from just loadglyph you get metrics of it, so you don't need to render it at all to get how big it is... THe advance is how wide the overall character is; that doesn't count how big the bitmap actually is if you're trying to get the actual rendered size of the character... but rather the cell size the character is expected to be in.
status = FT_Load_Glyph( face , glyph_index , 0 // | FT_LOAD_FORCE_AUTOHINT ); int ascent = CEIL( face->size->metrics.ascender ); int descent = CEIL( face->size->metrics.descender ); int height = CEIL( face->size->metrics.height ); /* left offset */ , face->glyph->bitmap_left // left offset? /* width */ , (face->face_flags & FT_FACE_FLAG_SCALABLE) ? (face->glyph->linearHoriAdvance >> 16) : (face->glyph->advance.x >> 6) On Fri, Sep 15, 2017 at 10:30 AM, Daniel Plakhotich < daniel.plakhot...@gmail.com> wrote: > Hello, > > I writing a program that process characters in two passes: first it > queries exact sizes (bounding boxes) of all bitmaps, then, after some > manipulations, the actual rendering is performed. Since there will be > a high amount of characters (probably the whole Unicode range), no > bitmap caching is performed to minimize the memory usage. > > Obviously, the step 1 doesn't actually need the bitmaps - only their > sizes, so I want to speed up it as much as possible and reduce the > memory fragmentation. According to the documentation, there are two > options: > 1. Usual FT_Render_Glyph() > 2. Use custom outline processing (FT_Outline_Render(), etc.) via > custom span functions > > It seems like the custom span rendering is the best choice, since I > can use a dummy function to get the size, then use a fixed-size buffer > to render each bitmap. I don't want to rely on the undocumented > implementation details of the library and make assumptions about how > FreeType manages memory of the root glyph slot, so this solution looks > better than FT_Render_Glyph(). > > My question is: which is the fastest way to get size of a rendered > bitmap? Is there a better way except the mentioned two? > > As I understand, there is no way to get exact size of a bitmap without > actually rendering it, because of various hinting strategies. But I'm > a bit confused by The FreeType Tutorial (II. Managing Glyphs: 4. > Simple Text Rendering: Kerning and Centering), which says: > > > In general, the above function does not compute an exact bounding box of > a string! As soon as hinting is involved, glyph dimensions must be derived > from the resulting outlines. For anti-aliased pixmaps, FT_Outline_Get_BBox > then yields proper results. In case you need 1-bit monochrome bitmaps, it > is even necessary to actually render the glyphs because the rules for the > conversion from outline to bitmap can also be controlled by hinting > instructions. > > Why it says the rendering is necessary for 1-bit bitmaps, but not > 8-bit? Maybe I missed something? > > _______________________________________________ > Freetype mailing list > Freetype@nongnu.org > https://lists.nongnu.org/mailman/listinfo/freetype > _______________________________________________ Freetype mailing list Freetype@nongnu.org https://lists.nongnu.org/mailman/listinfo/freetype
