forked from bartvdbraak/blender
982 lines
32 KiB
Python
982 lines
32 KiB
Python
#!BPY
|
|
"""
|
|
Name: '3D Studio (.3ds)...'
|
|
Blender: 244
|
|
Group: 'Import'
|
|
Tooltip: 'Import from 3DS file format (.3ds)'
|
|
"""
|
|
|
|
__author__= ['Bob Holcomb', 'Richard L?rk?ng', 'Damien McGinnes', 'Campbell Barton']
|
|
__url__ = ("blenderartists.org", "www.blender.org", "www.gametutorials.com", "lib3ds.sourceforge.net/")
|
|
__version__= '0.995'
|
|
__bpydoc__= '''\
|
|
|
|
3ds Importer
|
|
|
|
This script imports a 3ds file and the materials into Blender for editing.
|
|
|
|
Loader is based on 3ds loader from www.gametutorials.com (Thanks DigiBen).
|
|
|
|
|
|
0.995 by Campbell Barton<br>
|
|
- workaround for buggy mesh vert delete
|
|
- minor tweaks
|
|
|
|
0.99 by Bob Holcomb<br>
|
|
- added support for floating point color values that previously broke on import.
|
|
|
|
0.98 by Campbell Barton<br>
|
|
- import faces and verts to lists instead of a mesh, convert to a mesh later
|
|
- use new index mapping feature of mesh to re-map faces that were not added.
|
|
|
|
0.97 by Campbell Barton<br>
|
|
- Strip material names of spaces
|
|
- Added import as instance to import the 3ds into its own
|
|
scene and add a group instance to the current scene
|
|
- New option to scale down imported objects so they are within a limited bounding area.
|
|
|
|
0.96 by Campbell Barton<br>
|
|
- Added workaround for bug in setting UV's for Zero vert index UV faces.
|
|
- Removed unique name function, let blender make the names unique.
|
|
|
|
0.95 by Campbell Barton<br>
|
|
- Removed workarounds for Blender 2.41
|
|
- Mesh objects split by material- many 3ds objects used more then 16 per mesh.
|
|
- Removed a lot of unneeded variable creation.
|
|
|
|
0.94 by Campbell Barton<br>
|
|
- Face import tested to be about overall 16x speedup over 0.93.
|
|
- Material importing speedup.
|
|
- Tested with more models.
|
|
- Support some corrupt models.
|
|
|
|
0.93 by Campbell Barton<br>
|
|
- Tested with 400 3ds files from turbosquid and samples.
|
|
- Tactfully ignore faces that used the same verts twice.
|
|
- Rollback to 0.83 sloppy un-reorganized code, this broke UV coord loading.
|
|
- Converted from NMesh to Mesh.
|
|
- Faster and cleaner new names.
|
|
- Use external comprehensive image loader.
|
|
- Re intergrated 0.92 and 0.9 changes
|
|
- Fixes for 2.41 compat.
|
|
- Non textured faces do not use a texture flag.
|
|
|
|
0.92<br>
|
|
- Added support for diffuse, alpha, spec, bump maps in a single material
|
|
|
|
0.9<br>
|
|
- Reorganized code into object/material block functions<br>
|
|
- Use of Matrix() to copy matrix data<br>
|
|
- added support for material transparency<br>
|
|
|
|
0.83 2005-08-07: Campell Barton
|
|
- Aggressive image finding and case insensitivy for posisx systems.
|
|
|
|
0.82a 2005-07-22
|
|
- image texture loading (both for face uv and renderer)
|
|
|
|
0.82 - image texture loading (for face uv)
|
|
|
|
0.81a (fork- not 0.9) Campbell Barton 2005-06-08
|
|
- Simplified import code
|
|
- Never overwrite data
|
|
- Faster list handling
|
|
- Leaves import selected
|
|
|
|
0.81 Damien McGinnes 2005-01-09
|
|
- handle missing images better
|
|
|
|
0.8 Damien McGinnes 2005-01-08
|
|
- copies sticky UV coords to face ones
|
|
- handles images better
|
|
- Recommend that you run 'RemoveDoubles' on each imported mesh after using this script
|
|
|
|
'''
|
|
|
|
# ***** BEGIN GPL LICENSE BLOCK *****
|
|
#
|
|
# Script copyright (C) Bob Holcomb
|
|
#
|
|
# This program 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.
|
|
#
|
|
# This program 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 this program; if not, write to the Free Software Foundation,
|
|
# Inc., 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
|
|
#
|
|
# ***** END GPL LICENCE BLOCK *****
|
|
# --------------------------------------------------------------------------
|
|
|
|
# Importing modules
|
|
|
|
import Blender
|
|
import bpy
|
|
from Blender import Mesh, Object, Material, Image, Texture, Lamp, Mathutils
|
|
from Blender.Mathutils import Vector
|
|
import BPyImage
|
|
|
|
import BPyMessages
|
|
|
|
import struct
|
|
from struct import calcsize, unpack
|
|
|
|
import os
|
|
|
|
# If python version is less than 2.4, try to get set stuff from module
|
|
try:
|
|
set
|
|
except:
|
|
from sets import Set as set
|
|
|
|
BOUNDS_3DS= []
|
|
|
|
|
|
#this script imports uvcoords as sticky vertex coords
|
|
#this parameter enables copying these to face uv coords
|
|
#which shold be more useful.
