harfbuzz/src/hb-ot-shape-normalize.cc

409 lines
13 KiB
C++
Raw Normal View History

2011-07-21 06:51:18 +02:00
/*
* Copyright © 2011,2012 Google, Inc.
2011-07-21 06:51:18 +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
*/
#include "hb-ot-shape-normalize-private.hh"
#include "hb-ot-shape-complex-private.hh"
2011-07-21 06:51:18 +02:00
#include "hb-ot-shape-private.hh"
2011-07-21 21:25:01 +02:00
/*
* HIGHLEVEL DESIGN:
*
* This file exports one main function: _hb_ot_shape_normalize().
*
* This function closely reflects the Unicode Normalization Algorithm,
2012-03-07 02:47:50 +01:00
* yet it's different.
*
* Each shaper specifies whether it prefers decomposed (NFD) or composed (NFC).
* The logic however tries to use whatever the font can support.
2011-07-21 21:25:01 +02:00
*
* In general what happens is that: each grapheme is decomposed in a chain
2011-09-16 22:33:18 +02:00
* of 1:2 decompositions, marks reordered, and then recomposed if desired,
2011-07-21 21:25:01 +02:00
* so far it's like Unicode Normalization. However, the decomposition and
* recomposition only happens if the font supports the resulting characters.
*
* The goals are:
*
* - Try to render all canonically equivalent strings similarly. To really
* achieve this we have to always do the full decomposition and then
* selectively recompose from there. It's kinda too expensive though, so
* we skip some cases. For example, if composed is desired, we simply
* don't touch 1-character clusters that are supported by the font, even
* though their NFC may be different.
*
* - When a font has a precomposed character for a sequence but the 'ccmp'
2011-09-16 22:33:18 +02:00
* feature in the font is not adequate, use the precomposed character
2011-07-21 21:25:01 +02:00
* which typically has better mark positioning.
*
2011-09-28 22:20:09 +02:00
* - When a font does not support a combining mark, but supports it precomposed
2012-03-07 02:47:50 +01:00
* with previous base, use that. This needs the itemizer to have this
2012-03-07 16:21:24 +01:00
* knowledge too. We need to provide assistance to the itemizer.
2011-09-28 22:20:09 +02:00
*
2011-07-21 21:25:01 +02:00
* - When a font does not support a character but supports its decomposition,
* well, use the decomposition (preferring the canonical decomposition, but
* falling back to the compatibility decomposition if necessary). The
* compatibility decomposition is really nice to have, for characters like
* ellipsis, or various-sized space characters.
2011-07-21 21:25:01 +02:00
*
* - The complex shapers can customize the compose and decompose functions to
* offload some of their requirements to the normalizer. For example, the
* Indic shaper may want to disallow recomposing of two matras.
*
* - We try compatibility decomposition if decomposing through canonical
* decomposition alone failed to find a sequence that the font supports.
* We don't try compatibility decomposition recursively during the canonical
* decomposition phase. This has minimal impact. There are only a handful
* of Greek letter that have canonical decompositions that include characters
* with compatibility decomposition. Those can be found using this command:
*
* egrep "`echo -n ';('; grep ';<' UnicodeData.txt | cut -d';' -f1 | tr '\n' '|'; echo ') '`" UnicodeData.txt
2011-07-21 21:25:01 +02:00
*/
static bool
decompose_unicode (const hb_ot_shape_normalize_context_t *c,
hb_codepoint_t ab,
hb_codepoint_t *a,
hb_codepoint_t *b)
2012-08-07 22:51:48 +02:00
{
return c->unicode->decompose (ab, a, b);
2012-08-07 22:51:48 +02:00
}
static bool
compose_unicode (const hb_ot_shape_normalize_context_t *c,
hb_codepoint_t a,
hb_codepoint_t b,
