:00
|
|
|
#include "hb.hh"
|
|
|
|
#include "hb-ot-shape-complex-arabic.hh"
|
|
|
|
#include "hb-ot-shape.hh"
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2010-11-03 21:37:24 +01:00
|
|
|
|
|
|
|
/* buffer var allocations */
|
2012-08-10 03:48:55 +02:00
|
|
|
#define arabic_shaping_action() complex_var_u8_0() /* arabic shaping action */
|
2010-11-03 21:37:24 +01:00
|
|
|
|
2015-11-06 02:29:03 +01:00
|
|
|
#define HB_BUFFER_SCRATCH_FLAG_ARABIC_HAS_STCH HB_BUFFER_SCRATCH_FLAG_COMPLEX0
|
|
|
|
|
2015-11-07 01:27:44 +01:00
|
|
|
/* See:
|
2017-11-20 20:49:22 +01:00
|
|
|
* https://github.com/harfbuzz/harfbuzz/commit/6e6f82b6f3dde0fc6c3c7d991d9ec6cfff57823d#commitcomment-14248516 */
|
2015-11-07 01:27:44 +01:00
|
|
|
#define HB_ARABIC_GENERAL_CATEGORY_IS_WORD(gen_cat) \
|
2017-10-19 20:39:52 +02:00
|
|
|
(FLAG_UNSAFE (gen_cat) & \
|
2015-11-07 01:27:44 +01:00
|
|
|
(FLAG (HB_UNICODE_GENERAL_CATEGORY_UNASSIGNED) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_PRIVATE_USE) | \
|
|
|
|
/*FLAG (HB_UNICODE_GENERAL_CATEGORY_LOWERCASE_LETTER) |*/ \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_MODIFIER_LETTER) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_OTHER_LETTER) | \
|
|
|
|
/*FLAG (HB_UNICODE_GENERAL_CATEGORY_TITLECASE_LETTER) |*/ \
|
|
|
|
/*FLAG (HB_UNICODE_GENERAL_CATEGORY_UPPERCASE_LETTER) |*/ \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_SPACING_MARK) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_DECIMAL_NUMBER) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_LETTER_NUMBER) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_OTHER_NUMBER) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_CURRENCY_SYMBOL) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_MODIFIER_SYMBOL) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_MATH_SYMBOL) | \
|
|
|
|
FLAG (HB_UNICODE_GENERAL_CATEGORY_OTHER_SYMBOL)))
|
|
|
|
|
2010-11-03 21:37:24 +01:00
|
|
|
|
2015-07-22 11:32:30 +02:00
|
|
|
/*
|
|
|
|
* Joining types:
|
|
|
|
*/
|
|
|
|
|
2010-10-06 00:36:58 +02:00
|
|
|
/*
|
|
|
|
* Bits used in the joining tables
|
|
|
|
*/
|
2015-07-22 11:32:30 +02:00
|
|
|
enum hb_arabic_joining_type_t {
|
2010-10-06 00:36:58 +02:00
|
|
|
JOINING_TYPE_U = 0,
|
2013-02-15 15:27:02 +01:00
|
|
|
JOINING_TYPE_L = 1,
|
|
|
|
JOINING_TYPE_R = 2,
|
|
|
|
JOINING_TYPE_D = 3,
|
2010-10-06 00:36:58 +02:00
|
|
|
JOINING_TYPE_C = JOINING_TYPE_D,
|
2013-02-15 15:27:02 +01:00
|
|
|
JOINING_GROUP_ALAPH = 4,
|
|
|
|
JOINING_GROUP_DALATH_RISH = 5,
|
|
|
|
NUM_STATE_MACHINE_COLS = 6,
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2013-02-15 15:27:02 +01:00
|
|
|
JOINING_TYPE_T = 7,
|
|
|
|
JOINING_TYPE_X = 8 /* means: use general-category to choose between U or T. */
|
2010-10-06 00:36:58 +02:00
|
|
|
};
|
|
|
|
|
2011-04-21 00:50:27 +02:00
|
|
|
#include "hb-ot-shape-complex-arabic-table.hh"
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2011-04-11 21:47:40 +02:00
|
|
|
static unsigned int get_joining_type (hb_codepoint_t u, hb_unicode_general_category_t gen_cat)
|
2010-10-06 00:36:58 +02:00
|
|
|
{
|
2014-06-20 23:59:43 +02:00
|
|
|
unsigned int j_type = joining_type(u);
|
|
|
|
if (likely (j_type != JOINING_TYPE_X))
|
|
|
|
return j_type;
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2017-10-19 20:39:52 +02:00
|
|
|
return (FLAG_UNSAFE(gen_cat) &
|
2014-06-21 22:53:21 +02:00
|
|
|
(FLAG(HB_UNICODE_GENERAL_CATEGORY_NON_SPACING_MARK) |
|
|
|
|
FLAG(HB_UNICODE_GENERAL_CATEGORY_ENCLOSING_MARK) |
|
|
|
|
FLAG(HB_UNICODE_GENERAL_CATEGORY_FORMAT))
|
|
|
|
) ? JOINING_TYPE_T : JOINING_TYPE_U;
|
2010-10-06 00:36:58 +02:00
|
|
|
}
|
|
|
|
|
2014-07-17 19:36:09 +02:00
|
|
|
#define FEATURE_IS_SYRIAC(tag) hb_in_range<unsigned char> ((unsigned char) (tag), '2', '3')
|
|
|
|
|
2012-08-12 00:20:28 +02:00
|
|
|
static const hb_tag_t arabic_features[] =
|
2010-10-06 00:36:58 +02:00
|
|
|
{
|
|
|
|
HB_TAG('i','s','o','l'),
|
2014-07-17 19:36:09 +02:00
|
|
|
HB_TAG('f','i','n','a'),
|
2010-10-06 00:36:58 +02:00
|
|
|
HB_TAG('f','i','n','2'),
|
|
|
|
HB_TAG('f','i','n','3'),
|
2014-07-17 19:36:09 +02:00
|
|
|
HB_TAG('m','e','d','i'),
|
|
|
|
HB_TAG('m','e','d','2'),
|
|
|
|
HB_TAG('i','n','i','t'),
|
2010-10-06 00:36:58 +02:00
|
|
|
HB_TAG_NONE
|
|
|
|
};
|
|
|
|
|
|
|
|
|
|
|
|
/* Same order as the feature array */
|
2015-11-06 02:29:03 +01:00
|
|
|
enum arabic_action_t {
|
2010-10-06 00:36:58 +02:00
|
|
|
ISOL,
|
2014-07-17 19:36:09 +02:00
|
|
|