|
|
|
|
def createBlenderTexture(material, name, image):
|
|
texture= bpy.data.textures.new(name)
|
|
texture.setType('Image')
|
|
texture.image= image
|
|
material.setTexture(0, texture, Texture.TexCo.UV, Texture.MapTo.COL)
|
|
|
|
|
|
|
|
######################################################
|
|
# Data Structures
|
|
######################################################
|
|
|
|
#Some of the chunks that we will see
|
|
#----- Primary Chunk, at the beginning of each file
|
|
PRIMARY= long('0x4D4D',16)
|
|
|
|
#------ Main Chunks
|
|
OBJECTINFO = long('0x3D3D',16); #This gives the version of the mesh and is found right before the material and object information
|
|
VERSION = long('0x0002',16); #This gives the version of the .3ds file
|
|
EDITKEYFRAME= long('0xB000',16); #This is the header for all of the key frame info
|
|
|
|
#------ sub defines of OBJECTINFO
|
|
MATERIAL=45055 #0xAFFF // This stored the texture info
|
|
OBJECT=16384 #0x4000 // This stores the faces, vertices, etc...
|
|
|
|
#>------ sub defines of MATERIAL
|
|
#------ sub defines of MATERIAL_BLOCK
|
|
MAT_NAME = long('0xA000',16) # This holds the material name
|
|
MAT_AMBIENT = long('0xA010',16) # Ambient color of the object/material
|
|
MAT_DIFFUSE = long('0xA020',16) # This holds the color of the object/material
|
|
MAT_SPECULAR = long('0xA030',16) # SPecular color of the object/material
|
|
MAT_SHINESS = long('0xA040',16) # ??
|
|
MAT_TRANSPARENCY= long('0xA050',16) # Transparency value of material
|
|
MAT_SELF_ILLUM = long('0xA080',16) # Self Illumination value of material
|
|
MAT_WIRE = long('0xA085',16) # Only render's wireframe
|
|
|
|
MAT_TEXTURE_MAP = long('0xA200',16) # This is a header for a new texture map
|
|
MAT_SPECULAR_MAP= long('0xA204',16) # This is a header for a new specular map
|
|
MAT_OPACITY_MAP = long('0xA210',16) # This is a header for a new opacity map
|
|
MAT_REFLECTION_MAP= long('0xA220',16) # This is a header for a new reflection map
|
|
MAT_BUMP_MAP = long('0xA230',16) # This is a header for a new bump map
|
|
MAT_MAP_FILENAME = long('0xA300',16) # This holds the file name of the texture
|
|
|
|
MAT_FLOAT_COLOR = long ('0x0010', 16) #color defined as 3 floats
|
|
MAT_24BIT_COLOR = long ('0x0011', 16) #color defined as 3 bytes
|
|
|
|
#>------ sub defines of OBJECT
|
|
OBJECT_MESH = long('0x4100',16); # This lets us know that we are reading a new object
|
|
OBJECT_LAMP = long('0x4600',16); # This lets un know we are reading a light object
|
|
OBJECT_LAMP_SPOT = long('0x4610',16); # The light is a spotloght.
|
|
OBJECT_LAMP_OFF = long('0x4620',16); # The light off.
|
|
OBJECT_LAMP_ATTENUATE = long('0x4625',16);
|
|
OBJECT_LAMP_RAYSHADE = long('0x4627',16);
|
|
OBJECT_LAMP_SHADOWED = long('0x4630',16);
|
|
OBJECT_LAMP_LOCAL_SHADOW = long('0x4640',16);
|
|
OBJECT_LAMP_LOCAL_SHADOW2 = long('0x4641',16);
|
|
OBJECT_LAMP_SEE_CONE = long('0x4650',16);
|
|
OBJECT_LAMP_SPOT_RECTANGULAR= long('0x4651',16);
|
|
OBJECT_LAMP_SPOT_OVERSHOOT= long('0x4652',16);
|
|
OBJECT_LAMP_SPOT_PROJECTOR= long('0x4653',16);
|
|
OBJECT_LAMP_EXCLUDE= long('0x4654',16);
|
|
OBJECT_LAMP_RANGE= long('0x4655',16);
|
|
OBJECT_LAMP_ROLL= long('0x4656',16);
|
|
OBJECT_LAMP_SPOT_ASPECT= long('0x4657',16);
|
|
OBJECT_LAMP_RAY_BIAS= long('0x4658',16);
|
|
OBJECT_LAMP_INNER_RANGE= long('0x4659',16);
|
|
OBJECT_LAMP_OUTER_RANGE= long('0x465A',16);
|
|
OBJECT_LAMP_MULTIPLIER = long('0x465B',16);
|
|
OBJECT_LAMP_AMBIENT_LIGHT = long('0x4680',16);
|
|
|
|
|
|
|
|
OBJECT_CAMERA= long('0x4700',16); # This lets un know we are reading a camera object
|
|
|
|
#>------ sub defines of CAMERA
|
|
OBJECT_CAM_RANGES= long('0x4720',16); # The camera range values
|
|
|
|
#>------ sub defines of OBJECT_MESH
|
|
OBJECT_VERTICES = long('0x4110',16); # The objects vertices
|
|
OBJECT_FACES = long('0x4120',16); # The objects faces
|
|
OBJECT_MATERIAL = long('0x4130',16); # This is found if the object has a material, either texture map or color
|
|
OBJECT_UV = long('0x4140',16); # The UV texture coordinates
|
|
OBJECT_TRANS_MATRIX = long('0x4160',16); # The Object Matrix
|
|
|
|
global scn
|
|
scn= None
|
|
|
|
#the chunk class
|
|
class chunk:
|
|
ID=0
|
|
length=0
|
|
bytes_read=0
|
|
|
|
#we don't read in the bytes_read, we compute that
|
|
binary_format='<HI'
|
|
|
|
