# HG changeset patch # User Jason Rumney # Date 1207443885 0 # Node ID 46d39c0f7bf9171aaaaa1f6a905b28148fa7f223 # Parent 673422b6cd01e3771f3757cb86a9630561ec9f2a (uniscribe_shape): Increase items buffer size. Give up if glyph indices not supported. Use uniscribe obtained ABC widths for individual metrics. Map glyph clusters back to characters using fClusterStart flag. Return number of glyphs produced, not chars processed. diff -r 673422b6cd01 -r 46d39c0f7bf9 src/w32uniscribe.c --- a/src/w32uniscribe.c Sun Apr 06 00:03:38 2008 +0000 +++ b/src/w32uniscribe.c Sun Apr 06 01:04:45 2008 +0000 @@ -202,7 +202,7 @@ struct font * font; struct uniscribe_font_info * uniscribe_font; EMACS_UINT nchars; - int nitems, max_items, i, max_glyphs, done_glyphs, done_chars; + int nitems, max_items, i, max_glyphs, done_glyphs; wchar_t *chars; WORD *glyphs, *clusters; SCRIPT_ITEM *items; @@ -222,7 +222,7 @@ /* Get the chars from lgstring in a form we can use with uniscribe. */ max_glyphs = nchars = LGSTRING_LENGTH (lgstring); - done_glyphs = done_chars = 0; + done_glyphs = 0; chars = (wchar_t *) alloca (nchars * sizeof (wchar_t)); for (i = 0; i < nchars; i++) { @@ -238,7 +238,7 @@ /* First we need to break up the glyph string into runs of glyphs that can be treated together. First try a single run. */ max_items = 2; - items = (SCRIPT_ITEM *) xmalloc (sizeof (SCRIPT_ITEM) * max_items); + items = (SCRIPT_ITEM *) xmalloc (sizeof (SCRIPT_ITEM) * max_items + 1); bzero (&control, sizeof (control)); while ((result = ScriptItemize (chars, nchars, max_items, &control, NULL, @@ -247,7 +247,7 @@ /* If that wasn't enough, keep trying with one more run. */ max_items++; items = (SCRIPT_ITEM *) xrealloc (items, - sizeof (SCRIPT_ITEM) * max_items); + sizeof (SCRIPT_ITEM) * max_items + 1); } /* 0 = success in Microsoft's backwards world. */ @@ -257,6 +257,9 @@ return Qnil; } + /* TODO: When we get BIDI support, we need to call ScriptLayout here. + Requires that we know the surrounding context. */ + f = XFRAME (selected_frame); context = get_frame_dc (f); old_font = SelectObject (context, @@ -273,7 +276,7 @@ for (i = 0; i < nitems; i++) { - int nglyphs, nchars_in_run; + int nglyphs, nchars_in_run, rtl = items[i].a.fRTL ? -1 : 1; nchars_in_run = items[i+1].iCharPos - items[i].iCharPos; result = ScriptShape (context, &(uniscribe_font->cache), @@ -286,25 +289,36 @@ lgstring = Qnil; break; } - else if (result) + else if (result) /* Failure. */ { /* Can't shape this run - return results so far if any. */ break; } + else if (items[i].a.fNoGlyphIndex) + { + /* Glyph indices not supported by this font (or OS), means we + can't really do any meaningful shaping. */ + break; + } else { result = ScriptPlace (context, &(uniscribe_font->cache), glyphs, nglyphs, attributes, &(items[i].a), advances, offsets, &overall_metrics); - if (!result) + if (result == 0) /* Success. */ { - int j, nclusters; + int j, nclusters, from = 0, to = 0; + /* For tracking our mapping of characters to glyphs. + Special value -1 means not yet initialized, -2 means + we've run off the end. Anything else is an index + into the chars and clusters arrays. */ + int my_char = -1; for (j = 0; j < nglyphs; j++) { int lglyph_index = j + done_glyphs; Lisp_Object lglyph = LGSTRING_GLYPH (lgstring, lglyph_index); - GLYPHMETRICS metrics; + ABC char_metric; if (NILP (lglyph)) { @@ -312,93 +326,103 @@ LGSTRING_SET_GLYPH (lgstring, lglyph_index, lglyph); } LGLYPH_SET_CODE (lglyph, glyphs[j]); + + /* Detect clusters, for linking codes back to characters. */ + if (attributes[j].fClusterStart) + { + /* First time, set to appropriate end of run. */ + if (my_char == -1) + my_char = rtl > 0 ? 