FINA,
|
2010-10-06 00:36:58 +02:00
|
|
|
FIN2,
|
|
|
|
FIN3,
|
2014-07-17 19:36:09 +02:00
|
|
|
MEDI,
|
|
|
|
MED2,
|
|
|
|
INIT,
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
NONE,
|
|
|
|
|
2015-11-06 02:29:03 +01:00
|
|
|
ARABIC_NUM_FEATURES = NONE,
|
|
|
|
|
|
|
|
/* We abuse the same byte for other things... */
|
|
|
|
STCH_FIXED,
|
|
|
|
STCH_REPEATING,
|
2010-10-06 00:36:58 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
static const struct arabic_state_table_entry {
|
|
|
|
uint8_t prev_action;
|
|
|
|
uint8_t curr_action;
|
2011-06-15 15:49:58 +02:00
|
|
|
uint16_t next_state;
|
2010-10-06 00:36:58 +02:00
|
|
|
} arabic_state_table[][NUM_STATE_MACHINE_COLS] =
|
|
|
|
{
|
2013-02-15 15:27:02 +01:00
|
|
|
/* jt_U, jt_L, jt_R, jt_D, jg_ALAPH, jg_DALATH_RISH */
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
/* State 0: prev was U, not willing to join. */
|
2013-02-15 15:27:02 +01:00
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
/* State 1: prev was R or ISOL/ALAPH, not willing to join. */
|
2013-02-15 15:27:02 +01:00
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN2,5}, {NONE,ISOL,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2013-02-15 15:27:02 +01:00
|
|
|
/* State 2: prev was D/L in ISOL form, willing to join. */
|
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {INIT,FINA,1}, {INIT,FINA,3}, {INIT,FINA,4}, {INIT,FINA,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2013-02-15 15:27:02 +01:00
|
|
|
/* State 3: prev was D in FINA form, willing to join. */
|
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {MEDI,FINA,1}, {MEDI,FINA,3}, {MEDI,FINA,4}, {MEDI,FINA,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
/* State 4: prev was FINA ALAPH, not willing to join. */
|
2013-02-15 15:27:02 +01:00
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {MED2,ISOL,1}, {MED2,ISOL,2}, {MED2,FIN2,5}, {MED2,ISOL,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
/* State 5: prev was FIN2/FIN3 ALAPH, not willing to join. */
|
2013-02-15 15:27:02 +01:00
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {ISOL,ISOL,1}, {ISOL,ISOL,2}, {ISOL,FIN2,5}, {ISOL,ISOL,6}, },
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
/* State 6: prev was DALATH/RISH, not willing to join. */
|
2013-02-15 15:27:02 +01:00
|
|
|
{ {NONE,NONE,0}, {NONE,ISOL,2}, {NONE,ISOL,1}, {NONE,ISOL,2}, {NONE,FIN3,5}, {NONE,ISOL,6}, }
|
2010-10-06 00:36:58 +02:00
|
|
|
};
|
|
|
|
|
|
|
|
|
2013-10-16 13:42:38 +02:00
|
|
|
static void
|
2018-09-27 00:42:37 +02:00
|
|
|
flip_joiners (const hb_ot_shape_plan_t *plan,
|
2013-10-16 13:42:38 +02:00
|
|
|
hb_font_t *font,
|
|
|
|
hb_buffer_t *buffer);
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
static void
|
|
|
|
arabic_fallback_shape (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_font_t *font,
|
|
|
|
hb_buffer_t *buffer);
|
2010-10-12 22:50:36 +02:00
|
|
|
|
2015-11-06 02:29:03 +01:00
|
|
|
static void
|
|
|
|
record_stch (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_font_t *font,
|
|
|
|
hb_buffer_t *buffer);
|
|
|
|
|
2012-07-31 03:08:51 +02:00
|
|
|
static void
|
2012-08-02 15:38:28 +02:00
|
|
|
collect_features_arabic (hb_ot_shape_planner_t *plan)
|
2010-10-12 22:50:36 +02:00
|
|
|
{
|
2012-08-02 15:38:28 +02:00
|
|
|
hb_ot_map_builder_t *map = &plan->map;
|
|
|
|
|
2014-07-17 19:36:09 +02:00
|
|
|
/* We apply features according to the Arabic spec, with pauses
|
|
|
|
* in between most.
|
2011-07-08 03:07:41 +02:00
|
|
|
*
|
2014-07-17 19:36:09 +02:00
|
|
|
* The pause between init/medi/... and rlig is required. See eg:
|
2011-07-08 03:07:41 +02:00
|
|
|
* https://bugzilla.mozilla.org/show_bug.cgi?id=644184
|
2013-10-28 00:20:59 +01:00
|
|
|
*
|
2014-07-17 19:36:09 +02:00
|
|
|
* The pauses between init/medi/... themselves are not necessarily
|
|
|
|
* needed as only one of those features is applied to any character.
|
|
|
|
* The only difference it makes is when fonts have contextual
|
|
|
|
* substitutions. We now follow the order of the spec, which makes
|
|
|
|
* for better experience if that's what Uniscribe is doing.
|
|
|
|
*
|
|
|
|
* At least for Arabic, looks like Uniscribe has a pause between
|
|
|
|
* rlig and calt. Otherwise the IranNastaliq's ALLAH ligature won't
|
|
|
|
* work. However, testing shows that rlig and calt are applied
|
|
|
|
* together for Mongolian in Uniscribe. As such, we only add a
|
|
|
|
* pause for Arabic, not other scripts.
