//---------------------------------------------------------------------------- // Anti-Grain Geometry (AGG) - Version 2.5 // A high quality rendering engine for C++ // Copyright (C) 2002-2006 Maxim Shemanarev // Contact: mcseem@antigrain.com // mcseemagg@yahoo.com // http://antigrain.com // // AGG is free software; you can redistribute it and/or // modify it under the terms of the GNU General Public License // as published by the Free Software Foundation; either version 2 // of the License, or (at your option) any later version. // // AGG is distributed in the hope that it will be useful, // but WITHOUT ANY WARRANTY; without even the implied warranty of // MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the // GNU General Public License for more details. // // You should have received a copy of the GNU General Public License // along with AGG; if not, write to the Free Software // Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, // MA 02110-1301, USA. //---------------------------------------------------------------------------- // // Adaptation for 32-bit screen coordinates (scanline32_u) has been sponsored by // Liberty Technology Systems, Inc., visit http://lib-sys.com // // Liberty Technology Systems, Inc. is the provider of // PostScript and PDF technology for software developers. // //---------------------------------------------------------------------------- #ifndef AGG_SCANLINE_U_INCLUDED #define AGG_SCANLINE_U_INCLUDED #include "agg_array.h" namespace agg { //=============================================================scanline_u8 // // Unpacked scanline container class // // This class is used to transfer data from a scanline rasterizer // to the rendering buffer. It's organized very simple. The class stores // information of horizontal spans to render it into a pixel-map buffer. // Each span has staring X, length, and an array of bytes that determine the // cover-values for each pixel. // Before using this class you should know the minimal and maximal pixel // coordinates of your scanline. The protocol of using is: // 1. reset(min_x, max_x) // 2. add_cell() / add_span() - accumulate scanline. // When forming one scanline the next X coordinate must be always greater // than the last stored one, i.e. it works only with ordered coordinates. // 3. Call finalize(y) and render the scanline. // 3. Call reset_spans() to prepare for the new scanline. // // 4. Rendering: // // Scanline provides an iterator class that allows you to extract // the spans and the cover values for each pixel. Be aware that clipping // has not been done yet, so you should perform it yourself. // Use scanline_u8::iterator to render spans: //------------------------------------------------------------------------- // // int y = sl.y(); // Y-coordinate of the scanline // // ************************************ // ...Perform vertical clipping here... // ************************************ // // scanline_u8::const_iterator span = sl.begin(); // // unsigned char* row = m_rbuf->row(y); // The the address of the beginning // // of the current row // // unsigned num_spans = sl.num_spans(); // Number of spans. It's guaranteed that // // num_spans is always greater than 0. // // do // { // const scanline_u8::cover_type* covers = // span->covers; // The array of the cover values // // int num_pix = span->len; // Number of pixels of the span. // // Always greater than 0, still it's // // better to use "int" instead of // // "unsigned" because it's more // // convenient for clipping // int x = span->x; // // ************************************** // ...Perform horizontal clipping here... // ...you have x, covers, and pix_count.. // ************************************** // // unsigned char* dst = row + x; // Calculate the start address of the row. // // In this case we assume a simple // // grayscale image 1-byte per pixel. // do // { // *dst++ = *covers++; // Hypotetical rendering. // } // while(--num_pix); // // ++span; // } // while(--num_spans); // num_spans cannot be 0, so this loop is quite safe //------------------------------------------------------------------------ // // The question is: why should we accumulate