ADT/v18
ADT files contain terrain and object information for map tiles. They have a chunked structure just like the WDT files.
A map tile is split up into 16x16 = 256 map chunks. (not the same as file chunks, although each map chunk will have its own file chunk :) ) So there will be a few initial data chunks to specify textures, objects, models, etc. followed by 256 MCNK (mapchunk) chunks :) Each MCNK chunk has a small header of its own, and additional chunks within its data block, following the same id-size-data format.
An important note about the coordinate system used
Wow's main coordinate system is really weird but understanding it is very important in order to correctly interpret the ADT files. The best way to explain it is with screenshots: [[1]], [[2]].
It's important to remember that:
- Unlike most coordinate systems Z is used for the height, Y is used for the width and X is used for the depth.
- Y is "reversed". The more you go to the right (if north is up), the more it decreases.
- The center of the axis is in the center of the map.
- The top-left corner of the map has X = 17066, Y = 17066
- The bottom-right corner of the map has X = -17066, Y = -17066
- The bottom-left corner of the map has X = -17006, Y = 17066
- The top-right corner of the map has X = 17006, Y = -17066
Just to be absolutely clear, assuming you playing a character that is not flying or swimming and is facing north:
- Forward = D3DXVECTOR3(1, 0, 0);
- Right = D3DXVECTOR3(0, -1, 0);
- Up = D3DXVECTOR3(0, 0, 1);
This is the coordinate system used internally in all of the network packets and on most chunks in ADT files. I believe (but I'm not sure) that some chunks in this file use a different coordinate system (in particular for objects like M2's or WMO's that don't have an absolute position but only a relative one).
MODF and MDDF do indeed use a different coordinate system, even for the absolute Position. (Anyone clarify on the reasoning behind this?)
Unfortunately though whoever wrote this article decided to use yet another coordinate system to explain things, different from Wow's.. So it's kind of a mess!
Map size, blocks, chunks
Introduction
All maps are divided into 64x64 blocks for a total of 4096 (some of which may be unused). Each block are divided into 16x16 chunks (not to be confused with for example the file chunks, such as the "MHDR" chunk.. Completely different thing!). While like I said blocks can be unused, each block will always use all of its 16x16 chunks.
Map size
Each block is 533.33333 yards long (yes, strange number, but that's the way it is). Since like we said above the map is divided into 64x64 blocks the total X and Y of the map will be 34133.33312 yards (but remember that since the center of the axis is at the center of the map, the maximum X and Y will be 17066.66656).
Since each block has 16x16 chunks, the size of a chunk will be 33.3333 yards.
ADT files and blocks
There is an .adt file for each existing block. If a block is unused it won't have an .adt file. The file will be: World/Maps/<InternalMapName>/<InternalMapName>_<BlockY>_<BlockX>.adt.
- <InternalMapName> - The second field in Map.dbc
- <BlockY> - obtained with the following formula: floor((32 - (y / 533.33333)))
- <BlockX> - obtained with the following formula: floor((32 - (x / 533.33333)))
MHDR chunk
- Header chunk.
Contains offsets (relative to 0x14) for some other chunks that appear in the file. This information is very important as WoW only takes this for parsing the files.
struct SMAreaHeader // 03-29-2005 By ObscuR, --schlumpf_ 02:35, 8 August 2009 (CEST) { /*000h*/ UINT32 flags; // &1: MFBO, &2: unknown. in some Northrend ones. /*004h*/ MCIN* mcin; /*008h*/ MTEX* mtex; /*00Ch*/ MMDX* mmdx; /*010h*/ MMID* mmid; /*014h*/ MWMO* mwmo; /*018h*/ MWID* mwid; /*01Ch*/ MDDF* mddf; /*020h*/ MODF* modf; /*024h*/ MFBO* mfbo; // tbc, wotlk; only when flags&1 /*028h*/ MH2O* mh2o; // wotlk /*02Ch*/ MTFX* mtfx; // wotlk /*030h*/ UINT32 pad4; /*034h*/ UINT32 pad5; /*038h*/ UINT32 pad6; /*03Ch*/ UINT32 pad7; /*040h*/ };
MCIN chunk
- Index for MCNK chunks. Contains 256 records of 16 bytes, which have the following format:
struct SMChunkInfo // 03-29-2005 By ObscuR { /*000h*/ MCNK* mcnk; // absolute offset. /*004h*/ UINT32 size; // the size of the MCNK chunk, this is refering to. /*008h*/ UINT32 Unused_flags; // these two are always 0. only set in the client. /*00Ch*/ UINT32 Unused_asyncId; /*010h*/ };
MTEX chunk
- List of textures used by the terrain in this map tile.
A contiguous block of zero-terminated strings, that are complete filenames with paths. The textures will later be identified by their position in this list.