|
2017-07-14 18:34:47 +02:00
|
|
|
*
|
|
|
|
* A pause after calt is required to make KFGQPC Uthmanic Script HAFS
|
2017-11-20 20:49:22 +01:00
|
|
|
* work correctly. See https://github.com/harfbuzz/harfbuzz/issues/505
|
2011-07-08 03:07:41 +02:00
|
|
|
*/
|
|
|
|
|
2013-10-16 13:42:38 +02:00
|
|
|
|
2018-09-25 00:01:53 +02:00
|
|
|
map->enable_feature (HB_TAG('s','t','c','h'));
|
2015-11-06 02:29:03 +01:00
|
|
|
map->add_gsub_pause (record_stch);
|
|
|
|
|
2018-09-25 00:01:53 +02:00
|
|
|
map->enable_feature (HB_TAG('c','c','m','p'));
|
|
|
|
map->enable_feature (HB_TAG('l','o','c','l'));
|
2011-07-08 06:22:40 +02:00
|
|
|
|
2017-10-15 12:11:08 +02:00
|
|
|
map->add_gsub_pause (nullptr);
|
2011-04-06 20:04:56 +02:00
|
|
|
|
2012-08-12 00:20:28 +02:00
|
|
|
for (unsigned int i = 0; i < ARABIC_NUM_FEATURES; i++)
|
2014-07-17 19:36:09 +02:00
|
|
|
{
|
|
|
|
bool has_fallback = plan->props.script == HB_SCRIPT_ARABIC && !FEATURE_IS_SYRIAC (arabic_features[i]);
|
2018-09-25 00:01:53 +02:00
|
|
|
map->add_feature (arabic_features[i], has_fallback ? F_HAS_FALLBACK : F_NONE);
|
2017-10-15 12:11:08 +02:00
|
|
|
map->add_gsub_pause (nullptr);
|
2014-07-17 19:36:09 +02:00
|
|
|
}
|
2011-07-08 03:07:41 +02:00
|
|
|
|
2018-09-27 00:42:37 +02:00
|
|
|
/* Normally, Unicode says a ZWNJ means "don't ligate". In Arabic script
|
|
|
|
* however, it says a ZWJ should also mean "don't ligate". So we convert
|
|
|
|
* a ZWJ to a ZWNJ for GSUB. We want to revert it back to ZWJ before
|
|
|
|
* GPOS processing though. So we just flip their roles, and flip back
|
|
|
|
* later. Note that this makes a ZWNJ into ZWJ for GSUB stage, which
|
|
|
|
* means it would *not* break ligatures. But since ligatures around
|
|
|
|
* ZWNJ are rare, we don't care.
|
|
|
|
*
|
|
|
|
* Since we don't currently have a way to apply a pause before GPOS
|
|
|
|
* starts, let's just do this dance around a few required GUSB features. */
|
|
|
|
map->add_gsub_pause (flip_joiners);
|
|
|
|
|
2018-09-25 00:03:34 +02:00
|
|
|
map->add_feature (HB_TAG('r','l','i','g'), F_GLOBAL | F_HAS_FALLBACK);
|
2018-09-27 00:42:37 +02:00
|
|
|
|
2014-07-17 19:36:09 +02:00
|
|
|
if (plan->props.script == HB_SCRIPT_ARABIC)
|
|
|
|
map->add_gsub_pause (arabic_fallback_shape);
|
2011-07-08 03:07:41 +02:00
|
|
|
|
2017-07-14 18:34:47 +02:00
|
|
|
/* No pause after rclt. See 98460779bae19e4d64d29461ff154b3527bf8420. */
|
2018-09-25 00:01:53 +02:00
|
|
|
map->enable_feature (HB_TAG('r','c','l','t'));
|
|
|
|
map->enable_feature (HB_TAG('c','a','l','t'));
|
2018-09-27 00:42:37 +02:00
|
|
|
|
|
|
|
/* And undo here. */
|
|
|
|
map->add_gsub_pause (flip_joiners);
|
2011-07-08 03:07:41 +02:00
|
|
|
|
2014-07-17 21:57:37 +02:00
|
|
|
/* The spec includes 'cswh'. Earlier versions of Windows
|
|
|
|
* used to enable this by default, but testing suggests
|
|
|
|
* that Windows 8 and later do not enable it by default,
|
|
|
|
* and spec now says 'Off by default'.
|
|
|
|
* We disabled this in ae23c24c32.
|
|
|
|
* Note that IranNastaliq uses this feature extensively
|
|
|
|
* to fixup broken glyph sequences. Oh well...
|
|
|
|
* Test case: U+0643,U+0640,U+0631. */
|
2018-09-25 00:01:53 +02:00
|
|
|
//map->enable_feature (HB_TAG('c','s','w','h'));
|
|
|
|
map->enable_feature (HB_TAG('m','s','e','t'));
|
2010-10-12 22:50:36 +02:00
|
|
|
}
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
#include "hb-ot-shape-complex-arabic-fallback.hh"
|
|
|
|
|
2012-08-12 00:20:28 +02:00
|
|
|
struct arabic_shape_plan_t
|
|
|
|
{
|
|
|
|
ASSERT_POD ();
|
|
|
|
|
2012-08-16 17:46:46 +02:00
|
|
|
/* The "+ 1" in the next array is to accommodate for the "NONE" command,
|
|
|
|
* which is not an OpenType feature, but this simplifies the code by not
|
|
|
|
* having to do a "if (... < NONE) ..." and just rely on the fact that
|
|
|
|
* mask_array[NONE] == 0. */
|
2012-08-16 17:35:50 +02:00
|
|
|
hb_mask_t mask_array[ARABIC_NUM_FEATURES + 1];
|
2012-09-06 04:19:28 +02:00
|
|
|
|
2018-08-09 09:22:37 +02:00
|
|
|
hb_atomic_ptr_t<arabic_fallback_plan_t> fallback_plan;
|
2015-11-06 02:29:03 +01:00
|
|
|
|
|
|
|
unsigned int do_fallback : 1;
|
|
|
|
unsigned int has_stch : 1;
|
2012-08-12 00:20:28 +02:00
|
|
|
};
|
|
|
|
|
2015-07-22 11:32:30 +02:00
|
|
|
void *
|
2012-08-12 00:20:28 +02:00
|
|
|
data_create_arabic (const hb_ot_shape_plan_t *plan)
|
|
|
|
{
|
|
|
|
arabic_shape_plan_t *arabic_plan = (arabic_shape_plan_t *) calloc (1, sizeof (arabic_shape_plan_t));
|
|
|
|
if (unlikely (!arabic_plan))
|
2017-10-15 12:11:08 +02:00
|
|
|
return nullptr;
|
2012-08-12 00:20:28 +02:00
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
arabic_plan->do_fallback = plan->props.script == HB_SCRIPT_ARABIC;
|
2015-11-06 02:29:03 +01:00
|
|
|
arabic_plan->has_stch = !!plan->map.get_1_mask (HB_TAG ('s','t','c','h'));