def __init__(self):
|
|
self.ID=0
|
|
self.length=0
|
|
self.bytes_read=0
|
|
|
|
def dump(self):
|
|
print 'ID: ', self.ID
|
|
print 'ID in hex: ', hex(self.ID)
|
|
print 'length: ', self.length
|
|
print 'bytes_read: ', self.bytes_read
|
|
|
|
def read_chunk(file, chunk):
|
|
temp_data=file.read(calcsize(chunk.binary_format))
|
|
data=unpack(chunk.binary_format, temp_data)
|
|
chunk.ID=data[0]
|
|
chunk.length=data[1]
|
|
#update the bytes read function
|
|
chunk.bytes_read=6
|
|
|
|
#if debugging
|
|
#chunk.dump()
|
|
|
|
def read_string(file):
|
|
#read in the characters till we get a null character
|
|
s=''
|
|
while not s.endswith('\x00'):
|
|
s+=unpack( '<c', file.read(1) )[0]
|
|
#print 'string: ',s
|
|
|
|
#remove the null character from the string
|
|
return s[:-1]
|
|
|
|
######################################################
|
|
# IMPORT
|
|
######################################################
|
|
def process_next_object_chunk(file, previous_chunk):
|
|
new_chunk=chunk()
|
|
temp_chunk=chunk()
|
|
|
|
while (previous_chunk.bytes_read<previous_chunk.length):
|
|
#read the next chunk
|
|
read_chunk(file, new_chunk)
|
|
|
|
def skip_to_end(file, skip_chunk):
|
|
buffer_size=skip_chunk.length-skip_chunk.bytes_read
|
|
binary_format='%ic' % buffer_size
|
|
temp_data=file.read(calcsize(binary_format))
|
|
skip_chunk.bytes_read+=buffer_size
|
|
|
|
|
|
def add_texture_to_material(image, texture, material, mapto):
|
|
if mapto=='DIFFUSE':
|
|
map=Texture.MapTo.COL
|
|
elif mapto=='SPECULAR':
|
|
map=Texture.MapTo.SPEC
|
|
elif mapto=='OPACITY':
|
|
map=Texture.MapTo.ALPHA
|
|
elif mapto=='BUMP':
|
|
map=Texture.MapTo.NOR
|
|
else:
|
|
print '/tError: Cannot map to "%s"\n\tassuming diffuse color. modify material "%s" later.' % (mapto, material.name)
|
|
map=Texture.MapTo.COL
|
|
|
|
if image: texture.setImage(image) # double check its an image.
|
|
free_tex_slots= [i for i, tex in enumerate( material.getTextures() ) if tex==None]
|
|
if not free_tex_slots:
|
|
print '/tError: Cannot add "%s" map. 10 Texture slots alredy used.' % mapto
|
|
else:
|
|
material.setTexture(free_tex_slots[0],texture,Texture.TexCo.UV,map)
|
|
|
|
|
|
def process_next_chunk(file, previous_chunk, importedObjects, IMAGE_SEARCH):
|
|
#print previous_chunk.bytes_read, 'BYTES READ'
|
|
contextObName= None
|
|
contextLamp= [None, None] # object, Data
|
|
contextMaterial= None
|
|
contextMatrix_rot= None # Blender.Mathutils.Matrix(); contextMatrix.identity()
|
|
#contextMatrix_tx= None # Blender.Mathutils.Matrix(); contextMatrix.identity()
|
|
contextMesh_vertls= None
|
|
contextMesh_facels= None
|
|
contextMeshMaterials= {} # matname:[face_idxs]
|
|
|
|
TEXTURE_DICT={}
|
|
MATDICT={}
|
|
TEXMODE= Mesh.FaceModes['TEX']
|
|
|
|
# Localspace variable names, faster.
|
|
STRUCT_SIZE_1CHAR= calcsize('c')
|
|
STRUCT_SIZE_2FLOAT= calcsize('2f')
|
|
STRUCT_SIZE_3FLOAT= calcsize('3f')
|
|
STRUCT_SIZE_UNSIGNED_SHORT= calcsize('H')
|
|
STRUCT_SIZE_4UNSIGNED_SHORT= calcsize('4H')
|
|
STRUCT_SIZE_4x3MAT= calcsize('ffffffffffff')
|
|
_STRUCT_SIZE_4x3MAT= calcsize('fffffffffffff')
|
|
# STRUCT_SIZE_4x3MAT= calcsize('ffffffffffff')
|
|
# print STRUCT_SIZE_4x3MAT, ' STRUCT_SIZE_4x3MAT'
|
|
|
|
def putContextMesh(myContextMesh_vertls, myContextMesh_facels, myContextMeshMaterials):
|
|
|
|
materialFaces= set() # faces that have a material. Can optimize?
|
|
|
|
# Now make copies with assigned materils.
|
|
|
|
def makeMeshMaterialCopy(matName, faces):
|
|
'''
|
|
Make a new mesh with only face the faces that use this material.
|
|
faces can be any iterable object - containing ints.
|
|
'''
|
|
|
|
faceVertUsers = [False] * len(myContextMesh_vertls)
|
|
ok=0
|
|
for fIdx in faces:
|
|
for vindex in myContextMesh_facels[fIdx]:
|
|
faceVertUsers[vindex] = True
|
|
if matName != None: # if matName is none then this is a set(), meaning we are using the untextured faces and do not need to store textured faces.
|
|
materialFaces.add(fIdx)
|
|
ok=1
|
|
|
|
if not ok:
|
|
return
|
|
|
|
myVertMapping = {}
|
|
vertMappingIndex = 0
|
|
|
|
vertsToUse = [i for i in xrange(len(myContextMesh_vertls)) if faceVertUsers[i]]
|
|
myVertMapping = dict( [ (ii, i) for i, ii in enumerate(vertsToUse) ] )
|
|
|
|
tempName= '%s_%s' % (contextObName, matName) # matName may be None.