MMDX chunk
- List of filenames for M2 models that appear in this map tile. A contiguous block of zero-terminated strings.
MMID chunk
- Lists the relative offsets of string beginnings in the above MMDX chunk.
(sort of redundant)One 32-bit integer per offset.
- This will be referenced in the offsets in MDDF --Cromon 16:38, 28 August 2009 (CEST)
MWMO chunk
- List of filenames for WMOs (world map objects) that appear in this map tile. A contiguous block of zero-terminated strings.
MWID chunk
- Lists the relative offsets of string beginnings in the above MWMO chunk.
(again, redundant)One 32-bit integer per offset.
- Again referenced in MODF
MDDF chunk
- Placement information for doodads (M2 models). 36 bytes per model instance.
- WoW does not take the entries here and render those! It render per chunk using the nDoodadRefs and MCRF --Cromon
Offset Type Description 0x00 uint32 ID (index in the MMID list) 0x04 uint32 unique identifier for this instance 0x08 Vec3F Position (X,Y,Z) 0x14 Vec3F Orientation (A,B,C) 0x20 uint16 scale factor * 1024 (it's scale / 1024 for the other way around) 0x22 uint16 flags, known: &1 (sets the internal flags to 0x801 instead of 0x1. (WDOODADDEF.var0xC))
struct SMDoodadDef // 03-31-2005 By ObscuR { /*000h*/ UINT32 nameId; /*004h*/ UINT32 uniqueId; /*008h*/ float pos[3]; /*00Ch*/ /*010h*/ /*014h*/ float rot[3]; /*018h*/ /*01Ch*/ /*020h*/ UINT16 scale; /*022h*/ UINT16 flags; /*024h*/ };
Flags:
&1 is set for biodomes in netherstorm. &2 is set for some clovers and shrubbery in northrend.
Both are only used in Expansion01 respective Northrend.
The instance information specifies the actual M2 model to use, its absolute position and orientation within the world. The orientation is defined by rotations (in degrees) about the 3 axes as such (this order of operations is for OpenGL, so the transformations "actually" happen in reverse):
- Rotate around the Y axis by B-90
- Rotate around the Z axis by -A
- Rotate around the X axis by C
If you can't get those working, try these (The other ones didn't work for me, these does. My coordinate system is equal to WoW's (X is depth and Z is up))
- Rotate around Z (up) by B + 180 (or minus if you will)
- Rotate around X (depth) by +C
- Rotate around Y by +A
-MaiN
Got it perfect with
- Rotate around X by C
- Rotate around Y by -A
- Rotate around Z by B+180
X being depth forwards, Y width leftwards and Z height upwards. -eLaps
MODF chunk
Offset Type Description 0x00 uint32 ID (index in the MWID list) 0x04 uint32 unique identifier for this instance 0x08 3 floats Position (X,Y,Z) 0x14 3 floats Orientation (A,B,C) 0x20 3 floats Upper Extents 0x2C 3 floats Lower Extents 0x38 uint16 Flags 0x3A uint16 Doodad set index 0x3C uint16 Name set? 0x3E uint16 Padding (old: 0x3C uint32 Name set); it reads only a WORD into the WMAPOBJDEF structure. I don't know about the rest. Oo
To the flags: known: &1 (does something with the extends, no idea what exactely: paste, is set eg for World\wmo\Azeroth\Buildings\GuardTower\GuardTower.wmo and World\wmo\Azeroth\Buildings\GuardTower\GuardTower_destroyed.wmo in DeathknightStart_43_27. Maybe used for Phasing?
it's also used in patch 3.3.0a in: world\wmo\dungeon\icecrownraid\icecrownraid_arthas_precipice.wmo and world\wmo\dungeon\icecrownraid\icecrownraid_arthas_precipice_phase0.wmo in IcecrownCitadel_32_30. These flags turns those WMOs invisible in-game.)
struct SMMapObjDef // 03-29-2005 By ObscuR { /*000h*/ UINT32 nameId; /*004h*/ UINT32 uniqueId; /*008h*/ float pos[3]; /*00Ch*/ /*010h*/ /*014h*/ float rot[3]; /*018h*/ /*01Ch*/ /*020h*/ float extents[6]; /*024h*/ /*028h*/ /*02Ch*/ /*030h*/ /*034h*/ /*038h*/ UINT16 flags; /*03Ah*/ UINT16 doodadSetIndex; /*03Ch*/ UINT16 nameSet; /*03Eh*/ UINT16 pad; };
The positioning and orientation is done the same way as in the MDDF chunk. There is no scaling. Two additional positions and two integers are also given. They might or might not be used for lighting...?