|
2012-08-12 00:20:28 +02:00
|
|
|
for (unsigned int i = 0; i < ARABIC_NUM_FEATURES; i++) {
|
|
|
|
arabic_plan->mask_array[i] = plan->map.get_1_mask (arabic_features[i]);
|
2014-07-17 19:36:09 +02:00
|
|
|
arabic_plan->do_fallback = arabic_plan->do_fallback &&
|
2014-08-05 20:19:01 +02:00
|
|
|
(FEATURE_IS_SYRIAC (arabic_features[i]) ||
|
|
|
|
plan->map.needs_fallback (arabic_features[i]));
|
2012-08-12 00:20:28 +02:00
|
|
|
}
|
|
|
|
|
|
|
|
return arabic_plan;
|
|
|
|
}
|
|
|
|
|
2015-07-22 11:32:30 +02:00
|
|
|
void
|
2012-08-12 00:20:28 +02:00
|
|
|
data_destroy_arabic (void *data)
|
|
|
|
{
|
2012-09-06 04:19:28 +02:00
|
|
|
arabic_shape_plan_t *arabic_plan = (arabic_shape_plan_t *) data;
|
2012-04-10 23:20:05 +02:00
|
|
|
|
2018-08-09 09:22:37 +02:00
|
|
|
arabic_fallback_plan_destroy (arabic_plan->fallback_plan.get ());
|
2012-08-29 17:11:54 +02:00
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
free (data);
|
2012-04-10 22:44:38 +02:00
|
|
|
}
|
|
|
|
|
2012-07-31 03:08:51 +02:00
|
|
|
static void
|
2012-08-12 00:34:13 +02:00
|
|
|
arabic_joining (hb_buffer_t *buffer)
|
2010-10-06 00:36:58 +02:00
|
|
|
{
|
2011-07-08 04:25:25 +02:00
|
|
|
unsigned int count = buffer->len;
|
2014-07-17 20:22:11 +02:00
|
|
|
hb_glyph_info_t *info = buffer->info;
|
2012-09-26 03:32:35 +02:00
|
|
|
unsigned int prev = (unsigned int) -1, state = 0;
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2012-09-26 03:32:35 +02:00
|
|
|
/* Check pre-context */
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
for (unsigned int i = 0; i < buffer->context_len[0]; i++)
|
|
|
|
{
|
|
|
|
unsigned int this_type = get_joining_type (buffer->context[0][i], buffer->unicode->general_category (buffer->context[0][i]));
|
2012-09-26 03:32:35 +02:00
|
|
|
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
if (unlikely (this_type == JOINING_TYPE_T))
|
|
|
|
continue;
|
2012-09-26 03:32:35 +02:00
|
|
|
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
|
|
|
|
state = entry->next_state;
|
|
|
|
break;
|
|
|
|
}
|
2012-09-26 03:32:35 +02:00
|
|
|
|
2010-10-12 23:37:44 +02:00
|
|
|
for (unsigned int i = 0; i < count; i++)
|
|
|
|
{
|
2014-07-17 20:22:11 +02:00
|
|
|
unsigned int this_type = get_joining_type (info[i].codepoint, _hb_glyph_info_get_general_category (&info[i]));
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2010-10-27 16:40:39 +02:00
|
|
|
if (unlikely (this_type == JOINING_TYPE_T)) {
|
2014-07-17 20:22:11 +02:00
|
|
|
info[i].arabic_shaping_action() = NONE;
|
2010-10-06 00:36:58 +02:00
|
|
|
continue;
|
2010-10-27 16:40:39 +02:00
|
|
|
}
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2010-10-12 23:37:44 +02:00
|
|
|
const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2012-09-26 03:32:35 +02:00
|
|
|
if (entry->prev_action != NONE && prev != (unsigned int) -1)
|
2016-05-02 14:47:45 +02:00
|
|
|
{
|
2014-07-17 20:22:11 +02:00
|
|
|
info[prev].arabic_shaping_action() = entry->prev_action;
|
2016-05-02 14:47:45 +02:00
|
|
|
buffer->unsafe_to_break (prev, i + 1);
|
|
|
|
}
|
2010-10-06 00:36:58 +02:00
|
|
|
|
2014-07-17 20:22:11 +02:00
|
|
|
info[i].arabic_shaping_action() = entry->curr_action;
|
2010-10-06 00:36:58 +02:00
|
|
|
|
|
|
|
prev = i;
|
|
|
|
state = entry->next_state;
|
|
|
|
}
|
|
|
|
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
for (unsigned int i = 0; i < buffer->context_len[1]; i++)
|
|
|
|
{
|
|
|
|
unsigned int this_type = get_joining_type (buffer->context[1][i], buffer->unicode->general_category (buffer->context[1][i]));
|
2012-09-26 03:32:35 +02:00
|
|
|
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
if (unlikely (this_type == JOINING_TYPE_T))
|
|
|
|
continue;
|
2012-09-26 03:32:35 +02:00
|
|
|
|
Make it easier to use HB_BUFFER_FLAG_BOT/EOT
Previously, we expected users to provide BOT/EOT flags when the
text *segment* was at paragraph boundaries. This meant that for
clients that provide full paragraph to HarfBuzz (eg. Pango), they
had code like this:
hb_buffer_set_flags (hb_buffer,
(item_offset == 0 ? HB_BUFFER_FLAG_BOT : 0) |
(item_offset + item_length == paragraph_length ?
HB_BUFFER_FLAG_EOT : 0));
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
After this change such clients can simply say:
hb_buffer_set_flags (hb_buffer,
HB_BUFFER_FLAG_BOT | HB_BUFFER_FLAG_EOT);
hb_buffer_add_utf8 (hb_buffer,
paragraph_text, paragraph_length,
item_offset, item_length);
Ie, HarfBuzz itself checks whether the segment is at the beginning/end
of the paragraph. Clients that only pass item-at-a-time to HarfBuzz
continue not setting any flags whatsoever.
Another way to put it is: if there's pre-context text in the buffer,
HarfBuzz ignores the BOT flag. If there's post-context, it ignores
EOT flag.