|
|
bmesh = bpy.data.meshes.new(tempName)
|
|
|
|
if matName == None:
|
|
img= None
|
|
else:
|
|
bmat = MATDICT[matName][1]
|
|
bmesh.materials= [bmat]
|
|
try: img= TEXTURE_DICT[bmat.name]
|
|
except: img= None
|
|
|
|
bmesh_verts = bmesh.verts
|
|
bmesh_verts.extend( [Vector()] )
|
|
bmesh_verts.extend( [myContextMesh_vertls[i] for i in vertsToUse] )
|
|
# +1 because of DUMMYVERT
|
|
face_mapping= bmesh.faces.extend( [ [ bmesh_verts[ myVertMapping[vindex]+1] for vindex in myContextMesh_facels[fIdx]] for fIdx in faces ], indexList=True )
|
|
|
|
if bmesh.faces and (contextMeshUV or img):
|
|
bmesh.faceUV= 1
|
|
for ii, i in enumerate(faces):
|
|
|
|
# Mapped index- faces may have not been added- if so, then map to the correct index
|
|
# BUGGY API - face_mapping is not always the right length
|
|
map_index= face_mapping[ii]
|
|
|
|
if map_index != None:
|
|
targetFace= bmesh.faces[map_index]
|
|
if contextMeshUV:
|
|
# v.index-1 because of the DUMMYVERT
|
|
targetFace.uv= [contextMeshUV[vindex] for vindex in myContextMesh_facels[i]]
|
|
if img:
|
|
targetFace.image= img
|
|
|
|
# bmesh.transform(contextMatrix)
|
|
ob = SCN_OBJECTS.new(bmesh, tempName)
|
|
'''
|
|
if contextMatrix_tx:
|
|
ob.setMatrix(contextMatrix_tx)
|
|
'''
|
|
|
|
if contextMatrix_rot:
|
|
ob.setMatrix(contextMatrix_rot)
|
|
|
|
importedObjects.append(ob)
|
|
|
|
|
|
for matName, faces in myContextMeshMaterials.iteritems():
|
|
makeMeshMaterialCopy(matName, faces)
|
|
|
|
if len(materialFaces)!=len(myContextMesh_facels):
|
|
# Invert material faces.
|
|
makeMeshMaterialCopy(None, set(range(len( myContextMesh_facels ))) - materialFaces)
|
|
#raise 'Some UnMaterialed faces', len(contextMesh.faces)
|
|
|
|
#a spare chunk
|
|
new_chunk= chunk()
|
|
temp_chunk= chunk()
|
|
|
|
#loop through all the data for this chunk (previous chunk) and see what it is
|
|
while (previous_chunk.bytes_read<previous_chunk.length):
|
|
#print '\t', previous_chunk.bytes_read, 'keep going'
|
|
#read the next chunk
|
|
#print 'reading a chunk'
|
|
read_chunk(file, new_chunk)
|
|
|
|
#is it a Version chunk?
|
|
if (new_chunk.ID==VERSION):
|
|
#print 'if (new_chunk.ID==VERSION):'
|
|
#print 'found a VERSION chunk'
|
|
#read in the version of the file
|
|
#it's an unsigned short (H)
|
|
temp_data= file.read(calcsize('I'))
|
|
version = unpack('<I', temp_data)[0]
|
|
new_chunk.bytes_read+= 4 #read the 4 bytes for the version number
|
|
#this loader works with version 3 and below, but may not with 4 and above
|
|
if (version>3):
|
|
print '\tNon-Fatal Error: Version greater than 3, may not load correctly: ', version
|
|
|
|
#is it an object info chunk?
|
|
elif (new_chunk.ID==OBJECTINFO):
|
|
#print 'elif (new_chunk.ID==OBJECTINFO):'
|
|
# print 'found an OBJECTINFO chunk'
|
|
process_next_chunk(file, new_chunk, importedObjects, IMAGE_SEARCH)
|
|
|
|
#keep track of how much we read in the main chunk
|
|
new_chunk.bytes_read+=temp_chunk.bytes_read
|
|
|
|
#is it an object chunk?
|
|
elif (new_chunk.ID==OBJECT):
|
|
tempName= read_string(file)
|
|
contextObName= tempName
|
|
new_chunk.bytes_read += len(tempName)+1
|
|
|
|
#is it a material chunk?
|
|
elif (new_chunk.ID==MATERIAL):
|
|
#print 'elif (new_chunk.ID==MATERIAL):'
|
|
contextMaterial= bpy.data.materials.new('Material')
|
|
|
|
elif (new_chunk.ID==MAT_NAME):
|
|
#print 'elif (new_chunk.ID==MAT_NAME):'
|
|
material_name= read_string(file)
|
|
|
|
#plus one for the null character that ended the string
|
|
new_chunk.bytes_read+= len(material_name)+1
|
|
|
|
contextMaterial.name= material_name.rstrip() # remove trailing whitespace
|
|
MATDICT[material_name]= (contextMaterial.name, contextMaterial)
|
|
|
|
elif (new_chunk.ID==MAT_AMBIENT):
|
|
#print 'elif (new_chunk.ID==MAT_AMBIENT):'
|
|
read_chunk(file, temp_chunk)
|
|
if (temp_chunk.ID==MAT_FLOAT_COLOR):
|
|
temp_data=file.read(calcsize('3f'))
|
|
temp_chunk.bytes_read+=12
|
|
contextMaterial.mirCol=[float(col) for col in unpack('<3f', temp_data)]
|
|
elif (temp_chunk.ID==MAT_24BIT_COLOR):
|
|
temp_data=file.read(calcsize('3B'))
|
|
temp_chunk.bytes_read+= 3
|
|
contextMaterial.mirCol= [float(col)/255 for col in unpack('<3B', temp_data)] # data [0,1,2] == rgb
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
|
|
elif (new_chunk.ID==MAT_DIFFUSE):
|
|
#print 'elif (new_chunk.ID==MAT_DIFFUSE):'
|
|
read_chunk(file, temp_chunk)
|
|
if (temp_chunk.ID==MAT_FLOAT_COLOR):
|
|
temp_data=file.read(calcsize('3f'))
|
|
temp_chunk.bytes_read+=12
|
|
contextMaterial.rgbCol=[float(col) for col in unpack('<3f', temp_data)]
|
|
elif (temp_chunk.ID==MAT_24BIT_COLOR):
|
|
temp_data=file.read(calcsize('3B'))
|
|
temp_chunk.bytes_read+= 3
|
|
