// Copyright (C) 2015-2016 ChaosForge Ltd
// http://chaosforge.org/
//
// This file is part of Nova libraries. 
// For conditions of distribution and use, see copying.txt file in root folder.

#include "nv/image/png_writer.hh"

#include "nv/image/miniz.hh"

// TODO: remove
#include <stdarg.h>  

using namespace nv;

#if NV_COMPILER == NV_CLANG
#pragma clang diagnostic ignored "-Wunused-macros"
#pragma clang diagnostic ignored "-Wold-style-cast"
#pragma clang diagnostic ignored "-Wsign-conversion"
#pragma clang diagnostic ignored "-Wreserved-id-macro"
#pragma clang diagnostic ignored "-Wmissing-prototypes"
#define NULL 0
#endif

enum
{
	STBI_default = 0, // only used for req_comp

	STBI_grey = 1,
	STBI_grey_alpha = 2,
	STBI_rgb = 3,
	STBI_rgb_alpha = 4
};

#define STBI_MALLOC(sz)     nvmalloc(sz)
#define STBI_REALLOC(p,sz)  nvrealloc(p,sz)
#define STBI_FREE(p)        nvfree(p)
#define STBI_MEMMOVE(d,s,c) nvmemmove(d,s,c)

typedef void stbi_write_func( void *context, void *data, int size );

typedef struct
{
	stbi_write_func *func;
	void *context;
} stbi__write_context;

// initialize a callback-based context
static void stbi__start_write_callbacks( stbi__write_context *s, stbi_write_func *c, void *context )
{
	s->func = c;
	s->context = context;
}

typedef unsigned int stbiw_uint32;
typedef int stb_image_write_test[sizeof( stbiw_uint32 ) == 4 ? 1 : -1];

template < typename T >
inline uchar8 byte_cast( T x )
{
	return uchar8( ( x ) & 255 );
}

#define stbi__err(x,y)  0
#define stbi__errpuc(x,y)  ((unsigned char *)(size_t) (stbi__err(x,y)?NULL:NULL))

static void stbiw__writefv( stbi__write_context *s, const char *fmt, va_list v )
{
	while ( *fmt )
	{
		switch ( *fmt++ )
		{
		case ' ': break;
		case '1': { unsigned char x = byte_cast( va_arg( v, int ) );
			s->func( s->context, &x, 1 );
			break; }
		case '2': { int x = va_arg( v, int );
			unsigned char b[2];
			b[0] = byte_cast( x );
			b[1] = byte_cast( x >> 8 );
			s->func( s->context, b, 2 );
			break; }
		case '4': { stbiw_uint32 x = va_arg( v, int );
			unsigned char b[4];
			b[0] = byte_cast( x );
			b[1] = byte_cast( x >> 8 );
			b[2] = byte_cast( x >> 16 );
			b[3] = byte_cast( x >> 24 );
			s->func( s->context, b, 4 );
			break; }
		default:
			byte_cast( 0 );
			return;
		}
	}
}

static void stbiw__writef( stbi__write_context *s, const char *fmt, ... )
{
	va_list v;
	va_start( v, fmt );
	stbiw__writefv( s, fmt, v );
	va_end( v );
}

static void stbiw__write3( stbi__write_context *s, unsigned char a, unsigned char b, unsigned char c )
{
	unsigned char arr[3];
	arr[0] = a, arr[1] = b, arr[2] = c;
	s->func( s->context, arr, 3 );
}

static void stbiw__write_pixel( stbi__write_context *s, int rgb_dir, int comp, int write_alpha, int expand_mono, unsigned char *d )
{
	unsigned char bg[3] = { 255, 0, 255 }, px[3];
	int k;

	if ( write_alpha < 0 )
		s->func( s->context, &d[comp - 1], 1 );

	switch ( comp )
	{
	case 1:
		s->func( s->context, d, 1 );
		break;
	case 2:
		if ( expand_mono )
			stbiw__write3( s, d[0], d[0], d[0] ); // monochrome bmp
		else
			s->func( s->context, d, 1 );  // monochrome TGA
		break;
	case 4:
		if ( !write_alpha )
		{
			// composite against pink background
			for ( k = 0; k < 3; ++k )
				px[k] = bg[k] + ( ( d[k] - bg[k] ) * d[3] ) / 255;
			stbiw__write3( s, px[1 - rgb_dir], px[1], px[1 + rgb_dir] );
			break;
		}
		/* FALLTHROUGH */
	case 3:
		stbiw__write3( s, d[1 - rgb_dir], d[1], d[1 + rgb_dir] );
		break;
	}
	if ( write_alpha > 0 )
		s->func( s->context, &d[comp - 1], 1 );
}

static void stbiw__write_pixels( stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, void *data, int write_alpha, int scanline_pad, int expand_mono )
{
	stbiw_uint32 zero = 0;
	int i, j, j_end;