The unique identifier is important for WMOs, because multiple map tiles might want to draw the same WMO. This identifier is used to ensure that each specific instance can only be drawn once. (a unique identifier is required because the model name is not usable for this purpose, since it is possible to have more than one instance of the same WMO, like some bridges in Darkshore)
MH2O chunk
This is a new chunk in WotLK. The old MCLQ subchunk died for it (still, its parsed by the client for backwards compatibility). Its now holding all values for the adt in one chunk. The chunk itself is seperated in three parts. A header, the data-block and the other referenced data.
Note: all offsets here refer to a position in the ADT file relative to the end of the MH2O header. So if you find 'O2HM' at 0x1000 then you add 0x1008 (the extra eight bytes being for the two four byte parameters to the chunk header) to all of these offsets to get their absolute position in the file.
Structure
Header
The header is a list of 256 entries of this structure:
0x00 int32 ofsInformation An offset to the MH2O_Information struct(s) for this chunk. 0x04 int32 layerCount 0 if the chunk has no liquids. If > 1, the offsets will point to arrays. 0x08 int32 ofsRender An offset to the 64 bit render mask.
struct MH2O_Header { MH2O_Information *Information; // points to an array with layers entries. uint32 layerCount; uint64 *Render; // the blocks to render. 8*8 bits (=8 bytes). regardless of the information on width and height. }
Information
This block is made of 0x18 bytes and is at ofsInformation for all layers.
0x00 int16 LiquidType Points to LiquidType.dbc 0x02 int16 flags 0x04 float heightLevel1 The global liquid-height of this chunk. Which is always in there twice. Blizzard knows why. 0x08 float heightLevel2 (Actually these values are not always identical, I think they define the highest and lowest points in the heightmap) 0x0C byte xOffset The X offset of the liquid square (0-7) 0x0D byte yOffset The Y offset of the liquid square (0-7) 0x0E byte width The width of the liquid square (1-8) 0x0F byte height The height of the liquid square (1-8) 0x10 int32 ofsMask2 Offset to some data. 0x14 int32 ofsHeightmap Offset to MH2O_HeightmapData structure for this chunk.
struct MH2O_Information { uint16 LiquidType; uint16 flags; float heightLevel1; float heightLevel2; byte xOffset; byte yOffset; byte width; byte height; uint8* Mask2; // points to an array of bits with information about the mask. w*h bits (=h bytes). MH2O_HeightmapData *HeightmapData; // note that if flags & 1 != 1, this chunk is "ocean" and the information stored // at HeightmapData does not correspond to the MH2O_HeightmapData structure. // Use heightLevel1 (or 2) for height in this case. }
Data
This block holds w*h float values defining the actual water heights. After them, there are the transparency bytes. These bytes define, if the ocean is full blue or if the ground can be seen.
struct MH2O_HeightmapData { // if type & 1 != 1, this chunk is "ocean". in this case, do not use this structure. float heightMap[]; // w*h char transparency[]; // w*h }
Explanation: MH2O_Header.Render
Indepth and with example. Thanks Slartibartfast.
The full heightmap that covers a whole chunk would be created from 9x9 float values, effectively creating 8x8 quadratic pieces. But since WotLK and the introduction of the MH2O chunk there is no more need to define the full heightmap if only part of a chunk is actually covered with water (such as with a thin river). Instead, MH2O_Information.x, .y, .width and .height define the size and location of a "liquid rectangle" which can be smaller than a full chunk.
An example: let's say there's a river crossing a chunk like this ('x' is the river):
++++++++ ++++++++ xxxxxx++ ++xxxxxx ++++++++ ++++++++ ++++++++ ++++++++
This would lead to xOffset = 0, yOffset = 2, width = 8 and height = 2. The data at MH2o_HeightmapData.heightMap would then list 27 float values for the height map (a 9x3 height map which results in 8x2 quads, as shown in the picture above).
The data pointed to by MH2O_Header.Render would finally define which of the quads should be rendered. Its length is always eight bytes, and each byte is a bitmap which contains a 1 for each quad to be rendered and a 0 for each invisible quad. In our example this would be the following bytes: 0x00 (binary 00000000), 0x00 (binary 00000000), 0xFC (binary 11111100), 0x3F (binary 00111111), 0x00 (binary 00000000), 0x00 (binary 00000000), 0x00 (binary 00000000), 0x00 (binary 00000000).
Note that it is always possible to omit MH2O_Header.Render and/or MH2O_Information.HeightmapData to save some bytes in the ADT file! If MH2O_Render.Render is not given, the whole "liquid rectangle" is to be rendered. If MH2O_Information.HeightmapData is not given, then the height map consists only of values equal to heightLevel1 (I am not 100% sure of this one, but this approach seems to work fine for me).