2014-08-02 22:17:44 +02:00
|
|
|
const arabic_state_table_entry *entry = &arabic_state_table[state][this_type];
|
|
|
|
if (entry->prev_action != NONE && prev != (unsigned int) -1)
|
|
|
|
info[prev].arabic_shaping_action() = entry->prev_action;
|
|
|
|
break;
|
|
|
|
}
|
2014-07-17 20:16:38 +02:00
|
|
|
}
|
2012-09-26 03:32:35 +02:00
|
|
|
|
2014-07-17 20:16:38 +02:00
|
|
|
static void
|
|
|
|
mongolian_variation_selectors (hb_buffer_t *buffer)
|
|
|
|
{
|
|
|
|
/* Copy arabic_shaping_action() from base to Mongolian variation selectors. */
|
|
|
|
unsigned int count = buffer->len;
|
|
|
|
hb_glyph_info_t *info = buffer->info;
|
|
|
|
for (unsigned int i = 1; i < count; i++)
|
2017-06-19 12:17:09 +02:00
|
|
|
if (unlikely (hb_in_range<hb_codepoint_t> (info[i].codepoint, 0x180Bu, 0x180Du)))
|
2014-07-17 20:16:38 +02:00
|
|
|
info[i].arabic_shaping_action() = info[i - 1].arabic_shaping_action();
|
2012-08-12 00:34:13 +02:00
|
|
|
}
|
|
|
|
|
2015-07-22 11:32:30 +02:00
|
|
|
void
|
|
|
|
setup_masks_arabic_plan (const arabic_shape_plan_t *arabic_plan,
|
|
|
|
hb_buffer_t *buffer,
|
|
|
|
hb_script_t script)
|
2012-08-12 00:34:13 +02:00
|
|
|
{
|
2014-07-17 20:16:38 +02:00
|
|
|
HB_BUFFER_ALLOCATE_VAR (buffer, arabic_shaping_action);
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
arabic_joining (buffer);
|
2015-07-22 11:32:30 +02:00
|
|
|
if (script == HB_SCRIPT_MONGOLIAN)
|
2014-07-17 20:16:38 +02:00
|
|
|
mongolian_variation_selectors (buffer);
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
unsigned int count = buffer->len;
|
2014-07-17 20:22:11 +02:00
|
|
|
hb_glyph_info_t *info = buffer->info;
|
2012-09-06 04:19:28 +02:00
|
|
|
for (unsigned int i = 0; i < count; i++)
|
2014-07-17 20:22:11 +02:00
|
|
|
info[i].mask |= arabic_plan->mask_array[info[i].arabic_shaping_action()];
|
2012-08-12 00:34:13 +02:00
|
|
|
}
|
2011-07-28 22:48:43 +02:00
|
|
|
|
2015-07-22 11:32:30 +02:00
|
|
|
static void
|
|
|
|
setup_masks_arabic (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_buffer_t *buffer,
|
|
|
|
hb_font_t *font HB_UNUSED)
|
|
|
|
{
|
|
|
|
const arabic_shape_plan_t *arabic_plan = (const arabic_shape_plan_t *) plan->data;
|
|
|
|
setup_masks_arabic_plan (arabic_plan, buffer, plan->props.script);
|
|
|
|
}
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
|
2013-10-16 13:42:38 +02:00
|
|
|
static void
|
2018-09-27 00:42:37 +02:00
|
|
|
flip_joiners (const hb_ot_shape_plan_t *plan HB_UNUSED,
|
2013-10-16 13:42:38 +02:00
|
|
|
hb_font_t *font HB_UNUSED,
|
|
|
|
hb_buffer_t *buffer)
|
|
|
|
{
|
2018-09-30 09:54:40 +02:00
|
|
|
if (!(buffer->scratch_flags & HB_BUFFER_SCRATCH_FLAG_HAS_JOINERS))
|
|
|
|
return;
|
|
|
|
|
2013-10-16 13:42:38 +02:00
|
|
|
unsigned int count = buffer->len;
|
2014-07-17 20:22:11 +02:00
|
|
|
hb_glyph_info_t *info = buffer->info;
|
2013-10-16 13:42:38 +02:00
|
|
|
for (unsigned int i = 0; i < count; i++)
|
2018-09-27 00:42:37 +02:00
|
|
|
if (_hb_glyph_info_is_joiner (&info[i]))
|
2014-07-17 20:22:11 +02:00
|
|
|
_hb_glyph_info_flip_joiners (&info[i]);
|
2013-10-16 13:42:38 +02:00
|
|
|
}
|
|
|
|
|
2012-08-12 00:34:13 +02:00
|
|
|
static void
|
2012-09-06 04:19:28 +02:00
|
|
|
arabic_fallback_shape (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_font_t *font,
|
|
|
|
hb_buffer_t *buffer)
|
2012-08-12 00:34:13 +02:00
|
|
|
{
|
|
|
|
const arabic_shape_plan_t *arabic_plan = (const arabic_shape_plan_t *) plan->data;
|
|
|
|
|
2012-09-06 04:19:28 +02:00
|
|
|
if (!arabic_plan->do_fallback)
|
|
|
|
return;
|
|
|
|
|
|
|
|
retry:
|
2018-08-09 09:22:37 +02:00
|
|
|
arabic_fallback_plan_t *fallback_plan = arabic_plan->fallback_plan.get ();
|
2012-09-06 04:19:28 +02:00
|
|
|
if (unlikely (!fallback_plan))
|
2012-08-12 00:34:13 +02:00
|
|
|
{
|
2012-09-06 04:19:28 +02:00
|
|
|
/* This sucks. We need a font to build the fallback plan... */
|
|
|
|
fallback_plan = arabic_fallback_plan_create (plan, font);
|
2018-08-09 09:22:37 +02:00
|
|
|
if (unlikely (!arabic_plan->fallback_plan.cmpexch (nullptr, fallback_plan)))
|
|
|
|
{
|
2012-09-06 04:19:28 +02:00
|
|
|
arabic_fallback_plan_destroy (fallback_plan);
|
|
|
|
goto retry;
|
|
|
|
}
|
2012-08-12 00:34:13 +02:00
|
|
|
}
|
2012-09-06 04:19:28 +02:00
|
|
|
|
|
|
|
arabic_fallback_plan_shape (fallback_plan, font, buffer);
|
2010-10-06 00:36:58 +02:00
|
|
|
}
|
|
|
|
|
2015-11-06 02:29:03 +01:00
|
|
|
/*
|
|
|
|
* Stretch feature: "stch".