	if ( y <= 0 )
		return;

	if ( vdir < 0 )
		j_end = -1, j = y - 1;
	else
		j_end = y, j = 0;

	for ( ; j != j_end; j += vdir )
	{
		for ( i = 0; i < x; ++i )
		{
			unsigned char *d = (unsigned char *)data + ( j*x + i )*comp;
			stbiw__write_pixel( s, rgb_dir, comp, write_alpha, expand_mono, d );
		}
		s->func( s->context, &zero, scanline_pad );
	}
}

static int stbiw__outfile( stbi__write_context *s, int rgb_dir, int vdir, int x, int y, int comp, int expand_mono, void *data, int alpha, int pad, const char *fmt, ... )
{
	if ( y < 0 || x < 0 )
	{
		return 0;
	}
	else
	{
		va_list v;
		va_start( v, fmt );
		stbiw__writefv( s, fmt, v );
		va_end( v );
		stbiw__write_pixels( s, rgb_dir, vdir, x, y, comp, data, alpha, pad, expand_mono );
		return 1;
	}
}


static unsigned int stbiw__crc32( unsigned char *buffer, int len )
{
	static unsigned int crc_table[256] =
	{
		0x00000000, 0x77073096, 0xEE0E612C, 0x990951BA, 0x076DC419, 0x706AF48F, 0xE963A535, 0x9E6495A3,
		0x0eDB8832, 0x79DCB8A4, 0xE0D5E91E, 0x97D2D988, 0x09B64C2B, 0x7EB17CBD, 0xE7B82D07, 0x90BF1D91,
		0x1DB71064, 0x6AB020F2, 0xF3B97148, 0x84BE41DE, 0x1ADAD47D, 0x6DDDE4EB, 0xF4D4B551, 0x83D385C7,
		0x136C9856, 0x646BA8C0, 0xFD62F97A, 0x8A65C9EC, 0x14015C4F, 0x63066CD9, 0xFA0F3D63, 0x8D080DF5,
		0x3B6E20C8, 0x4C69105E, 0xD56041E4, 0xA2677172, 0x3C03E4D1, 0x4B04D447, 0xD20D85FD, 0xA50AB56B,
		0x35B5A8FA, 0x42B2986C, 0xDBBBC9D6, 0xACBCF940, 0x32D86CE3, 0x45DF5C75, 0xDCD60DCF, 0xABD13D59,
		0x26D930AC, 0x51DE003A, 0xC8D75180, 0xBFD06116, 0x21B4F4B5, 0x56B3C423, 0xCFBA9599, 0xB8BDA50F,
		0x2802B89E, 0x5F058808, 0xC60CD9B2, 0xB10BE924, 0x2F6F7C87, 0x58684C11, 0xC1611DAB, 0xB6662D3D,
		0x76DC4190, 0x01DB7106, 0x98D220BC, 0xEFD5102A, 0x71B18589, 0x06B6B51F, 0x9FBFE4A5, 0xE8B8D433,
		0x7807C9A2, 0x0F00F934, 0x9609A88E, 0xE10E9818, 0x7F6A0DBB, 0x086D3D2D, 0x91646C97, 0xE6635C01,
		0x6B6B51F4, 0x1C6C6162, 0x856530D8, 0xF262004E, 0x6C0695ED, 0x1B01A57B, 0x8208F4C1, 0xF50FC457,
		0x65B0D9C6, 0x12B7E950, 0x8BBEB8EA, 0xFCB9887C, 0x62DD1DDF, 0x15DA2D49, 0x8CD37CF3, 0xFBD44C65,
		0x4DB26158, 0x3AB551CE, 0xA3BC0074, 0xD4BB30E2, 0x4ADFA541, 0x3DD895D7, 0xA4D1C46D, 0xD3D6F4FB,
		0x4369E96A, 0x346ED9FC, 0xAD678846, 0xDA60B8D0, 0x44042D73, 0x33031DE5, 0xAA0A4C5F, 0xDD0D7CC9,
		0x5005713C, 0x270241AA, 0xBE0B1010, 0xC90C2086, 0x5768B525, 0x206F85B3, 0xB966D409, 0xCE61E49F,
		0x5EDEF90E, 0x29D9C998, 0xB0D09822, 0xC7D7A8B4, 0x59B33D17, 0x2EB40D81, 0xB7BD5C3B, 0xC0BA6CAD,
		0xEDB88320, 0x9ABFB3B6, 0x03B6E20C, 0x74B1D29A, 0xEAD54739, 0x9DD277AF, 0x04DB2615, 0x73DC1683,
		0xE3630B12, 0x94643B84, 0x0D6D6A3E, 0x7A6A5AA8, 0xE40ECF0B, 0x9309FF9D, 0x0A00AE27, 0x7D079EB1,
		0xF00F9344, 0x8708A3D2, 0x1E01F268, 0x6906C2FE, 0xF762575D, 0x806567CB, 0x196C3671, 0x6E6B06E7,
		0xFED41B76, 0x89D32BE0, 0x10DA7A5A, 0x67DD4ACC, 0xF9B9DF6F, 0x8EBEEFF9, 0x17B7BE43, 0x60B08ED5,
		0xD6D6A3E8, 0xA1D1937E, 0x38D8C2C4, 0x4FDFF252, 0xD1BB67F1, 0xA6BC5767, 0x3FB506DD, 0x48B2364B,
		0xD80D2BDA, 0xAF0A1B4C, 0x36034AF6, 0x41047A60, 0xDF60EFC3, 0xA867DF55, 0x316E8EEF, 0x4669BE79,
		0xCB61B38C, 0xBC66831A, 0x256FD2A0, 0x5268E236, 0xCC0C7795, 0xBB0B4703, 0x220216B9, 0x5505262F,
		0xC5BA3BBE, 0xB2BD0B28, 0x2BB45A92, 0x5CB36A04, 0xC2D7FFA7, 0xB5D0CF31, 0x2CD99E8B, 0x5BDEAE1D,
		0x9B64C2B0, 0xEC63F226, 0x756AA39C, 0x026D930A, 0x9C0906A9, 0xEB0E363F, 0x72076785, 0x05005713,
		0x95BF4A82, 0xE2B87A14, 0x7BB12BAE, 0x0CB61B38, 0x92D28E9B, 0xE5D5BE0D, 0x7CDCEFB7, 0x0BDBDF21,
		0x86D3D2D4, 0xF1D4E242, 0x68DDB3F8, 0x1FDA836E, 0x81BE16CD, 0xF6B9265B, 0x6FB077E1, 0x18B74777,
		0x88085AE6, 0xFF0F6A70, 0x66063BCA, 0x11010B5C, 0x8F659EFF, 0xF862AE69, 0x616BFFD3, 0x166CCF45,
		0xA00AE278, 0xD70DD2EE, 0x4E048354, 0x3903B3C2, 0xA7672661, 0xD06016F7, 0x4969474D, 0x3E6E77DB,
		0xAED16A4A, 0xD9D65ADC, 0x40DF0B66, 0x37D83BF0, 0xA9BCAE53, 0xDEBB9EC5, 0x47B2CF7F, 0x30B5FFE9,
		0xBDBDF21C, 0xCABAC28A, 0x53B39330, 0x24B4A3A6, 0xBAD03605, 0xCDD70693, 0x54DE5729, 0x23D967BF,
		0xB3667A2E, 0xC4614AB8, 0x5D681B02, 0x2A6F2B94, 0xB40BBE37, 0xC30C8EA1, 0x5A05DF1B, 0x2D02EF8D
	};