--Namreeb 15:55, 26 October 2009 (HST) --Slartibartfast 00:41, 30 October 2008 (CEST)
MCNK chunk
- After the above mentioned chunks come 256 individual MCNK chunks, row by row, starting from top-left (northwest). The MCNK chunks have a large block of data that starts with a header, and then has sub-chunks of its own.
Each map chunk has 9x9 vertices, and in between them 8x8 additional vertices, several texture layers, normal vectors, a shadow map, etc.
The MCNK header is 128 bytes large.
struct SMChunk // --schlumpf_ 17:01, 10 August 2009 (CEST), based on: 03-29-2005 By ObscuR, 11-08-2008 by Tharo { /*0x000*/ UINT32 flags; /*0x004*/ UINT32 IndexX; /*0x008*/ UINT32 IndexY; /*0x00C*/ UINT32 nLayers; // maximum 4 /*0x010*/ UINT32 nDoodadRefs; /*0x014*/ MCVT* ofsHeight; /*0x018*/ MCNR* ofsNormal; /*0x01C*/ MCLY* ofsLayer; /*0x020*/ MCRF* ofsRefs; /*0x024*/ MCAL* ofsAlpha; /*0x028*/ UINT32 sizeAlpha; /*0x02C*/ MCSH* ofsShadow; // only with flags&0x1 /*0x030*/ UINT32 sizeShadow; /*0x034*/ UINT32 areaid; /*0x038*/ UINT32 nMapObjRefs; /*0x03C*/ UINT32 holes; /*0x040*/ UINT2[8][8] ReallyLowQualityTextureingMap; // the content is the layer being on top, I guess. /*0x044*/ /*0x048*/ /*0x04C*/ /*0x050*/ UINT32 predTex; // 03-29-2005 By ObscuR; TODO: Investigate /*0x054*/ UINT32 noEffectDoodad; // 03-29-2005 By ObscuR; TODO: Investigate /*0x058*/ MCSE* ofsSndEmitters; /*0x05C*/ UINT32 nSndEmitters; //will be set to 0 in the client if ofsSndEmitters doesn't point to MCSE! /*0x060*/ MCLQ* ofsLiquid; /*0x064*/ UINT32 sizeLiquid; // 8 when not used; only read if >8. /*0x068*/ Vec3f position; /*0x06C*/ /*0x070*/ /*0x074*/ MCCV* ofsMCCV; // only with flags&0x40, had UINT32 textureId; in ObscuR's structure. /*0x078*/ MCLV* ofsMCLV; // introduced in Cataclysm /*0x07C*/ UINT32 unused; // currently unused /*0x080*/ }; enum // 03-29-2005 By ObscuR { FLAG_MCSH, FLAG_IMPASS, FLAG_LQ_RIVER, FLAG_LQ_OCEAN, FLAG_LQ_MAGMA, FLAG_MCCV, };
The X' and Z' coordinates specify the top left corner of the map chunk, but they are in an alternate coordinate system! (yay for consistency!) This one has its 0,0 point in the middle of the world, and has the axes' directions reversed.
X = 32*533.3333 - X' Z = 32*533.3333 - Z'
The Y base coordinate acts as a 'zero point' for the height values in the upcoming height map - that is, all height values are added to this Y height.
The X, Y, Z coordinates are in wow's main coordinates system as explained above. The struck explanation converts the coordinate system to one used by the author but that is never used in wow.
About the holes in the terrain: This is a bitmapped field, the least significant 16 bits are used row-wise in the following arrangement with a 1 bit meaning that the map chunk has a hole in that part of its area:
0x1 0x2 0x4 0x8 0x10 0x20 0x40 0x80 0x100 0x200 0x400 0x800 0x1000 0x2000 0x4000 0x8000
With this I've been able to fix some holes, the most obvious one being the Gates of Ironforge but there are some glitches for some other holes like the Lakeshire Inn - maybe this would require the more detailed height maps or for me to use a less cheap way to omit the hole triangles. :)
flags:
Flag Meaning 0x1 MCSH chunk available 0x2 Impassable? 0x4 River 0x8 Ocean 0x10 Magma 0x20 Slime? 0x40 MCCV chunk available 0x8000 Unknown, but heavily used in TBC.
MCVT sub-chunk
These are the actual height values for the 9x9+8x8 vertices. 145 floats in the following order/arrangement:.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140 141 142 143 144 145
The inner 8 vertices are only rendered in WoW when its using the up-close LoD. Otherwise, it only renders the outer 9. Nonsense? If I only change one of these it looks like: [3].