|
|
|
|
* See example here:
|
2018-04-12 11:10:45 +02:00
|
|
|
* https://docs.microsoft.com/en-us/typography/script-development/syriac
|
2015-11-06 02:29:03 +01:00
|
|
|
* We implement this in a generic way, such that the Arabic subtending
|
|
|
|
* marks can use it as well.
|
|
|
|
*/
|
|
|
|
|
|
|
|
static void
|
|
|
|
record_stch (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_font_t *font,
|
|
|
|
hb_buffer_t *buffer)
|
|
|
|
{
|
|
|
|
const arabic_shape_plan_t *arabic_plan = (const arabic_shape_plan_t *) plan->data;
|
|
|
|
if (!arabic_plan->has_stch)
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* 'stch' feature was just applied. Look for anything that multiplied,
|
|
|
|
* and record it for stch treatment later. Note that rtlm, frac, etc
|
|
|
|
* are applied before stch, but we assume that they didn't result in
|
|
|
|
* anything multiplying into 5 pieces, so it's safe-ish... */
|
|
|
|
|
|
|
|
unsigned int count = buffer->len;
|
|
|
|
hb_glyph_info_t *info = buffer->info;
|
|
|
|
for (unsigned int i = 0; i < count; i++)
|
|
|
|
if (unlikely (_hb_glyph_info_multiplied (&info[i])))
|
|
|
|
{
|
|
|
|
unsigned int comp = _hb_glyph_info_get_lig_comp (&info[i]);
|
|
|
|
info[i].arabic_shaping_action() = comp % 2 ? STCH_REPEATING : STCH_FIXED;
|
|
|
|
buffer->scratch_flags |= HB_BUFFER_SCRATCH_FLAG_ARABIC_HAS_STCH;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
apply_stch (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_buffer_t *buffer,
|
|
|
|
hb_font_t *font)
|
|
|
|
{
|
|
|
|
if (likely (!(buffer->scratch_flags & HB_BUFFER_SCRATCH_FLAG_ARABIC_HAS_STCH)))
|
|
|
|
return;
|
|
|
|
|
|
|
|
/* The Arabic shaper currently always processes in RTL mode, so we should
|
|
|
|
* stretch / position the stretched pieces to the left / preceding glyphs. */
|
|
|
|
|
|
|
|
/* We do a two pass implementation:
|
|
|
|
* First pass calculates the exact number of extra glyphs we need,
|
|
|
|
* We then enlarge buffer to have that much room,
|
|
|
|
* Second pass applies the stretch, copying things to the end of buffer.
|
|
|
|
*/
|
|
|
|
|
|
|
|
int sign = font->x_scale < 0 ? -1 : +1;
|
|
|
|
unsigned int extra_glyphs_needed = 0; // Set during MEASURE, used during CUT
|
2015-11-06 07:28:30 +01:00
|
|
|
typedef enum { MEASURE, CUT } step_t;
|
2015-11-06 02:29:03 +01:00
|
|
|
|
2015-11-06 07:28:30 +01:00
|
|
|
for (step_t step = MEASURE; step <= CUT; step = (step_t) (step + 1))
|
2015-11-06 02:29:03 +01:00
|
|
|
{
|
|
|
|
unsigned int count = buffer->len;
|
|
|
|
hb_glyph_info_t *info = buffer->info;
|
|
|
|
hb_glyph_position_t *pos = buffer->pos;
|
|
|
|
unsigned int new_len = count + extra_glyphs_needed; // write head during CUT
|
|
|
|
unsigned int j = new_len;
|
|
|
|
for (unsigned int i = count; i; i--)
|
|
|
|
{
|
|
|
|
if (!hb_in_range<unsigned> (info[i - 1].arabic_shaping_action(), STCH_FIXED, STCH_REPEATING))
|
|
|
|
{
|
|
|
|
if (step == CUT)
|
|
|
|
{
|
|
|
|
--j;
|
|
|
|
info[j] = info[i - 1];
|
|
|
|
pos[j] = pos[i - 1];
|
|
|
|
}
|
|
|
|
continue;
|
|
|
|
}
|
|
|
|
|
|
|
|
/* Yay, justification! */
|
|
|
|
|
|
|
|
hb_position_t w_total = 0; // Total to be filled
|
|
|
|
hb_position_t w_fixed = 0; // Sum of fixed tiles
|
|
|
|
hb_position_t w_repeating = 0; // Sum of repeating tiles
|
|
|
|
int n_fixed = 0;
|
|
|
|
int n_repeating = 0;
|
|
|
|
|
|
|
|
unsigned int end = i;
|
|
|
|
while (i &&
|
|
|
|
hb_in_range<unsigned> (info[i - 1].arabic_shaping_action(), STCH_FIXED, STCH_REPEATING))
|
|
|
|
{
|
|
|
|
i--;
|
2015-12-17 12:04:33 +01:00
|
|
|
hb_position_t width = font->get_glyph_h_advance (info[i].codepoint);
|
2015-11-06 02:29:03 +01:00
|
|
|
if (info[i].arabic_shaping_action() == STCH_FIXED)
|
|
|
|
{
|
2015-12-17 12:04:33 +01:00
|
|
|
w_fixed += width;
|
2015-11-06 02:29:03 +01:00
|
|
|
n_fixed++;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2015-12-17 12:04:33 +01:00
|
|
|
w_repeating += width;
|
2015-11-06 02:29:03 +01:00
|
|
|
n_repeating++;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
unsigned int start = i;
|
|
|
|
unsigned int context = i;
|
|
|
|
while (context &&
|
|
|
|
!hb_in_range<unsigned> (info[context - 1].arabic_shaping_action(), STCH_FIXED, STCH_REPEATING) &&
|
|
|
|
(_hb_glyph_info_is_default_ignorable (&info[context - 1]) ||
|
2015-11-07 01:27:44 +01:00
|
|
|
HB_ARABIC_GENERAL_CATEGORY_IS_WORD (_hb_glyph_info_get_general_category (&info[context - 1]))))
|
2015-11-06 02:29:03 +01:00
|
|
|
{
|
|
|
|
context--;
|
|
|
|
w_total += pos[context].x_advance;
|
|
|
|
}
|
|
|
|
i++; // Don't touch i again.