	unsigned int crc = ~0u;
	int i;
	for ( i = 0; i < len; ++i )
		crc = ( crc >> 8 ) ^ crc_table[buffer[i] ^ ( crc & 0xff )];
	return ~crc;
}

#define stbiw__wpng4(o,a,b,c,d) ((o)[0]=byte_cast(a),(o)[1]=byte_cast(b),(o)[2]=byte_cast(c),(o)[3]=byte_cast(d),(o)+=4)
#define stbiw__wp32(data,v) stbiw__wpng4(data, (v)>>24,(v)>>16,(v)>>8,(v));
#define stbiw__wptag(data,s) stbiw__wpng4(data, s[0],s[1],s[2],s[3])

static void stbiw__wpcrc( unsigned char **data, int len )
{
	unsigned int crc = stbiw__crc32( *data - len - 4, len + 4 );
	stbiw__wp32( *data, crc );
}

static unsigned char stbiw__paeth( int a, int b, int c )
{
	int p = a + b - c, pa = abs( p - a ), pb = abs( p - b ), pc = abs( p - c );
	if ( pa <= pb && pa <= pc ) return byte_cast( a );
	if ( pb <= pc ) return byte_cast( b );
	return byte_cast( c );
}

unsigned char * stbi_zlib_compress( unsigned char *data, int data_len, int *out_len, int quality )
{
	unsigned long expected = nv::miniz_bound( data_len );
	unsigned char* result = (unsigned char *)STBI_MALLOC( expected );
	unsigned long rlen = expected;
	nv::miniz_compress( result, &rlen, data, data_len, 6 );
	*out_len = rlen;
	return result;
}

unsigned char *stbi_write_png_to_mem( const unsigned char *pixels, int stride_bytes, int x, int y, int n, int *out_len )
{
	int ctype[5] = { -1, 0, 4, 2, 6 };
	unsigned char sig[8] = { 137,80,78,71,13,10,26,10 };
	unsigned char *out, *o, *filt, *zlib;
	signed char *line_buffer;
	int i, j, k, p, zlen;