contextMaterial.rgbCol= [float(col)/255 for col in unpack('<3B', temp_data)] # data [0,1,2] == rgb
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
|
|
elif (new_chunk.ID==MAT_SPECULAR):
|
|
#print 'elif (new_chunk.ID==MAT_SPECULAR):'
|
|
read_chunk(file, temp_chunk)
|
|
if (temp_chunk.ID==MAT_FLOAT_COLOR):
|
|
temp_data=file.read(calcsize('3f'))
|
|
temp_chunk.bytes_read+=12
|
|
contextMaterial.mirCol=[float(col) for col in unpack('<3f', temp_data)]
|
|
elif (temp_chunk.ID==MAT_24BIT_COLOR):
|
|
temp_data=file.read(calcsize('3B'))
|
|
temp_chunk.bytes_read+= 3
|
|
contextMaterial.mirCol= [float(col)/255 for col in unpack('<3B', temp_data)] # data [0,1,2] == rgb
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
|
|
elif (new_chunk.ID==MAT_TEXTURE_MAP):
|
|
#print 'elif (new_chunk.ID==MAT_TEXTURE_MAP):'
|
|
new_texture= bpy.data.textures.new('Diffuse')
|
|
new_texture.setType('Image')
|
|
img = None
|
|
while (new_chunk.bytes_read<new_chunk.length):
|
|
#print 'MAT_TEXTURE_MAP..while', new_chunk.bytes_read, new_chunk.length
|
|
read_chunk(file, temp_chunk)
|
|
|
|
if (temp_chunk.ID==MAT_MAP_FILENAME):
|
|
texture_name=read_string(file)
|
|
#img= TEXTURE_DICT[contextMaterial.name]= BPyImage.comprehensiveImageLoad(texture_name, FILENAME)
|
|
img= TEXTURE_DICT[contextMaterial.name]= BPyImage.comprehensiveImageLoad(texture_name, FILENAME, PLACE_HOLDER= False, RECURSIVE= IMAGE_SEARCH)
|
|
new_chunk.bytes_read += (len(texture_name)+1) #plus one for the null character that gets removed
|
|
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
|
|
#add the map to the material in the right channel
|
|
if img:
|
|
add_texture_to_material(img, new_texture, contextMaterial, 'DIFFUSE')
|
|
|
|
elif (new_chunk.ID==MAT_SPECULAR_MAP):
|
|
#print 'elif (new_chunk.ID==MAT_SPECULAR_MAP):'
|
|
new_texture= bpy.data.textures.new('Specular')
|
|
new_texture.setType('Image')
|
|
img = None
|
|
while (new_chunk.bytes_read<new_chunk.length):
|
|
read_chunk(file, temp_chunk)
|
|
|
|
if (temp_chunk.ID==MAT_MAP_FILENAME):
|
|
texture_name= read_string(file)
|
|
#img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME)
|
|
img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME, PLACE_HOLDER= False, RECURSIVE= IMAGE_SEARCH)
|
|
new_chunk.bytes_read+= (len(texture_name)+1) #plus one for the null character that gets removed
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
|
|
#add the map to the material in the right channel
|
|
if img:
|
|
add_texture_to_material(img, new_texture, contextMaterial, 'SPECULAR')
|
|
|
|
elif (new_chunk.ID==MAT_OPACITY_MAP):
|
|
#print 'new_texture=Blender.Texture.New('Opacity')'
|
|
new_texture= bpy.data.textures.new('Opacity')
|
|
new_texture.setType('Image')
|
|
img = None
|
|
while (new_chunk.bytes_read<new_chunk.length):
|
|
read_chunk(file, temp_chunk)
|
|
|
|
if (temp_chunk.ID==MAT_MAP_FILENAME):
|
|
texture_name= read_string(file)
|
|
#img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME)
|
|
img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME, PLACE_HOLDER= False, RECURSIVE= IMAGE_SEARCH)
|
|
new_chunk.bytes_read += (len(texture_name)+1) #plus one for the null character that gets removed
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
|
|
new_chunk.bytes_read+= temp_chunk.bytes_read
|
|
#add the map to the material in the right channel
|
|
if img:
|
|
add_texture_to_material(img, new_texture, contextMaterial, 'OPACITY')
|
|
|
|
elif (new_chunk.ID==MAT_BUMP_MAP):
|
|
#print 'elif (new_chunk.ID==MAT_BUMP_MAP):'
|
|
new_texture= bpy.data.textures.new('Bump')
|
|
new_texture.setType('Image')
|
|
img = None
|
|
while (new_chunk.bytes_read<new_chunk.length):
|
|
read_chunk(file, temp_chunk)
|
|
|
|
if (temp_chunk.ID==MAT_MAP_FILENAME):
|
|
texture_name= read_string(file)
|
|
#img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME)
|
|
img= BPyImage.comprehensiveImageLoad(texture_name, FILENAME, PLACE_HOLDER= False, RECURSIVE= IMAGE_SEARCH)
|
|
new_chunk.bytes_read += (len(texture_name)+1) #plus one for the null character that gets removed
|
|
else:
|
|
skip_to_end(file, temp_chunk)
|
|
|
|
new_chunk.bytes_read+=temp_chunk.bytes_read
|
|
|
|
#add the map to the material in the right channel
|
|
if img:
|
|
add_texture_to_material(img, new_texture, contextMaterial, 'BUMP')
|
|
|
|
elif (new_chunk.ID==MAT_TRANSPARENCY):
|
|
#print 'elif (new_chunk.ID==MAT_TRANSPARENCY):'
|
|
read_chunk(file, temp_chunk)
|
|
temp_data=file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
|
|
temp_chunk.bytes_read+=2
|
|
contextMaterial.alpha= 1-(float(unpack('<H', temp_data)[0])/100)
|
|
new_chunk.bytes_read+=temp_chunk.bytes_read
|
|
|
|
|
|
elif (new_chunk.ID==OBJECT_LAMP): # Basic lamp support.