Ok, after a further look into it, WoW uses Squares out of 4 of the Outer(called NoLoD)-Vertices with one of the Inner(called LoD)-Vertices in the Center:
1 2 10 18 19
So to render them in OpenGL you can use something like this:
gl.glBegin(GL.GL_TRIANGLE_STRIP); for(int x=0;x<8;x++){ for(int y=0;y<8;y++){ float nL1=mcvt.getValNoLOD(x, y); float nL2=mcvt.getValNoLOD(x, y+1); float nL3=mcvt.getValNoLOD(x+1, y); float nL4=mcvt.getValNoLOD(x+1, y+1); float L=mcvt.getValLOD(x, y); gl.glVertex3f( y, x, nL1); gl.glVertex3f( y+1, x, nL2); gl.glVertex3f(y+0.5f, x+0.5f, L); gl.glVertex3f( y, x, nL1); gl.glVertex3f( y, x+1,nL3); gl.glVertex3f(y+0.5f, x+0.5f,L); gl.glVertex3f( y, x+1, nL3); gl.glVertex3f( y+1, x+1, nL4); gl.glVertex3f(y+0.5f, x+0.5f,L); gl.glVertex3f( y+1, x,nL2); gl.glVertex3f( y+1, x+1, nL4); gl.glVertex3f(y+0.5f, x+0.5f, L); } } gl.glEnd();
Although it seems there is still a mistake :/ --Tigurius
Old ones:
To stripify try this one: ( stripsize is now : 16*18 + 7*2 + 8*2 )
void stripify(V *in, V *out) { for (int row=0; row<8; row++) { V *thisrow = &in[row*9*2]; V *nextrow = &in[row*9*2 + 9]; V *overrow = &in[(row+1)*9*2]; if (row>0) *out++ = thisrow[0];// jump end for (int col=0; col<8; col++) { *out++ = thisrow[col]; *out++ = nextrow[col]; } *out++ = thisrow[8]; *out++ = overrow[8]; *out++ = overrow[8];// jump start *out++ = thisrow[0];// jump end *out++ = thisrow[0]; for (int col=0; col<8; col++) { *out++ = overrow[col]; *out++ = nextrow[col]; } if (row<8) *out++ = overrow[8]; if (row<7) *out++ = overrow[8];// jump start } }
or try this one (made by tharo)
// to make it not TOO complicated u get data as 9*9 and 8*9 chain. // the 9th value is never used but calculation is more easy now ^^ private int stripify(Point3d[] in, Point3d[] out) { int outc=0; for (int row=0; row<8; row++) { int thisrow = row*9*2; int nextrow = row*9*2 + 9; int overrow = (row+1) *9*2; for(int col=0; col<8; col++) { out[outc++] = in[thisrow+col]; out[outc++] = in[nextrow+col]; } out[outc++] = in[thisrow+8]; for(int col=8; col>0; col--) { out[outc++] = in[overrow+col]; out[outc++] = in[nextrow+col-1]; } out[outc++] = in[overrow]; out[outc++] = in[thisrow]; out[outc++] = in[nexttow]; out[outc++] = in[overrow]; } for(int row=8; row>=0; row--) { out[outc++] = in[row*9*2]; } return outc; }
These points look like they might be better organized as a triangle fan instead of a strip. This is my untested guess:
float wowData[145]; int off = 9; float x, y; for (y = 0; y < 8; ++y, off += 9) { for (x = 0; x < 8; ++x, ++off) { glBegin(GL_TRIANGLE_FAN); glVertex3f(x, y, wowData[off]); glVertex3f(x - 0.5f, y - 0.5f, wowData[off - 9]); glVertex3f(x + 0.5f, y - 0.5f, wowData[off - 8]); glVertex3f(x + 0.5f, y + 0.5f, wowData[off + 9]); glVertex3f(x - 0.5f, y + 0.5f, wowData[off + 8]); glVertex3f(x - 0.5f, y - 0.5f, wowData[off - 9]); glEnd(); } }
--Kelmar
MCLV sub-chunk
It looks like an array of Colors, one per heightmap vertex (i.e. a sequence of 145 x unsigned int). Most of them are 0xFF000000, but some differ - e.g. 0xFF060102. In almost all cases RGB components are very low (0x00 - 0x10).
MCCV sub-chunk
This is a new feature with vertex shading. You can manipulate the color of the vertices by adding this layer of colors blended onto the terrain.
The chunk contains 145 color values, each one is designated to one of the heightmap vertices. Each single color is defined in 4 bytes:
Offset Type Description 0x0 byte red value 0x1 byte green value 0x2 byte blue value 0x3 byte alpha value
The color components are one byte each, with 0x7F being 1.0, 0x00 being 0.0 and 0xFF being 2.0 (I am not sure about the 2.0 factor, but the visible change when setting a component to 0xFF suggests this value).
When rendering, these color settings manipulate the final color of a pixel in a chunk by multiplication. Usually all the component values are set to 0x7F (= multiplication with 1.0), which means there's no modification at all. But if set to (for example) 0x00, a chunk can be made completely black, or if set to 0xFF, a chunk is rendered brighter than normal.