hb_codepoint_t *ab)
2012-08-07 22:51:48 +02:00
{
return c->unicode->compose (a, b, ab);
2012-08-07 22:51:48 +02:00
}
static inline void
set_glyph (hb_glyph_info_t &info, hb_font_t *font)
{
2012-08-10 09:28:50 +02:00
font->get_glyph (info.codepoint, 0, &info.glyph_index());
}
2012-08-10 03:31:52 +02:00
static inline void
output_char (hb_buffer_t *buffer, hb_codepoint_t unichar, hb_codepoint_t glyph)
{
buffer->cur().glyph_index() = glyph;
2012-08-10 03:31:52 +02:00
buffer->output_glyph (unichar);
_hb_glyph_info_set_unicode_props (&buffer->prev(), buffer->unicode);
}
2011-07-22 17:07:05 +02:00
2012-08-10 03:31:52 +02:00
static inline void
next_char (hb_buffer_t *buffer, hb_codepoint_t glyph)
2012-08-10 03:31:52 +02:00
{
buffer->cur().glyph_index() = glyph;
2012-08-10 03:31:52 +02:00
buffer->next_glyph ();
}
static inline void
skip_char (hb_buffer_t *buffer)
{
buffer->skip_glyph ();
}
/* Returns 0 if didn't decompose, number of resulting characters otherwise. */
static inline unsigned int
decompose (const hb_ot_shape_normalize_context_t *c, bool shortest, hb_codepoint_t ab)
2011-07-21 06:51:18 +02:00
{
hb_codepoint_t a, b, a_glyph, b_glyph;
2011-07-22 17:07:05 +02:00
if (!c->decompose (c, ab, &a, &b) ||
(b && !c->font->get_glyph (b, 0, &b_glyph)))
return 0;
2011-07-22 17:07:05 +02:00
bool has_a = c->font->get_glyph (a, 0, &a_glyph);
2011-07-22 22:15:32 +02:00
if (shortest && has_a) {
/* Output a and b */
output_char (c->buffer, a, a_glyph);
if (likely (b)) {
output_char (c->buffer, b, b_glyph);
return 2;
}
return 1;
2011-07-22 22:15:32 +02:00
}
2011-07-21 06:51:18 +02:00
unsigned int ret;
if ((ret = decompose (c, shortest, a))) {
if (b) {
output_char (c->buffer, b, b_glyph);
return ret + 1;
}
return ret;
2011-07-22 22:15:32 +02:00
}
2011-07-22 17:07:05 +02:00
2011-07-22 22:15:32 +02:00
if (has_a) {
output_char (c->buffer, a, a_glyph);
if (likely (b)) {
output_char (c->buffer, b, b_glyph);
return 2;
}
return 1;
2011-07-22 17:07:05 +02:00
}
return 0;
2011-07-21 17:31:08 +02:00
}
/* Returns 0 if didn't decompose, number of resulting characters otherwise. */
static inline bool
decompose_compatibility (const hb_ot_shape_normalize_context_t *c, hb_codepoint_t u)
2011-07-21 18:16:45 +02:00
{
unsigned int len, i;
hb_codepoint_t decomposed[HB_UNICODE_MAX_DECOMPOSITION_LEN];
hb_codepoint_t glyphs[HB_UNICODE_MAX_DECOMPOSITION_LEN];
len = c->buffer->unicode->decompose_compatibility (u, decomposed);
if (!len)
return 0;
for (i = 0; i < len; i++)
if (!c->font->get_glyph (decomposed[i], 0, &glyphs[i]))
return 0;
for (i = 0; i < len; i++)
output_char (c->buffer, decomposed[i], glyphs[i]);
return len;
2011-07-21 18:16:45 +02:00
}
/* Returns true if recomposition may be benefitial. */
static inline bool
decompose_current_character (const hb_ot_shape_normalize_context_t *c, bool shortest)
2011-07-21 17:31:08 +02:00
{
hb_buffer_t * const buffer = c->buffer;
2011-07-22 22:15:32 +02:00
hb_codepoint_t glyph;
unsigned int len = 1;
2011-07-22 22:15:32 +02:00
/* Kind of a cute waterfall here... */
if (shortest && c->font->get_glyph (buffer->cur().codepoint, 0, &glyph))
next_char (buffer, glyph);
else if ((len = decompose (c, shortest, buffer->cur().codepoint)))
2012-08-10 03:31:52 +02:00
skip_char (buffer);
else if (!shortest && c->font->get_glyph (buffer->cur().codepoint, 0, &glyph))
next_char (buffer, glyph);
else if ((len = decompose_compatibility (c, buffer->cur().codepoint)))
2012-08-10 03:31:52 +02:00
skip_char (buffer);
2012-08-10 09:28:50 +02:00
else
next_char (buffer, glyph); /* glyph is initialized in earlier branches. */
/*
* A recomposition would only be useful if we decomposed into at least three
* characters...