|
|
|
|
temp_data=file.read(STRUCT_SIZE_3FLOAT)
|
|
|
|
x,y,z=unpack('<3f', temp_data)
|
|
new_chunk.bytes_read+=STRUCT_SIZE_3FLOAT
|
|
|
|
contextLamp[1]= bpy.data.lamps.new()
|
|
contextLamp[0]= SCN_OBJECTS.new(contextLamp[1])
|
|
importedObjects.append(contextLamp[0])
|
|
|
|
#print 'number of faces: ', num_faces
|
|
#print x,y,z
|
|
contextLamp[0].setLocation(x,y,z)
|
|
|
|
# Reset matrix
|
|
contextMatrix_rot= None
|
|
#contextMatrix_tx= None
|
|
#print contextLamp.name,
|
|
|
|
|
|
elif (new_chunk.ID==OBJECT_MESH):
|
|
# print 'Found an OBJECT_MESH chunk'
|
|
if contextMesh_facels != None: # Write context mesh if we have one.
|
|
putContextMesh(contextMesh_vertls, contextMesh_facels, contextMeshMaterials)
|
|
|
|
contextMesh_vertls= []; contextMesh_facels= []
|
|
|
|
contextMeshMaterials= {} # matname:[face_idxs]
|
|
contextMeshUV= None
|
|
#contextMesh.vertexUV= 1 # Make sticky coords.
|
|
# Reset matrix
|
|
contextMatrix_rot= None
|
|
#contextMatrix_tx= None
|
|
|
|
elif (new_chunk.ID==OBJECT_VERTICES):
|
|
'''
|
|
Worldspace vertex locations
|
|
'''
|
|
# print 'elif (new_chunk.ID==OBJECT_VERTICES):'
|
|
temp_data=file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
num_verts=unpack('<H', temp_data)[0]
|
|
new_chunk.bytes_read+=2
|
|
|
|
# print 'number of verts: ', num_verts
|
|
def getvert():
|
|
temp_data= unpack('<3f', file.read(STRUCT_SIZE_3FLOAT))
|
|
new_chunk.bytes_read += STRUCT_SIZE_3FLOAT #12: 3 floats x 4 bytes each
|
|
return temp_data
|
|
|
|
#contextMesh.verts.extend( [Vector(),] ) # DUMMYVERT! - remove when blenders internals are fixed.
|
|
contextMesh_vertls= [getvert() for i in xrange(num_verts)]
|
|
|
|
#print 'object verts: bytes read: ', new_chunk.bytes_read
|
|
|
|
elif (new_chunk.ID==OBJECT_FACES):
|
|
# print 'elif (new_chunk.ID==OBJECT_FACES):'
|
|
temp_data= file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
num_faces= unpack('<H', temp_data)[0]
|
|
new_chunk.bytes_read+= 2
|
|
#print 'number of faces: ', num_faces
|
|
|
|
def getface():
|
|
# print '\ngetting a face'
|
|
temp_data= file.read(STRUCT_SIZE_4UNSIGNED_SHORT)
|
|
new_chunk.bytes_read+= STRUCT_SIZE_4UNSIGNED_SHORT #4 short ints x 2 bytes each
|
|
v1,v2,v3,dummy= unpack('<4H', temp_data)
|
|
return v1, v2, v3
|
|
|
|
contextMesh_facels= [ getface() for i in xrange(num_faces) ]
|
|
|
|
|
|
elif (new_chunk.ID==OBJECT_MATERIAL):
|
|
# print 'elif (new_chunk.ID==OBJECT_MATERIAL):'
|
|
material_name= read_string(file)
|
|
new_chunk.bytes_read += len(material_name)+1 # remove 1 null character.
|
|
|
|
temp_data=file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
num_faces_using_mat = unpack('<H', temp_data)[0]
|
|
new_chunk.bytes_read += STRUCT_SIZE_UNSIGNED_SHORT
|
|
|
|
def getmat():
|
|
temp_data= file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
new_chunk.bytes_read+= STRUCT_SIZE_UNSIGNED_SHORT
|
|
return unpack('<H', temp_data)[0]
|
|
|
|
contextMeshMaterials[material_name]= [ getmat() for i in xrange(num_faces_using_mat) ]
|
|
|
|
#look up the material in all the materials
|
|
|
|
elif (new_chunk.ID==OBJECT_UV):
|
|
temp_data=file.read(STRUCT_SIZE_UNSIGNED_SHORT)
|
|
num_uv=unpack('<H', temp_data)[0]
|
|
new_chunk.bytes_read+= 2
|
|
|
|
def getuv():
|
|
temp_data=file.read(STRUCT_SIZE_2FLOAT)
|
|
new_chunk.bytes_read += STRUCT_SIZE_2FLOAT #2 float x 4 bytes each
|
|
return Vector( unpack('<2f', temp_data) )
|
|
|
|
contextMeshUV= [ getuv() for i in xrange(num_uv) ]
|
|
|
|
elif (new_chunk.ID== OBJECT_TRANS_MATRIX):
|
|
# How do we know the matrix size? 54 == 4x4 48 == 4x3
|
|
temp_data=file.read(STRUCT_SIZE_4x3MAT)
|
|
data= list( unpack('<ffffffffffff', temp_data) )
|
|
new_chunk.bytes_read += STRUCT_SIZE_4x3MAT
|
|
|
|
contextMatrix_rot= Blender.Mathutils.Matrix(\
|
|
data[:3] + [0],\
|
|
data[3:6] + [0],\
|
|
data[6:9] + [0],\
|
|
data[9:] + [1])
|
|
|
|
|
|
'''
|
|
contextMatrix_rot= Blender.Mathutils.Matrix(\
|
|
data[:3] + [0],\
|
|
data[3:6] + [0],\
|
|
data[6:9] + [0],\
|
|
[0,0,0,1])
