mirror of
https://git.code.sf.net/p/quake/quakeforge
synced 2024-11-11 07:42:18 +00:00
99c58cca73
Not sure if it actually works as the clients don't render the result properly (can't see anything where the model should be), but the output model does import back into blender properly.
368 lines
13 KiB
Python
368 lines
13 KiB
Python
# vim:ts=4:et
|
|
# ##### BEGIN GPL LICENSE BLOCK #####
|
|
#
|
|
# 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., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
|
|
#
|
|
# ##### END GPL LICENSE BLOCK #####
|
|
|
|
# <pep8 compliant>
|
|
|
|
import bpy
|
|
from bpy_extras.object_utils import object_data_add
|
|
from mathutils import Vector,Matrix
|
|
|
|
from .quakepal import palette
|
|
from .mdl import MDL
|
|
from .qfplist import pldata
|
|
|
|
def make_verts(mdl, framenum, subframenum=0):
|
|
frame = mdl.frames[framenum]
|
|
if frame.type:
|
|
frame = frame.frames[subframenum]
|
|
verts = []
|
|
s = Vector(mdl.scale)
|
|
o = Vector(mdl.scale_origin)
|
|
m = Matrix(((s.x, 0, 0,o.x),
|
|
( 0,s.y, 0,o.y),
|
|
( 0, 0,s.z,o.z),
|
|
( 0, 0, 0, 1)))
|
|
for v in frame.verts:
|
|
verts.append(m * Vector(v.r))
|
|
return verts
|
|
|
|
def make_faces(mdl):
|
|
faces = []
|
|
uvs = []
|
|
for tri in mdl.tris:
|
|
tv = list(tri.verts)
|
|
sts = []
|
|
for v in tri.verts:
|
|
stv = mdl.stverts[v]
|
|
s = stv.s
|
|
t = stv.t
|
|
if stv.onseam and not tri.facesfront:
|
|
s += mdl.skinwidth / 2
|
|
# quake textures are top to bottom, but blender images
|
|
# are bottom to top
|
|
sts.append((s * 1.0 / mdl.skinwidth, 1 - t * 1.0 / mdl.skinheight))
|
|
# blender's and quake's vertex order seem to be opposed
|
|
tv.reverse()
|
|
sts.reverse()
|
|
# annoyingly, blender can't have 0 in the final vertex, so rotate the
|
|
# face vertices and uvs
|
|
if not tv[2]:
|
|
tv = [tv[2]] + tv[:2]
|
|
sts = [sts[2]] + sts[:2]
|
|
faces.append (tv)
|
|
uvs.append(sts)
|
|
return faces, uvs
|
|
|
|
def load_skins(mdl):
|
|
def load_skin(skin, name):
|
|
skin.name = name
|
|
img = bpy.data.images.new(name, mdl.skinwidth, mdl.skinheight)
|
|
mdl.images.append(img)
|
|
p = [0.0] * mdl.skinwidth * mdl.skinheight * 4
|
|
d = skin.pixels
|
|
for j in range(mdl.skinheight):
|
|
for k in range(mdl.skinwidth):
|
|
c = palette[d[j * mdl.skinwidth + k]]
|
|
# quake textures are top to bottom, but blender images
|
|
# are bottom to top
|
|
l = ((mdl.skinheight - 1 - j) * mdl.skinwidth + k) * 4
|
|
p[l + 0] = c[0] / 255.0
|
|
p[l + 1] = c[1] / 255.0
|
|
p[l + 2] = c[2] / 255.0
|
|
p[l + 3] = 1.0
|
|
img.pixels[:] = p[:]
|
|
img.pack(True)
|
|
img.use_fake_user = True
|
|
|
|
mdl.images=[]
|
|
for i, skin in enumerate(mdl.skins):
|
|
if skin.type:
|
|
for j, subskin in enumerate(skin.skins):
|
|
load_skin (subskin, "%s_%d_%d" % (mdl.name, i, j))
|
|
else:
|
|
load_skin (skin, "%s_%d" % (mdl.name, i))
|
|
|
|
def setup_skins (mdl, uvs):
|
|
load_skins (mdl)
|
|
img = mdl.images[0] # use the first skin for now
|
|
uvlay = mdl.mesh.uv_textures.new(mdl.name)
|
|
uvloop = mdl.mesh.uv_loop_layers[0]
|
|
for i, texpoly in enumerate(uvlay.data):
|
|
poly = mdl.mesh.polygons[i]
|
|
mdl_uv = uvs[i]
|
|
texpoly.image = img
|
|
for j,k in enumerate(poly.loop_indices):
|
|
uvloop.data[k].uv = mdl_uv[j]
|
|
mat = bpy.data.materials.new(mdl.name)
|
|
mat.diffuse_color = (1,1,1)
|
|
mat.use_raytrace = False
|
|
tex = bpy.data.textures.new(mdl.name, 'IMAGE')
|
|
tex.extension = 'CLIP'
|
|
tex.use_preview_alpha = True
|
|
tex.image = img
|
|
mat.texture_slots.add()