|
|
|
|
|
2017-10-15 12:11:08 +02:00
|
|
|
DEBUG_MSG (ARABIC, nullptr, "%s stretch at (%d,%d,%d)",
|
2015-11-06 02:29:03 +01:00
|
|
|
step == MEASURE ? "measuring" : "cutting", context, start, end);
|
2017-10-15 12:11:08 +02:00
|
|
|
DEBUG_MSG (ARABIC, nullptr, "rest of word: count=%d width %d", start - context, w_total);
|
|
|
|
DEBUG_MSG (ARABIC, nullptr, "fixed tiles: count=%d width=%d", n_fixed, w_fixed);
|
|
|
|
DEBUG_MSG (ARABIC, nullptr, "repeating tiles: count=%d width=%d", n_repeating, w_repeating);
|
2015-11-06 02:29:03 +01:00
|
|
|
|
|
|
|
/* Number of additional times to repeat each repeating tile. */
|
|
|
|
int n_copies = 0;
|
|
|
|
|
2015-12-17 12:04:33 +01:00
|
|
|
hb_position_t w_remaining = w_total - w_fixed;
|
2015-11-06 02:29:03 +01:00
|
|
|
if (sign * w_remaining > sign * w_repeating && sign * w_repeating > 0)
|
2015-12-05 18:47:37 +01:00
|
|
|
n_copies = (sign * w_remaining) / (sign * w_repeating) - 1;
|
|
|
|
|
|
|
|
/* See if we can improve the fit by adding an extra repeat and squeezing them together a bit. */
|
|
|
|
hb_position_t extra_repeat_overlap = 0;
|
|
|
|
hb_position_t shortfall = sign * w_remaining - sign * w_repeating * (n_copies + 1);
|
2017-02-08 01:53:21 +01:00
|
|
|
if (shortfall > 0 && n_repeating > 0)
|
2015-12-05 18:47:37 +01:00
|
|
|
{
|
|
|
|
++n_copies;
|
|
|
|
hb_position_t excess = (n_copies + 1) * sign * w_repeating - sign * w_remaining;
|
|
|
|
if (excess > 0)
|
|
|
|
extra_repeat_overlap = excess / (n_copies * n_repeating);
|
|
|
|
}
|
2015-11-06 02:29:03 +01:00
|
|
|
|
|
|
|
if (step == MEASURE)
|
|
|
|
{
|
|
|
|
extra_glyphs_needed += n_copies * n_repeating;
|
2017-10-15 12:11:08 +02:00
|
|
|
DEBUG_MSG (ARABIC, nullptr, "will add extra %d copies of repeating tiles", n_copies);
|
2015-11-06 02:29:03 +01:00
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
2017-11-09 02:04:31 +01:00
|
|
|
buffer->unsafe_to_break (context, end);
|
2015-12-17 12:04:33 +01:00
|
|
|
hb_position_t x_offset = 0;
|
2015-11-06 02:29:03 +01:00
|
|
|
for (unsigned int k = end; k > start; k--)
|
|
|
|
{
|
2015-12-17 12:04:33 +01:00
|
|
|
hb_position_t width = font->get_glyph_h_advance (info[k - 1].codepoint);
|
2015-11-06 02:29:03 +01:00
|
|
|
|
|
|
|
unsigned int repeat = 1;
|
|
|
|
if (info[k - 1].arabic_shaping_action() == STCH_REPEATING)
|
|
|
|
repeat += n_copies;
|
|
|
|
|
2017-10-15 12:11:08 +02:00
|
|
|
DEBUG_MSG (ARABIC, nullptr, "appending %d copies of glyph %d; j=%d",
|
2015-11-06 02:29:03 +01:00
|
|
|
repeat, info[k - 1].codepoint, j);
|
|
|
|
for (unsigned int n = 0; n < repeat; n++)
|
|
|
|
{
|
2015-12-17 12:04:33 +01:00
|
|
|
x_offset -= width;
|
2015-12-05 18:47:37 +01:00
|
|
|
if (n > 0)
|
|
|
|
x_offset += extra_repeat_overlap;
|
2015-11-06 02:29:03 +01:00
|
|
|
pos[k - 1].x_offset = x_offset;
|
|
|
|
/* Append copy. */
|
|
|
|
--j;
|
|
|
|
info[j] = info[k - 1];
|
|
|
|
pos[j] = pos[k - 1];
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
if (step == MEASURE)
|
|
|
|
{
|
|
|
|
if (unlikely (!buffer->ensure (count + extra_glyphs_needed)))
|
|
|
|
break;
|
|
|
|
}
|
|
|
|
else
|
|
|
|
{
|
|
|
|
assert (j == 0);
|
|
|
|
buffer->len = new_len;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
|
|
static void
|
|
|
|
postprocess_glyphs_arabic (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_buffer_t *buffer,
|
|
|
|
hb_font_t *font)
|
|
|
|
{
|
|
|
|
apply_stch (plan, buffer, font);
|
|
|
|
|
|
|
|
HB_BUFFER_DEALLOCATE_VAR (buffer, arabic_shaping_action);
|
|
|
|
}
|
2012-09-06 04:19:28 +02:00
|
|
|
|
2018-04-12 11:10:45 +02:00
|
|
|
/* https://unicode.org/reports/tr53/tr53-1.pdf */
|
2017-10-04 14:47:10 +02:00
|
|
|
|
|
|
|
static hb_codepoint_t
|
|
|
|
modifier_combining_marks[] =
|
|
|
|
{
|
|
|
|
0x0654u, /* ARABIC HAMZA ABOVE */
|
|
|
|
0x0655u, /* ARABIC HAMZA BELOW */
|
|
|
|
0x0658u, /* ARABIC MARK NOON GHUNNA */
|
|
|
|
0x06DCu, /* ARABIC SMALL HIGH SEEN */
|
|
|
|
0x06E3u, /* ARABIC SMALL LOW SEEN */
|
|
|
|
0x06E7u, /* ARABIC SMALL HIGH YEH */
|
|
|
|
0x06E8u, /* ARABIC SMALL HIGH NOON */
|
|
|
|
0x08F3u, /* ARABIC SMALL HIGH WAW */
|
|
|
|
};
|
|
|
|
|
|
|
|
static inline bool
|
|
|
|
info_is_mcm (const hb_glyph_info_t &info)
|
|
|
|
{
|
|
|
|
hb_codepoint_t u = info.codepoint;
|
|
|
|