|
|
'''
|
|
|
|
'''
|
|
contextMatrix_rot= Blender.Mathutils.Matrix(\
|
|
data[:3] ,\
|
|
data[3:6],\
|
|
data[6:9])
|
|
'''
|
|
|
|
'''
|
|
contextMatrix_rot = Blender.Mathutils.Matrix()
|
|
m = 0
|
|
for j in xrange(4):
|
|
for i in xrange(3):
|
|
contextMatrix_rot[j][i] = data[m]
|
|
m+=1
|
|
|
|
contextMatrix_rot[0][3]=0;
|
|
contextMatrix_rot[1][3]=0;
|
|
contextMatrix_rot[2][3]=0;
|
|
contextMatrix_rot[3][3]=1;
|
|
'''
|
|
|
|
#contextMatrix_rot.resize4x4()
|
|
#print "MTX"
|
|
#print contextMatrix_rot
|
|
contextMatrix_rot.invert()
|
|
#print contextMatrix_rot
|
|
#contextMatrix_tx = Blender.Mathutils.TranslationMatrix(0.5 * Blender.Mathutils.Vector(data[9:]))
|
|
#contextMatrix_tx.invert()
|
|
|
|
#tx.invert()
|
|
|
|
#contextMatrix = contextMatrix * tx
|
|
#contextMatrix = contextMatrix *tx
|
|
|
|
elif (new_chunk.ID==MAT_MAP_FILENAME):
|
|
texture_name=read_string(file)
|
|
try:
|
|
TEXTURE_DICT[contextMaterial.name]
|
|
except:
|
|
#img= TEXTURE_DICT[contextMaterial.name]= BPyImage.comprehensiveImageLoad(texture_name, FILENAME)
|
|
img= TEXTURE_DICT[contextMaterial.name]= BPyImage.comprehensiveImageLoad(texture_name, FILENAME, PLACE_HOLDER= False, RECURSIVE= IMAGE_SEARCH)
|
|
|
|
new_chunk.bytes_read+= len(texture_name)+1 #plus one for the null character that gets removed
|
|
|
|
else: #(new_chunk.ID!=VERSION or new_chunk.ID!=OBJECTINFO or new_chunk.ID!=OBJECT or new_chunk.ID!=MATERIAL):
|
|
# print 'skipping to end of this chunk'
|
|
buffer_size=new_chunk.length-new_chunk.bytes_read
|
|
binary_format='%ic' % buffer_size
|
|
temp_data=file.read(calcsize(binary_format))
|
|
new_chunk.bytes_read+=buffer_size
|
|
|
|
|
|
#update the previous chunk bytes read
|
|
# print 'previous_chunk.bytes_read += new_chunk.bytes_read'
|
|
# print previous_chunk.bytes_read, new_chunk.bytes_read
|
|
previous_chunk.bytes_read += new_chunk.bytes_read
|
|
## print 'Bytes left in this chunk: ', previous_chunk.length-previous_chunk.bytes_read
|
|
|
|
# FINISHED LOOP
|
|
# There will be a number of objects still not added
|
|
if contextMesh_facels != None:
|
|
putContextMesh(contextMesh_vertls, contextMesh_facels, contextMeshMaterials)
|
|
|
|
def load_3ds(filename, PREF_UI= True):
|
|
global FILENAME, SCN_OBJECTS
|
|
|
|
if BPyMessages.Error_NoFile(filename):
|
|
return
|
|
|
|
print '\n\nImporting 3DS: "%s"' % (Blender.sys.expandpath(filename))
|
|
|
|
time1= Blender.sys.time()
|
|
|
|
FILENAME=filename
|
|
current_chunk=chunk()
|
|
|
|
file=open(filename,'rb')
|
|
|
|
#here we go!
|
|
# print 'reading the first chunk'
|
|
read_chunk(file, current_chunk)
|
|
if (current_chunk.ID!=PRIMARY):
|
|
print '\tFatal Error: Not a valid 3ds file: ', filename
|
|
file.close()
|
|
return
|
|
|
|
|
|
# IMPORT_AS_INSTANCE= Blender.Draw.Create(0)
|
|
IMPORT_CONSTRAIN_BOUNDS= Blender.Draw.Create(10.0)
|
|
IMAGE_SEARCH= Blender.Draw.Create(1)
|
|
|
|
# Get USER Options
|
|
pup_block= [\
|
|
('Size Constraint:', IMPORT_CONSTRAIN_BOUNDS, 0.0, 1000.0, 'Scale the model by 10 until it reacehs the size constraint. Zero Disables.'),\
|
|
('Image Search', IMAGE_SEARCH, 'Search subdirs for any assosiated images (Warning, may be slow)'),\
|
|
#('Group Instance', IMPORT_AS_INSTANCE, 'Import objects into a new scene and group, creating an instance in the current scene.'),\
|
|
]
|
|
|
|
if PREF_UI:
|
|
if not Blender.Draw.PupBlock('Import 3DS...', pup_block):
|
|
return
|
|
|
|
Blender.Window.WaitCursor(1)
|
|
|
|
IMPORT_CONSTRAIN_BOUNDS= IMPORT_CONSTRAIN_BOUNDS.val
|
|
# IMPORT_AS_INSTANCE= IMPORT_AS_INSTANCE.val
|
|
IMAGE_SEARCH = IMAGE_SEARCH.val
|
|
|
|
if IMPORT_CONSTRAIN_BOUNDS:
|
|
BOUNDS_3DS[:]= [1<<30, 1<<30, 1<<30, -1<<30, -1<<30, -1<<30]
|
|
else:
|
|
BOUNDS_3DS[:]= []
|
|
|
|
##IMAGE_SEARCH
|
|
|
|
scn= bpy.data.scenes.active
|
|
SCN_OBJECTS = scn.objects
|
|
SCN_OBJECTS.selected = [] # de select all
|
|
|
|
importedObjects= [] # Fill this list with objects
|
|
process_next_chunk(file, current_chunk, importedObjects, IMAGE_SEARCH)
