[HarfBuzz] HARFBUZZ - ARABIC SHAPING PROBLEM

[Sergey]

We are newers in HB and have problems with using HB on Windows in our project
for rendering Arabic/Indic and etc. languages. MAIN our goal while using
harfbuzz is to get appropriate ligatures/replacement glyph indexes and etc. We
build HB on msvc9, use FT-2.4.2 and ICU-4c-4_8_1. Our  implementation of  buffer
shaping example is next:



	/* Get our harfbuzz font/face structs */

	hb_font_t *hb_ft_font = hb_ft_font_create(Font_Mgr::Font_Face, NULL);

    hb_face_t *hb_ft_face = hb_ft_face_create(Font_Mgr::Font_Face, NULL);

	

	hb_ft_font_set_funcs (hb_ft_font); // IS IT AS A MUST or could be ommited??



	wchar_t *w_cut_text = L"عر بي";     // x0639, 0x0631, 0x0020, 0x0628, 0x064a

	

	// INPUT string is wchar_t - NOT SURE that further conversion works well!!

	

	uint16_t *cur_text_16 = (uint16_t*)w_cut_text;

	

	const char *cur_lang = "ar"; // OK



	/* Create a buffer for harfbuzz to use */

	hb_buffer_t *buffer = hb_buffer_create(); //OK



	hb_buffer_set_unicode_funcs(buffer, hb_icu_get_unicode_funcs()); 

	// WHAT IS THE PURPOSE OF USING hb_icu_get_unicode_funcs() or what own
implemtation should provide??

	

	hb_buffer_set_direction(buffer, HB_DIRECTION_RTL); //OK 

	hb_buffer_set_script(buffer, HB_SCRIPT_ARABIC); //OK

	hb_buffer_set_language(buffer, hb_language_from_string(cur_lang)); //OK



	/* Layout the text */

	hb_buffer_add_utf16 (buffer, cur_text_16, 5, 0, 5); // WHAT is the semantic
usage of 3 last parameters??

	

	hb_shape (hb_ft_font, buffer, NULL, 0);



	/* Hand the layout to render */

	unsigned int glyph_count = hb_buffer_get_length (buffer);

	hb_glyph_info_t *glyph_infos = hb_buffer_get_glyph_infos (buffer, NULL);

	hb_glyph_position_t *glyph_positions = hb_buffer_get_glyph_positions (buffer,
NULL);



	for (unsigned int i = 0; i < glyph_count; i++)

	{

		hb_glyph_info_t *info = &glyph_infos[i];

		hb_glyph_position_t *pos = &glyph_positions[i];



		printf ("cluster %d	glyph 0x%x at	(%d,%d)+(%d,%d)\n",

			info->cluster,

			info->codepoint,

			pos->x_offset,

			pos->x_offset,

			pos->x_advance,

			pos->y_advance);



	}

	 

	hb_buffer_destroy (buffer);

	hb_font_destroy (hb_ft_font);

	hb_face_destroy (hb_ft_face);



AS AN OUTPUT we got array of glyphs indexes in info->codepoint, but this array
is the same as array of glyphs indexes, that backs FT for text

 {'ع','ر',' ','ب','ي'}. 

 - > BUT DID NOT GET EXPECTED ligatures/replaced glyphs and etc. So, only string
revers is worked good.

What are we donig wrong??

Could you please give any ideas or provide a sample code for a such task!

THANK YOU VERY MUCH IN ADVANCED!

WE ARE LOOKING FORWARD RECEIVING FEEDBACK FROM YOU!

REGARDS,

Sergey.