You can see the direct effects of this in this video (see 3:25 to 3:45) from Blizzcon 09. Additionally, here is a screenshot showing some of the effects possible: Link to Screenshot
--Slartibartfast, Tigurius, --schlumpf_
Alpha seems to be without effect. Neither 0xFF nor 0x00 perform any visible change to 0x7F for the alphachannel. I will make some more investigations about that. Also it seems that they use a newer technology in the video posted above. With 145 possible vertices that can be colored you get a very sharp edged while in the video from Blizzcon you have no edges.
--Cromon
MCNR sub-chunk
- Normal vectors for each vertex, encoded as 3 signed bytes per normal, in the same order as specified above.
The size field of this chunk is wrong! Actually, it isn't wrong, but there are a few junk (unknown?) bytes between the MCNR chunk's end and the next chunk. The size of the MCNR chunk should be adjusted to 0x1C0 bytes instead of 0x1B3 bytes.
Normals are stored in X,Z,Y order, with 127 being 1.0 and -127 being -1.0. The vectors are normalized.
--Log 09:23, 28 May 2006 (EEST) Maybe the extra data are "edge flag" bitmaps that are used only by the client for smoothing normals around adjustment triangles ? I didn't check it, just a hint.
MCLY sub-chunk
Complete and right as of 19-AUG-09 (3.0.9 or higher)
- Texture layer definitions for this map chunk. 16 bytes per layer, up to 4 layers.
Every texture layer other than the first will have an alpha map to specify blending amounts. The first layer is rendered with full opacity. To know which alphamap is used, there is an offset into the MCAL chunk. That one is relative to MCAL.
You can animate these by setting the flags. Only simple linear animations are possible. You can specify the direction in 45° steps and the speed.
The textureId is just the array index of the filename array in the MTEX chunk.
For getting the right feeling when walking, you should set the effectId which links to GroundEffectTexture.dbc. It defines the little detaildoodads as well as the footstep sounds and if footprints are visible. You only need to have the upper layer with an id and you can set the id to -1 (int16!) to have no detaildoodads and footsteps at all (?).
Introduced in Wrath of the Lich King, terrain can now reflect a skybox. This is used for icecubes made out of ADTs to reflect something. You need to have the MTFX chunk in, if you want that. Look at an skybox Blizzard made to see how you should do it.
struct SMLayer // 03-29-2005 By ObscuR, --schlumpf_ 19:47, 19 August 2009 (CEST) { /*000h*/ UINT32 textureId; /*004h*/ UINT32 flags; /*008h*/ UINT32 offsetInMCAL; /*00Ch*/ INT32 effectId; // (actually int16 and padding) /*010h*/ };
Flag | Description | |
---|---|---|
0x001 | Animation: Rotate 45° clockwise. | |
0x002 | Animation: Rotate 90° clockwise. | |
0x004 | Animation: Rotate 180° clockwise. | |
0x008 | Animation: Make this faster. | |
0x010 | Animation: Faster!! | |
0x020 | Animation: Faster!!!! | |
0x040 | Animation: Animate this texture as told in the other bits. | |
0x080 | This will make the texture way brighter. Used for lava to make it "glow". | |
0x100 | Use alpha map - set for every layer after the first | |
0x200 | Alpha map is compressed (see MCAL chunk description) | |
0x400 | This makes the layer behave like its a reflection of the skybox. See below |
- Explanation for flag 0x400:
First of all you can see the effects in this video: Video The texture that became the 0x400 flag was the following: Image . Have a look at the bright points that wander with the toon as it moves. This should imitate the stars from the sky (that you can find in the texture).
There are some important things you should be aware when using the flag 0x400:
- It doesnt matter for which layer you set the flag 0x400, it will always affect the groundlayer.
- The common skyboxtextures need to have the same ration from width and height as the one posted above. If this isnt the case, it looks like that: Video You see that it doesnt really fit the shape.
- All of the skyboxtextures blizzard has need to specify a special flag to be decompressed correctly. This is done using the MTFX-chunk. If the texture has a 1 in MTFX it will be interpreted correctly, else it will be green.
--Cromon
MCRF sub-chunk
- A list of with MCNK.nDoodadRefs + MCNK.nMapObjRefs indices into the file's MDDF and MODF chunks, saying which MCNK subchunk those particular doodads and objects are drawn within. This MCRF list contains duplicates for map doodads that overlap areas.
As both, WMOs and M2s are referenced here, they get doodad indices first, then WMOs. If you have a doodad and a WMO in the ADT as well as the MCNK, you will have a {0,0} in MCRF with nDoodadRefs and MCNK.nMapObjRefs being 1.