|
|
|
|
|
|
# Link the objects into this scene.
|
|
# Layers= scn.Layers
|
|
|
|
# REMOVE DUMMYVERT, - remove this in the next release when blenders internal are fixed.
|
|
|
|
|
|
for ob in importedObjects:
|
|
if ob.type=='Mesh':
|
|
me= ob.getData(mesh=1)
|
|
me.verts.delete([me.verts[0],])
|
|
|
|
# Done DUMMYVERT
|
|
"""
|
|
if IMPORT_AS_INSTANCE:
|
|
name= filename.split('\\')[-1].split('/')[-1]
|
|
# Create a group for this import.
|
|
group_scn= Scene.New(name)
|
|
for ob in importedObjects:
|
|
group_scn.link(ob) # dont worry about the layers
|
|
|
|
grp= Blender.Group.New(name)
|
|
grp.objects= importedObjects
|
|
|
|
grp_ob= Object.New('Empty', name)
|
|
grp_ob.enableDupGroup= True
|
|
grp_ob.DupGroup= grp
|
|
scn.link(grp_ob)
|
|
grp_ob.Layers= Layers
|
|
grp_ob.sel= 1
|
|
else:
|
|
# Select all imported objects.
|
|
for ob in importedObjects:
|
|
scn.link(ob)
|
|
ob.Layers= Layers
|
|
ob.sel= 1
|
|
"""
|
|
|
|
if IMPORT_CONSTRAIN_BOUNDS!=0.0:
|
|
# Set bounds from objecyt bounding box
|
|
for ob in importedObjects:
|
|
if ob.type=='Mesh':
|
|
ob.makeDisplayList() # Why dosnt this update the bounds?
|
|
for v in ob.getBoundBox():
|
|
for i in (0,1,2):
|
|
if v[i] < BOUNDS_3DS[i]:
|
|
BOUNDS_3DS[i]= v[i] # min
|
|
|
|
if v[i] > BOUNDS_3DS[i+3]:
|
|
BOUNDS_3DS[i+3]= v[i] # min
|
|
|
|
# Get the max axis x/y/z
|
|
max_axis= max(BOUNDS_3DS[3]-BOUNDS_3DS[0], BOUNDS_3DS[4]-BOUNDS_3DS[1], BOUNDS_3DS[5]-BOUNDS_3DS[2])
|
|
# print max_axis
|
|
if max_axis < 1<<30: # Should never be false but just make sure.
|
|
|
|
# Get a new scale factor if set as an option
|
|
SCALE=1.0
|
|
while (max_axis*SCALE) > IMPORT_CONSTRAIN_BOUNDS:
|
|
SCALE/=10
|
|
|
|
# SCALE Matrix
|
|
SCALE_MAT= Blender.Mathutils.Matrix([SCALE,0,0,0],[0,SCALE,0,0],[0,0,SCALE,0],[0,0,0,1])
|
|
|
|
for ob in importedObjects:
|
|
ob.setMatrix(ob.matrixWorld*SCALE_MAT)
|
|
|
|
# Done constraining to bounds.
|
|
|
|
# Select all new objects.
|
|
print 'finished importing: "%s" in %.4f sec.' % (filename, (Blender.sys.time()-time1))
|
|
file.close()
|
|
Blender.Window.WaitCursor(0)
|
|
|
|
|
|
DEBUG= False
|
|
if __name__=='__main__' and not DEBUG:
|
|
Blender.Window.FileSelector(load_3ds, 'Import 3DS', '*.3ds')
|
|
|
|
# For testing compatibility
|
|
#load_3ds('/metavr/convert/vehicle/truck_002/TruckTanker1.3DS', False)
|
|
#load_3ds('/metavr/archive/convert/old/arranged_3ds_to_hpx-2/only-need-engine-trains/Engine2.3DS', False)
|
|
'''
|
|
|
|
else:
|
|
# DEBUG ONLY
|
|
TIME= Blender.sys.time()
|
|
import os
|
|
print 'Searching for files'
|
|
os.system('find /metavr/ -iname "*.3ds" > /tmp/temp3ds_list')
|
|
# os.system('find /storage/ -iname "*.3ds" > /tmp/temp3ds_list')
|
|
print '...Done'
|
|
file= open('/tmp/temp3ds_list', 'r')
|
|
lines= file.readlines()
|
|
file.close()
|
|
# sort by filesize for faster testing
|
|
lines_size = [(os.path.getsize(f[:-1]), f[:-1]) for f in lines]
|
|
lines_size.sort()
|
|
lines = [f[1] for f in lines_size]
|
|
|
|
|
|
def between(v,a,b):
|
|
if v <= max(a,b) and v >= min(a,b):
|
|
return True
|
|
return False
|
|
|
|
for i, _3ds in enumerate(lines):
|
|
if between(i, 650,800):
|
|
#_3ds= _3ds[:-1]
|
|
print 'Importing', _3ds, '\nNUMBER', i, 'of', len(lines)
|
|
_3ds_file= _3ds.split('/')[-1].split('\\')[-1]
|
|
newScn= Blender.Scene.New(_3ds_file)
|
|
newScn.makeCurrent()
|
|
load_3ds(_3ds, False)
|
|
|
|
print 'TOTAL TIME: %.6f' % (Blender.sys.time() - TIME)
|
|
|
|
''' |