the whole scanline when we // could render just separate spans when they're ready? // That's because using the scanline is generally faster. When is consists // of more than one span the conditions for the processor cash system // are better, because switching between two different areas of memory // (that can be very large) occurs less frequently. //------------------------------------------------------------------------ class scanline_u8 { public: typedef scanline_u8 self_type; typedef int8u cover_type; typedef int16 coord_type; //-------------------------------------------------------------------- struct span { coord_type x; coord_type len; cover_type* covers; }; typedef span* iterator; typedef const span* const_iterator; //-------------------------------------------------------------------- scanline_u8() : m_min_x(0), m_last_x(0x7FFFFFF0), m_cur_span(0) {} //-------------------------------------------------------------------- void reset(int min_x, int max_x) { unsigned max_len = max_x - min_x + 2; if(max_len > m_spans.size()) { m_spans.resize(max_len); m_covers.resize(max_len); } m_last_x = 0x7FFFFFF0; m_min_x = min_x; m_cur_span = &m_spans[0]; } //-------------------------------------------------------------------- void add_cell(int x, unsigned cover) { x -= m_min_x; m_covers[x] = (cover_type)cover; if(x == m_last_x+1) { m_cur_span->len++; } else { m_cur_span++; m_cur_span->x = (coord_type)(x + m_min_x); m_cur_span->len = 1; m_cur_span->covers = &m_covers[x]; } m_last_x = x; } //-------------------------------------------------------------------- void add_cells(int x, unsigned len, const cover_type* covers) { x -= m_min_x; memcpy(&m_covers[x], covers, len * sizeof(cover_type)); if(x == m_last_x+1) { m_cur_span->len += (coord_type)len; } else { m_cur_span++; m_cur_span->x = (coord_type)(x + m_min_x); m_cur_span->len = (coord_type)len; m_cur_span->covers = &m_covers[x]; } m_last_x = x + len - 1; } //-------------------------------------------------------------------- void add_span(int x, unsigned len, unsigned cover) { x -= m_min_x; memset(&m_covers[x], cover, len); if(x == m_last_x+1) { m_cur_span->len += (coord_type)len; } else { m_cur_span++; m_cur_span->x = (coord_type)(x + m_min_x); m_cur_span->len = (coord_type)len; m_cur_span->covers = &m_covers[x]; } m_last_x = x + len - 1; } //-------------------------------------------------------------------- void finalize(int y) { m_y = y; } //-------------------------------------------------------------------- void reset_spans() { m_last_x = 0x7FFFFFF0; m_cur_span = &m_spans[0]; } //-------------------------------------------------------------------- int y() const { return m_y; } unsigned num_spans() const { return unsigned(m_cur_span - &m_spans[0]); } const_iterator begin() const { return &m_spans[1]; } iterator begin() { return &m_spans[1]; } private: scanline_u8(const self_type&); const self_type& operator = (const self_type&); private: int m_min_x; int m_last_x; int m_y; pod_array<cover_type> m_covers; pod_array<span> m_spans; span* m_cur_span; }; //==========================================================scanline_u8_am // // The scanline container with alpha-masking // //------------------------------------------------------------------------ template<class AlphaMask> class scanline_u8_am : public scanline_u8 { public: typedef scanline_u8 base_type; typedef AlphaMask alpha_mask_type; typedef base_type::cover_type cover_type; typedef base_type::coord_type coord_type; scanline_u8_am() : base_type(), m_alpha_mask(0) {} scanline_u8_am(const AlphaMask& am) : base_type(), m_alpha_mask(&am) {} //-------------------------------------------------------------------- void finalize(int span_y) { base_type::finalize(span_y); if(m_alpha_mask) { typename base_type::iterator span = base_type::begin(); unsigned count = base_type::num_spans(); do { m_alpha_mask->combine_hspan(span->x, base_type::y(), span->covers, span->len); ++span; } while(--count); } } private: const AlphaMask* m_alpha_mask; }; //===========================================================scanline32_u8 class scanline32_u8 { public: typedef scanline32_u8 self_type; typedef int8u cover_type; typedef