*/
return len > 2;
}
static inline void
handle_variation_selector_cluster (const hb_ot_shape_normalize_context_t *c, unsigned int end)
{
hb_buffer_t * const buffer = c->buffer;
for (; buffer->idx < end - 1;) {
if (unlikely (buffer->unicode->is_variation_selector (buffer->cur(+1).codepoint))) {
/* The next two lines are some ugly lines... But work. */
c->font->get_glyph (buffer->cur().codepoint, buffer->cur(+1).codepoint, &buffer->cur().glyph_index());
buffer->replace_glyphs (2, 1, &buffer->cur().codepoint);
} else {
set_glyph (buffer->cur(), c->font);
buffer->next_glyph ();
}
}
if (likely (buffer->idx < end)) {
set_glyph (buffer->cur(), c->font);
buffer->next_glyph ();
}
2011-07-21 06:51:18 +02:00
}
/* Returns true if recomposition may be benefitial. */
static inline bool
decompose_multi_char_cluster (const hb_ot_shape_normalize_context_t *c, unsigned int end)
2011-07-21 06:51:18 +02:00
{
hb_buffer_t * const buffer = c->buffer;
2011-07-21 21:25:01 +02:00
/* TODO Currently if there's a variation-selector we give-up, it's just too hard. */
for (unsigned int i = buffer->idx; i < end; i++)
2012-08-01 23:13:10 +02:00
if (unlikely (buffer->unicode->is_variation_selector (buffer->info[i].codepoint))) {
handle_variation_selector_cluster (c, end);
return false;
}
2011-07-21 18:16:45 +02:00
while (buffer->idx < end)
decompose_current_character (c, false);
/* We can be smarter here and only return true if there are at least two ccc!=0 marks.
* But does not matter. */
return true;
}
static inline bool
decompose_cluster (const hb_ot_shape_normalize_context_t *c, bool short_circuit, unsigned int end)
{
if (likely (c->buffer->idx + 1 == end))
return decompose_current_character (c, short_circuit);
else
return decompose_multi_char_cluster (c, end);
2011-07-21 06:51:18 +02:00
}
static int
compare_combining_class (const hb_glyph_info_t *pa, const hb_glyph_info_t *pb)
{
unsigned int a = _hb_glyph_info_get_modified_combining_class (pa);
unsigned int b = _hb_glyph_info_get_modified_combining_class (pb);
return a < b ? -1 : a == b ? 0 : +1;
}
2011-07-22 17:07:05 +02:00
void
_hb_ot_shape_normalize (const hb_ot_shape_plan_t *plan,
hb_buffer_t *buffer,
hb_font_t *font)
2011-07-21 06:51:18 +02:00
{
hb_ot_shape_normalization_mode_t mode = plan->shaper->normalization_preference ?