for (unsigned int i = 0; i < ARRAY_LENGTH (modifier_combining_marks); i++)
|
|
|
|
if (u == modifier_combining_marks[i])
|
|
|
|
return true;
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
|
|
|
static void
|
|
|
|
reorder_marks_arabic (const hb_ot_shape_plan_t *plan,
|
|
|
|
hb_buffer_t *buffer,
|
|
|
|
unsigned int start,
|
|
|
|
unsigned int end)
|
|
|
|
{
|
|
|
|
hb_glyph_info_t *info = buffer->info;
|
|
|
|
|
2018-01-05 13:46:12 +01:00
|
|
|
DEBUG_MSG (ARABIC, buffer, "Reordering marks from %d to %d", start, end);
|
|
|
|
|
2017-10-04 14:47:10 +02:00
|
|
|
unsigned int i = start;
|
|
|
|
for (unsigned int cc = 220; cc <= 230; cc += 10)
|
|
|
|
{
|
2018-01-05 13:46:12 +01:00
|
|
|
DEBUG_MSG (ARABIC, buffer, "Looking for %d's starting at %d", cc, i);
|
2017-10-04 14:47:10 +02:00
|
|
|
while (i < end && info_cc(info[i]) < cc)
|
|
|
|
i++;
|
2018-01-05 13:46:12 +01:00
|
|
|
DEBUG_MSG (ARABIC, buffer, "Looking for %d's stopped at %d", cc, i);
|
2017-10-04 14:47:10 +02:00
|
|
|
|
|
|
|
if (i == end)
|
|
|
|
break;
|
|
|
|
|
|
|
|
if (info_cc(info[i]) > cc)
|
|
|
|
continue;
|
|
|
|
|
|
|
|
unsigned int j = i;
|
2018-01-05 13:46:12 +01:00
|
|
|
while (j < end && info_cc(info[j]) == cc && info_is_mcm (info[j]))
|
2017-10-04 14:47:10 +02:00
|
|
|
j++;
|
|
|
|
|
|
|
|
if (i == j)
|
|
|
|
continue;
|
|
|
|
|
2018-01-05 13:46:12 +01:00
|
|
|
DEBUG_MSG (ARABIC, buffer, "Found %d's from %d to %d", cc, i, j);
|
|
|
|
|
2017-10-04 14:47:10 +02:00
|
|
|
/* Shift it! */
|
2018-01-05 13:46:12 +01:00
|
|
|
DEBUG_MSG (ARABIC, buffer, "Shifting %d's: %d %d", cc, i, j);
|
2017-10-04 14:47:10 +02:00
|
|
|
hb_glyph_info_t temp[HB_OT_SHAPE_COMPLEX_MAX_COMBINING_MARKS];
|
|
|
|
assert (j - i <= ARRAY_LENGTH (temp));
|
2017-10-07 13:17:33 +02:00
|
|
|
buffer->merge_clusters (start, j);
|
2017-10-04 14:47:10 +02:00
|
|
|
memmove (temp, &info[i], (j - i) * sizeof (hb_glyph_info_t));
|
|
|
|
memmove (&info[start + j - i], &info[start], (i - start) * sizeof (hb_glyph_info_t));
|
|
|
|
memmove (&info[start], temp, (j - i) * sizeof (hb_glyph_info_t));
|
|
|
|
|
2018-01-05 13:46:12 +01:00
|
|
|
/* Renumber CC such that the reordered sequence is still sorted.
|
|
|
|
* 22 and 26 are chosen because they are smaller than all Arabic categories,
|
|
|
|
* and are folded back to 220/230 respectively during fallback mark positioning.
|
|
|
|
*
|
|
|
|
* We do this because the CGJ-handling logic in the normalizer relies on
|
|
|
|
* mark sequences having an increasing order even after this reordering.
|
|
|
|
* https://github.com/harfbuzz/harfbuzz/issues/554
|
|
|
|
* This, however, does break some obscure sequences, where the normalizer
|
|
|
|
* might compose a sequence that it should not. For example, in the seequence
|
|
|
|
* ALEF, HAMZAH, MADDAH, we should NOT try to compose ALEF+MADDAH, but with this
|
|
|
|
* renumbering, we will.
|
|
|
|
*/
|
|
|
|
unsigned int new_start = start + j - i;
|
|
|
|
unsigned int new_cc = cc == 220 ? HB_MODIFIED_COMBINING_CLASS_CCC22 : HB_MODIFIED_COMBINING_CLASS_CCC26;
|
|
|
|
while (start < new_start)
|
|
|
|
{
|
|
|
|
_hb_glyph_info_set_modified_combining_class (&info[start], new_cc);
|
|
|
|
start++;
|
|
|
|
}
|
2017-10-04 14:47:10 +02:00
|
|
|
|
|
|
|
i = j;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
2012-07-31 03:08:51 +02:00
|
|
|
const hb_ot_complex_shaper_t _hb_ot_complex_shaper_arabic =
|
|
|
|
{
|
|
|
|
collect_features_arabic,
|
2017-10-15 12:11:08 +02:00
|
|
|
nullptr, /* override_features */
|
2012-08-12 00:20:28 +02:00
|
|
|
data_create_arabic,
|
|
|
|
data_destroy_arabic,
|
2017-10-15 12:11:08 +02:00
|
|
|
nullptr, /* preprocess_text */
|
2015-11-06 02:29:03 +01:00
|
|
|
postprocess_glyphs_arabic,
|
2013-12-31 09:04:35 +01:00
|
|
|
HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT,
|
2017-10-15 12:11:08 +02:00
|
|
|
nullptr, /* decompose */
|
|
|
|
nullptr, /* compose */
|
2012-07-31 03:08:51 +02:00
|
|
|
setup_masks_arabic,
|
2017-10-15 12:11:08 +02:00
|
|
|
nullptr, /* disable_otl */
|
2017-10-04 14:47:10 +02:00
|
|
|
reorder_marks_arabic,
|
2013-10-28 00:20:59 +01:00
|
|
|
HB_OT_SHAPE_ZERO_WIDTH_MARKS_BY_GDEF_LATE,
|
2012-11-14 22:48:26 +01:00
|
|
|
true, /* fallback_position */
|
2012-07-31 03:08:51 +02:00
|
|
|
};
|