|
|
ts = mat.texture_slots[0]
|
|
ts.texture = tex
|
|
ts.use_map_alpha = True
|
|
ts.texture_coords = 'UV'
|
|
mdl.mesh.materials.append(mat)
|
|
|
|
def make_shape_key(mdl, framenum, subframenum=0):
|
|
frame = mdl.frames[framenum]
|
|
name = "%s_%d" % (mdl.name, framenum)
|
|
if frame.type:
|
|
frame = frame.frames[subframenum]
|
|
name = "%s_%d_%d" % (mdl.name, framenum, subframenum)
|
|
if frame.name:
|
|
name = frame.name
|
|
else:
|
|
frame.name = name
|
|
frame.key = mdl.obj.shape_key_add(name)
|
|
frame.key.value = 0.0
|
|
mdl.keys.append (frame.key)
|
|
s = Vector(mdl.scale)
|
|
o = Vector(mdl.scale_origin)
|
|
m = Matrix(((s.x, 0, 0,o.x),
|
|
( 0,s.y, 0,o.y),
|
|
( 0, 0,s.z,o.z),
|
|
( 0, 0, 0, 1)))
|
|
for i, v in enumerate(frame.verts):
|
|
frame.key.data[i].co = m * Vector(v.r)
|
|
|
|
def build_shape_keys(mdl):
|
|
mdl.keys = []
|
|
mdl.obj.shape_key_add("Basis")
|
|
mdl.mesh.shape_keys.name = mdl.name
|
|
mdl.obj.active_shape_key_index = 0
|
|
for i, frame in enumerate(mdl.frames):
|
|
frame = mdl.frames[i]
|
|
if frame.type:
|
|
for j in range(len(frame.frames)):
|
|
make_shape_key(mdl, i, j)
|
|
else:
|
|
make_shape_key(mdl, i)
|
|
|
|
def set_keys(act, data):
|
|
for d in data:
|
|
key, co = d
|
|
dp = """key_blocks["%s"].value""" % key.name
|
|
fc = act.fcurves.new(data_path = dp)
|
|
fc.keyframe_points.add(len(co))
|
|
for i in range(len(co)):
|
|
fc.keyframe_points[i].co = co[i]
|
|
fc.keyframe_points[i].interpolation = 'LINEAR'
|
|
|
|
def build_actions(mdl):
|
|
sk = mdl.mesh.shape_keys
|
|
ad = sk.animation_data_create()
|
|
track = ad.nla_tracks.new ();
|
|
track.name = mdl.name
|
|
start_frame = 1.0
|
|
for frame in mdl.frames:
|
|
act = bpy.data.actions.new(frame.name)
|
|
data = []
|
|
other_keys = mdl.keys[:]
|
|
if frame.type:
|
|
for j, subframe in enumerate(frame.frames):
|
|
subframe.frameno = start_frame + j
|
|
co = []
|
|
if j > 1:
|
|
co.append ((1.0, 0.0))
|
|
if j > 0:
|
|
co.append ((j * 1.0, 0.0))
|
|
co.append (((j + 1) * 1.0, 1.0))
|
|
if j < len(frame.frames) - 2:
|
|
co.append (((j + 2) * 1.0, 0.0))
|
|
if j < len(frame.frames) - 1:
|
|
co.append ((len(frame.frames) * 1.0, 0.0))
|
|
data.append((subframe.key, co))
|
|
if subframe.key in other_keys:
|
|
del(other_keys[other_keys.index(subframe.key)])
|
|
co = [(1.0, 0.0), (len(frame.frames) * 1.0, 0.0)]
|
|
for k in other_keys:
|
|
data.append((k, co))
|
|
else:
|
|
sub.frameno = start_frame + j
|
|
data.append((frame.key, [(1.0, 1.0)]))
|
|
if frame.key in other_keys:
|
|
del(other_keys[other_keys.index(frame.key)])
|
|
co = [(1.0, 0.0)]
|
|
for k in other_keys:
|
|
data.append((k, co))
|
|
set_keys (act, data)
|
|
track.strips.new (act.name, start_frame, act)
|
|
start_frame += act.frame_range[1]
|
|
|
|
def merge_frames(mdl):
|
|
def get_base(name):
|
|
i = 0
|
|
while i < len(name) and name[i] not in "0123456789":
|
|
i += 1
|
|
return name[:i]
|
|
|
|
i = 0
|
|
while i < len(mdl.frames):
|
|
if mdl.frames[i].type:
|
|
i += 1
|
|
continue
|
|
base = get_base(mdl.frames[i].name)
|
|
j = i + 1
|
|
while j < len(mdl.frames):
|
|
if mdl.frames[j].type:
|
|
break
|
|
if get_base(mdl.frames[j].name) != base:
|
|
break
|
|
j += 1
|
|
f = MDL.Frame()
|
|
f.name = base
|
|
f.type = 1
|
|
f.frames = mdl.frames[i:j]
|
|
mdl.frames[i:j] = [f]
|
|
i += 1
|
|
|
|
def write_text(mdl):
|
|
header="""
|
|
/* This script represents the animation data within the model file. It
|
|
is generated automatically on import, and is optional when exporting.