The client uses those MCRF-entries to calculate collision. Only objects which are referenced in the current chunk of the toon get checked against collision (this is only for MDX, WMO seem to have different collision). If a doodad entry from MDDF or MODF gets never referenced in a chunks MCRF it wont be drawn at all, WoW doesnt take the MDDF and MODF to draw the objects.
--Cromon
MCSH sub-chunk
- Shadow map for static shadows on the terrain. Can be left out with the chunk&1 flag not set.
Thanks to Sylvain, the shadow maps work as follows: the shadows are stored per bit, not byte as 0 or 1 (off or on) so we have 8 bytes (which equates to 64 values) X 64 bytes (64 values in this case) which ends up as a square 64x64 shadowmap with either white or black. Note that the shadow values come LSB first.
MCAL sub-chunk
- Alpha maps for additional texture layers.
There are 3 kinds of alpha map here. The number of alpha map can reference MCLY's flag with 0x100.
- Uncompress mode size: 4096
- Uncompress mode size: 2048
- Self compress mode size: reference MCLY's offsetInMCAL can known the size
In order to know whether you deal with an uncompressed alphamap of 2048 bytes (preWotLK) or 4096 bytes (postWotLK), you should check the flags of the layer in the MCLY chunk.
Uncompress mode1
For every layer, a 32x64 array of alpha values. Can be used as a secondary texture with the modulation op to control the blending of the texture layers. For video cards with 2 texture units, this requires one pass per layer. (For 4 or more texture units, maybe this could be done faster using register combiners? Pixel shaders maybe?)
The size field of this chunk might be wrong for map chunks with zero texture layers. There are a couple of these in some of the development maps.
Similar to the shadow maps, the 32x64 bytes is really a 64x64 alpha map. There are 2 alpha values per byte, first 4 bits and second 4 bits. -Eric
Funny how this happened to work for the wrong sizes, too, because the upper 4 bits became the most significant in the alpha map, and the lower 4 appeared as noise. I sure didn't notice the difference :) It's fixed now of course. - Z.
Self compress mode
Flow - 21-10-2008:
In Wotlk these chunks are compressed.
I think it's a self made compression. Really simple. The size field of this chunk doesn't matter anymore.
How the decompression works
- read a byte
- check for sign bit
- if set we are in fill mode else we are in copy mode
- take the 7 lesser bits of the first byte as a count indicator
- fill mode: read the next byte an fill it by count in resulting alpha map
- copy mode: read the next count bytes and copy them in the resulting alpha map
- if the alpha map is complete we are done otherwise start at 1. again
Sample C++ code
// 21-10-2008 by Flow unsigned offI = 0; //offset IN buffer unsigned offO = 0; //offset OUT buffer char* buffIn; // pointer to data in adt file char buffOut[4096]; // the resulting alpha map while( offO < 4096 ) { // fill or copy mode bool fill = buffIn[offI] & 0x80; unsigned n = buffIn[offI] & 0x7F; offI++; for( unsigned k = 0; k < n; k++ ) { if (offO == 4096) break; buffOut[offO] = buffIn[offI]; offO++; if( !fill ) offI++; } if( fill ) offI++; }
Tested with Wotlk Beta build 8770
Uncompress mode2
Note: not all alpha maps are compressed; some of them are just stored as 4096 plain bytes. If an alpha map is compressed can be determined by checking the compression flag in the corresponding MCLY chunk entry. - Slartibartfast
Note: In order to know which compression you need to use you will need the WDT and the header of the Wdt.
if(layer[i].flags & 0x200)
--> Selfcompress mode
else if(wdtHeader.flags & 0x04)
--> 4096 bytes of plain data
else
--> 2048 compressed bytes like compress mode 1
Blend
Slartibartfast - 1 November 2008:
Blizzard has changed the way how the additional textures are blended onto the ground texture in Northrend (old continents still seem to be blended the old way; they also don't use the new alpha map format). They have gone from a "one-layer-per-step" approach to blending all the 4 textures in a single step according to the following formula:
finalColor = tex0 * (1.0 - (alpha1 + alpha2 + alpha3)) + tex1 * alpha1 + tex2 * alpha2 + tex3 * alpha3
So all the alpha values for the different layers including the ground layer add up to 1.0; the ground layer's alpha value is calculated to match this constraint.