int32 coord_type; //-------------------------------------------------------------------- struct span { span() {} span(coord_type x_, coord_type len_, cover_type* covers_) : x(x_), len(len_), covers(covers_) {} coord_type x; coord_type len; cover_type* covers; }; typedef pod_bvector<span, 4> span_array_type; //-------------------------------------------------------------------- class const_iterator { public: const_iterator(const span_array_type& spans) : m_spans(spans), m_span_idx(0) {} const span& operator*() const { return m_spans[m_span_idx]; } const span* operator->() const { return &m_spans[m_span_idx]; } void operator ++ () { ++m_span_idx; } private: const span_array_type& m_spans; unsigned m_span_idx; }; //-------------------------------------------------------------------- class iterator { public: iterator(span_array_type& spans) : m_spans(spans), m_span_idx(0) {} span& operator*() { return m_spans[m_span_idx]; } span* operator->() { return &m_spans[m_span_idx]; } void operator ++ () { ++m_span_idx; } private: span_array_type& m_spans; unsigned m_span_idx; }; //-------------------------------------------------------------------- scanline32_u8() : m_min_x(0), m_last_x(0x7FFFFFF0), m_covers() {} //-------------------------------------------------------------------- void reset(int min_x, int max_x) { unsigned max_len = max_x - min_x + 2; if(max_len > m_covers.size()) { m_covers.resize(max_len); } m_last_x = 0x7FFFFFF0; m_min_x = min_x; m_spans.remove_all(); } //-------------------------------------------------------------------- void add_cell(int x, unsigned cover) { x -= m_min_x; m_covers[x] = cover_type(cover); if(x == m_last_x+1) { m_spans.last().len++; } else { m_spans.add(span(coord_type(x + m_min_x), 1, &m_covers[x])); } m_last_x = x; } //-------------------------------------------------------------------- void add_cells(int x, unsigned len, const cover_type* covers) { x -= m_min_x; memcpy(&m_covers[x], covers, len * sizeof(cover_type)); if(x == m_last_x+1) { m_spans.last().len += coord_type(len); } else { m_spans.add(span(coord_type(x + m_min_x), coord_type(len), &m_covers[x])); } m_last_x = x + len - 1; } //-------------------------------------------------------------------- void add_span(int x, unsigned len, unsigned cover) { x -= m_min_x; memset(&m_covers[x], cover, len); if(x == m_last_x+1) { m_spans.last().len += coord_type(len); } else { m_spans.add(span(coord_type(x + m_min_x), coord_type(len), &m_covers[x])); } m_last_x = x + len - 1; } //-------------------------------------------------------------------- void finalize(int y) { m_y = y; } //-------------------------------------------------------------------- void reset_spans() { m_last_x = 0x7FFFFFF0; m_spans.remove_all(); } //-------------------------------------------------------------------- int y() const { return m_y; } unsigned num_spans() const { return m_spans.size(); } const_iterator begin() const { return const_iterator(m_spans); } iterator begin() { return iterator(m_spans); } private: scanline32_u8(const self_type&); const self_type& operator = (const self_type&); private: int m_min_x; int m_last_x; int m_y; pod_array<cover_type> m_covers; span_array_type m_spans; }; //========================================================scanline32_u8_am // // The scanline container with alpha-masking // //------------------------------------------------------------------------ template<class AlphaMask> class scanline32_u8_am : public scanline32_u8 { public: typedef scanline32_u8 base_type; typedef AlphaMask alpha_mask_type; typedef base_type::cover_type cover_type; typedef base_type::coord_type coord_type; scanline32_u8_am() : base_type(), m_alpha_mask(0) {} scanline32_u8_am(const AlphaMask& am) : base_type(), m_alpha_mask(&am) {} //-------------------------------------------------------------------- void finalize(int span_y) { base_type::finalize(span_y); if(m_alpha_mask) { typename base_type::iterator span = base_type::begin(); unsigned count = base_type::num_spans(); do { m_alpha_mask->combine_hspan(span->x, base_type::y(), span->covers, span->len); ++span; } while(--count); } } private: const AlphaMask* m_alpha_mask; }; } #endif |