	if ( stride_bytes == 0 )
		stride_bytes = x * n;

	filt = (unsigned char *)STBI_MALLOC( ( x*n + 1 ) * y ); if ( !filt ) return 0;
	line_buffer = (signed char *)STBI_MALLOC( x * n ); if ( !line_buffer ) { STBI_FREE( filt ); return 0; }

	bool flip_y = true;

	for ( j = 0; j < y; ++j )
	{
		int line = flip_y ? y - j - 1 : j;
		int step = flip_y ? -stride_bytes : stride_bytes;
		static int mapping[] = { 0,1,2,3,4 };
		static int firstmap[] = { 0,1,0,5,6 };
		int *mymap = j ? mapping : firstmap;
		int best = 0, bestval = 0x7fffffff;
		for ( p = 0; p < 2; ++p )
		{
			for ( k = p ? best : 0; k < 5; ++k )
			{
				int type = mymap[k], est = 0;
				const unsigned char *z = pixels + stride_bytes*line;
				for ( i = 0; i < n; ++i )
					switch ( type )
					{
					case 0: line_buffer[i] = z[i]; break;
					case 1: line_buffer[i] = z[i]; break;
					case 2: line_buffer[i] = z[i] - z[i - step]; break;
					case 3: line_buffer[i] = z[i] - ( z[i - step] >> 1 ); break;
					case 4: line_buffer[i] = (signed char)( z[i] - stbiw__paeth( 0, z[i - step], 0 ) ); break;
					case 5: line_buffer[i] = z[i]; break;
					case 6: line_buffer[i] = z[i]; break;
					}
				for ( i = n; i < x*n; ++i )
				{
					switch ( type )
					{
					case 0: line_buffer[i] = z[i]; break;
					case 1: line_buffer[i] = z[i] - z[i - n]; break;
					case 2: line_buffer[i] = z[i] - z[i - step]; break;
					case 3: line_buffer[i] = z[i] - ( ( z[i - n] + z[i - step] ) >> 1 ); break;
					case 4: line_buffer[i] = z[i] - stbiw__paeth( z[i - n], z[i - step], z[i - step - n] ); break;
					case 5: line_buffer[i] = z[i] - ( z[i - n] >> 1 ); break;
					case 6: line_buffer[i] = z[i] - stbiw__paeth( z[i - n], 0, 0 ); break;
					}
				}
				if ( p ) break;
				for ( i = 0; i < x*n; ++i )
					est += abs( (signed char)line_buffer[i] );
				if ( est < bestval ) { bestval = est; best = k; }
			}
		}
		// when we get here, best contains the filter type, and line_buffer contains the data
		//line = j;
		filt[j*( x*n + 1 )] = (unsigned char)best;
		STBI_MEMMOVE( filt + j*( x*n + 1 ) + 1, line_buffer, x*n );
	}
	STBI_FREE( line_buffer );
	zlib = stbi_zlib_compress( filt, y*( x*n + 1 ), &zlen, 8 ); // increase 8 to get smaller but use more memory
	STBI_FREE( filt );
	if ( !zlib ) return 0;

	// each tag requires 12 bytes of overhead
	out = (unsigned char *)STBI_MALLOC( 8 + 12 + 13 + 12 + zlen + 12 );
	if ( !out ) return 0;
	*out_len = 8 + 12 + 13 + 12 + zlen + 12;

	o = out;
	STBI_MEMMOVE( o, sig, 8 ); o += 8;
	stbiw__wp32( o, 13 ); // header length
	stbiw__wptag( o, "IHDR" );
	stbiw__wp32( o, x );
	stbiw__wp32( o, y );
	*o++ = 8;
	*o++ = byte_cast( ctype[n] );
	*o++ = 0;
	*o++ = 0;
	*o++ = 0;
	stbiw__wpcrc( &o, 13 );

	stbiw__wp32( o, zlen );
	stbiw__wptag( o, "IDAT" );
	STBI_MEMMOVE( o, zlib, zlen );
	o += zlen;
	STBI_FREE( zlib );
	stbiw__wpcrc( &o, zlen );

	stbiw__wp32( o, 0 );
	stbiw__wptag( o, "IEND" );
	stbiw__wpcrc( &o, 0 );

	NV_ASSERT( o == out + *out_len, "wrong output" );

	return out;
}

bool nv::png_writer::save( stream& s, image_data* data )
{
	ivec2 sz = data->get_size();
	int len = 0;
	unsigned char* out = stbi_write_png_to_mem( data->get_data(), 0, sz.x, sz.y, data->get_depth(), &len );
	if ( !out ) return false;
	s.write( out, 1, len );
	STBI_FREE( out );
	return true;
}