How to render
It is of course possible to devise different ways to render such terrain; one way I use and of which I know that it's working is a 2-pass-approach: first render all ground textures without blending, then use a fragment shader program to mix the 1-3 additional layer textures and render them with a glBlendFunc setting of (GL_ONE, GL_ONE_MINUS_SRC_ALPHA) on top of the ground texture already present in the framebuffer. The fragment program that mixes the textures would have to work like this short GLSL example:
gl_FragColor = texture2D(texture0, vec2(gl_TexCoord[0])) * texture2D(texture3, vec2(gl_TexCoord[3])).r + texture2D(texture1, vec2(gl_TexCoord[1])) * texture2D(texture3, vec2(gl_TexCoord[3])).g + texture2D(texture2, vec2(gl_TexCoord[2])) * texture2D(texture3, vec2(gl_TexCoord[3])).b;
(this example uses 4 texture units: texture0 - texture3; the first 3 of them contain the actual textures, while the fourth unit contains the alpha maps combined in one RGB texture)
MCLQ sub-chunk
- Water levels for this map chunk. This chunk is old and not really used anymore. Still, there is backwards compatibility in the client as old ADTs are not updated as it would be much data to patch it. I guess, it will be done in some expansion. You can fully use this chunk, even to have multiple water. You can have a lot of stacked water with this and the MH2O one. I advise you to implement the MH2O one as its better if you want to write a editor for ADT files.
The size of the chunk is in the mapchunk header. The type of liquid is given in the mapchunk flags, also in the header.
This information is old and incomplete as well as maybe wrong.
The first two floats specify the minimum and maximum liquid height level. After them comes a 9x9 height map for the water with the following format per vertex:
Offset Type Description 0x00 int16 ? 0x02 int16 ? 0x04 float height value
The unknown int values might be color or transparency info, or something entirely different... Most frequently they are 0.
Followed by 8x8 bytes of flags for every liquid "tile" between the 9x9 vertex grid. The value 0x0F means do not render. (the specific flag for this seems to be 8 but I'm not sure - but it fixes some places where there was extra "water" sticking into the rest of the scenery)
Finally, 0x54 bytes of additional data, no idea what it's used for.
MCSE sub-chunk
- Sound emitters.
This seems to be a bit different to that structure, ObscuR posted back then. From what I can see, WoW takes only 0x1C bytes per entry. Quite a big difference. This change might have happened, when they introduced the SoundEntriesAdvanced.dbc.
struct CWSoundEmitter // --schlumpf_ 21:44, 8 August 2009 (CEST) { /*000h*/ uint32 SoundEntriesAdvancedId; /*004h*/ Vec3F position; /*008h*/ /*00Ch*/ /*010h*/ Vec3F size; // I'm not really sure with this. I'm far too lazy to analyze this. Seems like noone ever needed these anyway. /*014h*/ /*018h*/ };
old one by ObscuR (for older versions)
struct CWSoundEmitter // 04-29-2005 By ObscuR (Maybe not accurate but should look like this :p ) { /*000h*/ UINT32 soundPointID; /*004h*/ UINT32 soundNameID; /*008h*/ float pos[3]; /*00Ch*/ /*010h*/ /*014h*/ float minDistance; /*018h*/ float maxDistance; /*01Ch*/ float cutoffDistance; /*020h*/ UINT16 startTime; /*022h*/ UINT16 endTime; /*024h*/ UINT16 groupSilenceMin; /*026h*/ UINT16 groupSilenceMax; /*028h*/ UINT16 playInstancesMin; /*02Ah*/ UINT16 playInstancesMax; /*02Ch*/ UINT16 loopCountMin; /*02Eh*/ UINT16 loopCountMax; /*030h*/ UINT16 interSoundGapMin; /*032h*/ UINT16 interSoundGapMax; /*034h*/ };
MFBO chunk
- A bounding box for flying.
This chunk is a "box" defining, where you can fly and where you can't. It also defines the height at which one you will fall into nowhere while your camera remains at the same position. Its actually two planes with 3*3 coordinates per plane.
Therefore the structure is:
struct plane{ short[3][3] height; }; struct MFBO { plane maximum; plane minimum; };
MTFX chunk
This chunk is an array of integers that are 1 or 0. 1 means that the texture at the same position in the MTEX array has to be handled differentely. The size of this chunk is always the same as there are entries in the MTEX chunk.
Simple as it is:
struct MTFX { uint32 mode[nMTEX]; }
The textures with this extended rendering mode are no normal ones, but skyboxes. These skyboxes are getting added as a reflection layer on the terrain. This is used for icecubes reflecting clouds etc. The layer being the reflection one needs to have the 0x400 flag in the MCLY chunk.
Textures that are listed here need to be parsed otherwise. This specifies some specifications about the BLP. If you use one of the Terrain Cube Maps in a layer it will be displayed green as the client cannot decompress it in the normal way. If you set an entry in MTFX the texture will be loaded.
An example ("TILESET\\Terrain Cube Maps\\TCB_CrystalSong_A.blp"):
Without MTFX set to 1: http://www.imagr.eu/up/4b794f10abbc1_WoWScrnShot_021510_144015.jpg
With MTFX set 1: http://www.imagr.eu/up/4b794f3894877_WoWScrnShot_021510_143827.jpg
--Cromon
Partitially retrieved from "http://wowdev.org/wiki/index.php/ADT"