summaryrefslogtreecommitdiff
path: root/core/src/fxge/agg/agg23/agg_path_storage.h
diff options
context:
space:
mode:
Diffstat (limited to 'core/src/fxge/agg/agg23/agg_path_storage.h')
-rw-r--r--core/src/fxge/agg/agg23/agg_path_storage.h172
1 files changed, 172 insertions, 0 deletions
diff --git a/core/src/fxge/agg/agg23/agg_path_storage.h b/core/src/fxge/agg/agg23/agg_path_storage.h
new file mode 100644
index 0000000000..64ea0b4753
--- /dev/null
+++ b/core/src/fxge/agg/agg23/agg_path_storage.h
@@ -0,0 +1,172 @@
+
+//----------------------------------------------------------------------------
+// Anti-Grain Geometry - Version 2.3
+// Copyright (C) 2002-2005 Maxim Shemanarev (http://www.antigrain.com)
+//
+// Permission to copy, use, modify, sell and distribute this software
+// is granted provided this copyright notice appears in all copies.
+// This software is provided "as is" without express or implied
+// warranty, and with no claim as to its suitability for any purpose.
+//
+//----------------------------------------------------------------------------
+// Contact: mcseem@antigrain.com
+// mcseemagg@yahoo.com
+// http://www.antigrain.com
+//----------------------------------------------------------------------------
+#ifndef AGG_PATH_STORAGE_INCLUDED
+#define AGG_PATH_STORAGE_INCLUDED
+#include "agg_basics.h"
+namespace agg
+{
+class path_storage : public CFX_Object
+{
+ enum block_scale_e {
+ block_shift = 8,
+ block_size = 1 << block_shift,
+ block_mask = block_size - 1,
+ block_pool = 256
+ };
+public:
+ class vertex_source : public CFX_Object
+ {
+ public:
+ vertex_source() {}
+ vertex_source(const path_storage& p) : m_path(&p), m_vertex_idx(0) {}
+ void rewind(unsigned path_id)
+ {
+ m_vertex_idx = path_id;
+ }
+ unsigned vertex(FX_FLOAT* x, FX_FLOAT* y)
+ {
+ return (m_vertex_idx < m_path->total_vertices()) ?
+ m_path->vertex(m_vertex_idx++, x, y) :
+ path_cmd_stop;
+ }
+ private:
+ const path_storage* m_path;
+ unsigned m_vertex_idx;
+ };
+ ~path_storage();
+ path_storage();
+ unsigned last_vertex(FX_FLOAT* x, FX_FLOAT* y) const;
+ unsigned prev_vertex(FX_FLOAT* x, FX_FLOAT* y) const;
+ void move_to(FX_FLOAT x, FX_FLOAT y);
+ void line_to(FX_FLOAT x, FX_FLOAT y);
+ void curve4(FX_FLOAT x_ctrl1, FX_FLOAT y_ctrl1,
+ FX_FLOAT x_ctrl2, FX_FLOAT y_ctrl2,
+ FX_FLOAT x_to, FX_FLOAT y_to);
+ template<class VertexSource>
+ void add_path(VertexSource& vs,
+ unsigned path_id = 0,
+ bool solid_path = true)
+ {
+ FX_FLOAT x, y;
+ unsigned cmd;
+ vs.rewind(path_id);
+ while(!is_stop(cmd = vs.vertex(&x, &y))) {
+ if(is_move_to(cmd) && solid_path && m_total_vertices) {
+ cmd = path_cmd_line_to;
+ }
+ add_vertex(x, y, cmd);
+ }
+ }
+ template<class VertexSource>
+ void add_path_curve(VertexSource& vs,
+ unsigned path_id = 0,
+ bool solid_path = true)
+ {
+ FX_FLOAT x, y;
+ unsigned cmd;
+ int flag;
+ vs.rewind(path_id);
+ while(!is_stop(cmd = vs.vertex_curve_flag(&x, &y, flag))) {
+ if(is_move_to(cmd) && solid_path && m_total_vertices) {
+ cmd = path_cmd_line_to | flag;
+ }
+ add_vertex(x, y, cmd | flag);
+ }
+ }
+ unsigned total_vertices() const
+ {
+ return m_total_vertices;
+ }
+ unsigned vertex(unsigned idx, FX_FLOAT* x, FX_FLOAT* y) const
+ {
+ unsigned nb = idx >> block_shift;
+ const FX_FLOAT* pv = m_coord_blocks[nb] + ((idx & block_mask) << 1);
+ *x = *pv++;
+ *y = *pv;
+ return m_cmd_blocks[nb][idx & block_mask];
+ }
+ unsigned command(unsigned idx) const
+ {
+ return m_cmd_blocks[idx >> block_shift][idx & block_mask];
+ }
+ unsigned getflag(unsigned idx) const
+ {
+ return m_cmd_blocks[idx >> block_shift][idx & block_mask] & path_flags_jr;
+ }
+ void rewind(unsigned path_id);
+ unsigned vertex(FX_FLOAT* x, FX_FLOAT* y);
+ void add_vertex(FX_FLOAT x, FX_FLOAT y, unsigned cmd);
+ void end_poly();
+private:
+ void allocate_block(unsigned nb);
+ unsigned char* storage_ptrs(FX_FLOAT** xy_ptr);
+private:
+ unsigned m_total_vertices;
+ unsigned m_total_blocks;
+ unsigned m_max_blocks;
+ FX_FLOAT** m_coord_blocks;
+ unsigned char** m_cmd_blocks;
+ unsigned m_iterator;
+};
+inline unsigned path_storage::vertex(FX_FLOAT* x, FX_FLOAT* y)
+{
+ if(m_iterator >= m_total_vertices) {
+ return path_cmd_stop;
+ }
+ return vertex(m_iterator++, x, y);
+}
+inline unsigned path_storage::prev_vertex(FX_FLOAT* x, FX_FLOAT* y) const
+{
+ if(m_total_vertices > 1) {
+ return vertex(m_total_vertices - 2, x, y);
+ }
+ return path_cmd_stop;
+}
+inline unsigned path_storage::last_vertex(FX_FLOAT* x, FX_FLOAT* y) const
+{
+ if(m_total_vertices) {
+ return vertex(m_total_vertices - 1, x, y);
+ }
+ return path_cmd_stop;
+}
+inline unsigned char* path_storage::storage_ptrs(FX_FLOAT** xy_ptr)
+{
+ unsigned nb = m_total_vertices >> block_shift;
+ if(nb >= m_total_blocks) {
+ allocate_block(nb);
+ }
+ *xy_ptr = m_coord_blocks[nb] + ((m_total_vertices & block_mask) << 1);
+ return m_cmd_blocks[nb] + (m_total_vertices & block_mask);
+}
+inline void path_storage::add_vertex(FX_FLOAT x, FX_FLOAT y, unsigned cmd)
+{
+ FX_FLOAT* coord_ptr = 0;
+ unsigned char* cmd_ptr = storage_ptrs(&coord_ptr);
+ *cmd_ptr = (unsigned char)cmd;
+ *coord_ptr++ = x;
+ *coord_ptr = y;
+ m_total_vertices++;
+}
+inline void path_storage::move_to(FX_FLOAT x, FX_FLOAT y)
+{
+ add_vertex(x, y, path_cmd_move_to);
+}
+inline void path_storage::line_to(FX_FLOAT x, FX_FLOAT y)
+{
+ add_vertex(x, y, path_cmd_line_to);
+}
+}
+#endif