0 : nchars_in_run - 1; + else if (my_char >= 0) + my_char += rtl; + while (my_char >= 0 && my_char < nchars_in_run + && clusters[my_char] < j) + my_char += rtl; + + if (my_char < 0 || my_char >= nchars_in_run) + my_char = -2; + + /* FROM and TO as char indices. This produces + much better results at small font sizes than + earlier attempts at using glyph indices for + FROM and TO, but the output still isn't quite + right. For example, on the first South Asia + line of etc/HELLO, the third example + (Kannada) is missing the last glyph. This + seems to be caused by the fact that more + glyphs are produced than there are characters + in the output (other scripts on that line + result in the same or fewer glyphs). */ + if (my_char < 0) + from = to = rtl > 0 ? nchars_in_run - 1: 0; + else + { + int k; + from = my_char; + to = rtl > 0 ? nchars_in_run : 0; + for (k = my_char + rtl; k >= 0 && k < nchars_in_run; + k += rtl) + { + if (clusters[k] > j) + { + to = k; + break; + } + } + } + } + + if (my_char < 0 || clusters[my_char] > j) + { + /* No mapping. */ + LGLYPH_SET_CHAR (lglyph, 0); + LGLYPH_SET_FROM (lglyph, items[i].iCharPos + from); + LGLYPH_SET_TO (lglyph, items[i].iCharPos + to); + } + else + { + LGLYPH_SET_CHAR (lglyph, chars[items[i].iCharPos + + my_char]); + LGLYPH_SET_FROM (lglyph, items[i].iCharPos + from); + LGLYPH_SET_TO (lglyph, items[i].iCharPos + to); + } + + /* Metrics. */ LGLYPH_SET_WIDTH (lglyph, advances[j]); + LGLYPH_SET_ASCENT (lglyph, font->ascent); + LGLYPH_SET_DESCENT (lglyph, font->descent); + + result = ScriptGetGlyphABCWidth (context, + &(uniscribe_font->cache), + glyphs[j], &char_metric); + + if (result == 0) /* Success. */ + { + LGLYPH_SET_LBEARING (lglyph, char_metric.abcA); + LGLYPH_SET_RBEARING (lglyph, (char_metric.abcA + + char_metric.abcB)); + } + else + { + LGLYPH_SET_LBEARING (lglyph, 0); + LGLYPH_SET_RBEARING (lglyph, advances[j]); + } + if (offsets[j].du || offsets[j].dv) { Lisp_Object vec; - /* Convert from logical inches. */ - int dpi = FRAME_W32_DISPLAY_INFO (f)->resy; - int dx = (int)(offsets[j].du * dpi / 72.27 + 0.5); - int dy = (int)(offsets[j].dv * dpi / 72.27 + 0.5); vec = Fmake_vector (make_number (3), Qnil); - ASET (vec, 0, make_number (dx)); - ASET (vec, 1, make_number (dy)); + ASET (vec, 0, make_number (offsets[j].du)); + ASET (vec, 1, make_number (offsets[j].dv)); /* Based on what ftfont.c does... */ ASET (vec, 2, make_number (advances[j])); LGLYPH_SET_ADJUSTMENT (lglyph, vec); } else LGLYPH_SET_ADJUSTMENT (lglyph, Qnil); - - if (GetGlyphOutlineW (context, glyphs[j], - GGO_METRICS | GGO_GLYPH_INDEX, - &metrics, 0, NULL, &transform) - != GDI_ERROR) - { - LGLYPH_SET_LBEARING (lglyph, metrics.gmptGlyphOrigin.x); - LGLYPH_SET_RBEARING (lglyph, - metrics.gmBlackBoxX - + metrics.gmptGlyphOrigin.x); - LGLYPH_SET_ASCENT (lglyph, metrics.gmBlackBoxY); - LGLYPH_SET_DESCENT (lglyph, - (metrics.gmCellIncY - - metrics.gmptGlyphOrigin.y - - metrics.gmBlackBoxY)); - } - else - { - /* Defaults based on what we know from elsewhere. */ - LGLYPH_SET_LBEARING (lglyph, 0); - LGLYPH_SET_RBEARING (lglyph, advances[j]); - LGLYPH_SET_ASCENT (lglyph, font->ascent); - LGLYPH_SET_DESCENT (lglyph, font->descent); - } - } - - /* Set character codes as indicated in clusters. */ - for (j = 0; j < nchars_in_run - 1; j++) - { - int k, start, end; - wchar_t this_char = *(chars + items[i].iCharPos + j); - - start = clusters[i]; - end = clusters[i+1]; - if (start > end) - { - int temp = start; - start = end; - end = temp; - } - for (k = start; k < end; k++) - { - Lisp_Object lglyph = LGSTRING_GLYPH (lgstring, - done_glyphs + k); - LGLYPH_SET_CHAR (lglyph, this_char); - } - } - /* Last one until end (or beginning) of string. */ - { - int k, start, end; - wchar_t last_char = *(chars + items[i].iCharPos - + nchars_in_run - 1); - start = clusters[nchars_in_run - 1]; - end = nglyphs; - if (start < clusters[nchars_in_run - 2]) - { - end = start + 1; - start = 0; - } - for (k = start; k < end; k++) - { - Lisp_Object lglyph = LGSTRING_GLYPH (lgstring, - done_glyphs + k); - LGLYPH_SET_CHAR (lglyph, last_char); - } - } - } + } } } done_glyphs += nglyphs; - done_chars += nchars_in_run; } xfree (items); @@ -408,7 +432,7 @@ if (NILP (lgstring)) return Qnil; else - return make_number (done_chars); + return make_number (done_glyphs); } /* Uniscribe implementation of encode_char for font backend.