plan->shaper->normalization_preference (&buffer->props) :
HB_OT_SHAPE_NORMALIZATION_MODE_DEFAULT;
const hb_ot_shape_normalize_context_t c = {
plan,
buffer,
font,
buffer->unicode,
plan->shaper->decompose ? plan->shaper->decompose : decompose_unicode,
plan->shaper->compose ? plan->shaper->compose : compose_unicode
};
bool short_circuit = mode != HB_OT_SHAPE_NORMALIZATION_MODE_DECOMPOSED &&
mode != HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_DIACRITICS_NO_SHORT_CIRCUIT;
bool can_use_recompose = false;
unsigned int count;
2011-07-21 21:25:01 +02:00
/* We do a fairly straightforward yet custom normalization process in three
* separate rounds: decompose, reorder, recompose (if desired). Currently
* this makes two buffer swaps. We can make it faster by moving the last
* two rounds into the inner loop for the first round, but it's more readable
* this way. */
2011-07-21 21:25:01 +02:00
2011-07-22 22:15:32 +02:00
/* First round, decompose */
buffer->clear_output ();
count = buffer->len;
2011-07-22 17:28:07 +02:00
for (buffer->idx = 0; buffer->idx < count;)
2011-07-21 21:25:01 +02:00
{
2011-07-21 06:51:18 +02:00
unsigned int end;
2011-07-22 17:28:07 +02:00
for (end = buffer->idx + 1; end < count; end++)
if (buffer->cur().cluster != buffer->info[end].cluster)
2011-07-21 06:51:18 +02:00
break;
2011-07-21 21:25:01 +02:00
can_use_recompose = decompose_cluster (&c, short_circuit, end) || can_use_recompose;
2011-07-21 06:51:18 +02:00
}
2011-07-22 17:28:07 +02:00
buffer->swap_buffers ();
2011-07-22 22:15:32 +02:00
if (mode != HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_FULL && !can_use_recompose)
2011-07-22 22:15:32 +02:00
return; /* Done! */
2011-07-22 22:15:32 +02:00
/* Second round, reorder (inplace) */
count = buffer->len;
for (unsigned int i = 0; i < count; i++)
{
if (_hb_glyph_info_get_modified_combining_class (&buffer->info[i]) == 0)
continue;
unsigned int end;
for (end = i + 1; end < count; end++)
if (_hb_glyph_info_get_modified_combining_class (&buffer->info[end]) == 0)
break;
/* We are going to do a bubble-sort. Only do this if the
* sequence is short. Doing it on long sequences can result
* in an O(n^2) DoS. */
if (end - i > 10) {
i = end;
continue;
}
hb_bubble_sort (buffer->info + i, end - i, compare_combining_class);
i = end;
}
2011-07-22 22:15:32 +02:00
if (mode == HB_OT_SHAPE_NORMALIZATION_MODE_DECOMPOSED)
return;
2011-07-22 22:15:32 +02:00
/* Third round, recompose */
/* As noted in the comment earlier, we don't try to combine
* ccc=0 chars with their previous Starter. */
2011-07-22 22:15:32 +02:00
buffer->clear_output ();
count = buffer->len;
unsigned int starter = 0;
buffer->next_glyph ();
while (buffer->idx < count)
{
hb_codepoint_t composed, glyph;
if (/* If mode is NOT COMPOSED_FULL (ie. it's COMPOSED_DIACRITICS), we don't try to
* compose a CCC=0 character with it's preceding starter. */
(mode == HB_OT_SHAPE_NORMALIZATION_MODE_COMPOSED_FULL ||
_hb_glyph_info_get_modified_combining_class (&buffer->cur()) != 0) &&
/* If there's anything between the starter and this char, they should have CCC
* smaller than this character's. */
(starter == buffer->out_len - 1 ||
_hb_glyph_info_get_modified_combining_class (&buffer->prev()) < _hb_glyph_info_get_modified_combining_class (&buffer->cur())) &&
/* And compose. */
c.compose (&c,
buffer->out_info[starter].codepoint,
buffer->cur().codepoint,
&composed) &&
/* And the font has glyph for the composite. */
2012-08-02 01:03:46 +02:00
font->get_glyph (composed, 0, &glyph))
{
2012-06-09 03:01:20 +02:00
/* Composes. */
buffer->next_glyph (); /* Copy to out-buffer. */
2012-06-09 03:01:20 +02:00
if (unlikely (buffer->in_error))
return;
buffer->merge_out_clusters (starter, buffer->out_len);
2012-08-10 03:31:52 +02:00
buffer->out_len--; /* Remove the second composable. */
2012-06-09 03:01:20 +02:00
buffer->out_info[starter].codepoint = composed; /* Modify starter and carry on. */
set_glyph (buffer->out_info[starter], font);
_hb_glyph_info_set_unicode_props (&buffer->out_info[starter], buffer->unicode);
2012-04-07 20:49:13 +02:00
continue;
}
2012-04-07 20:49:13 +02:00
/* Blocked, or doesn't compose. */
buffer->next_glyph ();
if (_hb_glyph_info_get_modified_combining_class (&buffer->prev()) == 0)
starter = buffer->out_len - 1;
2011-07-22 22:15:32 +02:00
}
buffer->swap_buffers ();
2011-07-21 06:51:18 +02:00
}