|
|
If no script is used when exporting, only a single frame and single
|
|
skin will be exported.
|
|
|
|
The fundamental format of the script is documented at
|
|
http://quakeforge.net/doxygen/property-list.html
|
|
|
|
The expected layout is a top-level dictionary with two expected
|
|
entries:
|
|
frames array of frame entries
|
|
skins array of skin entries
|
|
|
|
A frame entry is a dictionary with the following fields:
|
|
name The name of the frame to be written to the mdl file. In a
|
|
frame group, this will form the base for sub-frame names
|
|
(name + relative frame number: eg, frame1) if the
|
|
sub-frame does not have a name field. (string)
|
|
frameno The blender frame to use for the captured animation. In a
|
|
frame group, this will be used as the base frame for any
|
|
sub-frames that do not specify a frame. While fractional
|
|
frames are supported, YMMV. (string:float)
|
|
frames Array of frame entries. If present, the current frame
|
|
entry is a frame group, and the frame entries specify
|
|
sub-frames. (array of dictionary)
|
|
NOTE: only top-level frames may be frame groups
|
|
intervals Array of frame end times for frame groups. No meaning
|
|
in blender, but the quake engine uses them for client-side
|
|
animations. Times must be ascending, but any step > 0 is
|
|
valid. Ignored for single frames. If not present in a
|
|
frame group, the sub-frames of the group will be written
|
|
as single frames (in order to undo the auto-group feature
|
|
of the importer). Excess times will be ignored, missing
|
|
times will be generated at 0.1
|
|
second intervals.
|
|
(array of string:float).
|
|
|
|
A skin entry is a dictionary with the following fields:
|
|
name The name of the blender image to be used as the skin.
|
|
Ignored for skin groups (animated skins). (string)
|
|
skins Array of skin entries. If present, the current skin
|
|
entry is a skin group (animated skin), and the skin
|
|
entries specify sub-skin. (array of dictionary)
|
|
NOTE: only top-level skins may be skins groups
|
|
intervals Array of skin end times for skin groups. No meaning
|
|
in blender, but the quake engine uses them for client-side
|
|
animations. Times must be ascending, but any step > 0 is
|
|
valid. Ignored for single skins. If not present in a
|
|
skin group, it will be generated using 0.1 second
|
|
intervals. Excess times will be ignored, missing times
|
|
will be generated at 0.1 second intervals.
|
|
(array of string:float).
|
|
*/
|
|
"""
|
|
d={'frames':[], 'skins':[]}
|
|
for f in mdl.frames:
|
|
d['frames'].append(f.info())
|
|
for s in mdl.skins:
|
|
d['skins'].append(s.info())
|
|
pl = pldata()
|
|
string = header + pl.write(d)
|
|
|
|
txt = bpy.data.texts.new(mdl.name)
|
|
txt.from_string(string)
|
|
mdl.text = txt
|
|
|
|
def parse_flags(flags):
|
|
#NOTE these are in QuakeForge priority order; a little different to id.
|
|
# id has rocket and grenate between tracer2 and tracer3
|
|
if flags & MDL.EF_ROCKET:
|
|
return 'EF_ROCKET'
|
|
elif flags & MDL.EF_GRENADE:
|
|
return 'EF_GRENADE'
|
|
elif flags & MDL.EF_GIB:
|
|
return 'EF_GIB'
|
|
elif flags & MDL.EF_ZOMGIB:
|
|
return 'EF_ZOMGIB'
|
|
elif flags & MDL.EF_TRACER:
|
|
return 'EF_TRACER'
|
|
elif flags & MDL.EF_TRACER2:
|
|
return 'EF_TRACER2'
|
|
elif flags & MDL.EF_TRACER3:
|
|
return 'EF_TRACER3'
|
|
else:
|
|
return 'EF_NONE'
|
|
|
|
def set_properties(mdl):
|
|
mdl.obj.qfmdl.eyeposition = mdl.eyeposition
|
|
try:
|
|
mdl.obj.qfmdl.synctype = MDL.SYNCTYPE[mdl.synctype]
|
|
except IndexError:
|
|
mdl.obj.qfmdl.synctype = 'ST_SYNC'
|
|
mdl.obj.qfmdl.rotate = (mdl.flags & MDL.EF_ROTATE) and True or False
|
|
mdl.obj.qfmdl.effects = parse_flags(mdl.flags)
|
|
mdl.obj.qfmdl.script = mdl.text.name #FIXME really want the text object
|
|
mdl.obj.qfmdl.md16 = (mdl.ident == "MD16")
|
|
|
|
def import_mdl(operator, context, filepath):
|
|
bpy.context.user_preferences.edit.use_global_undo = False
|
|
|
|
for obj in bpy.context.scene.objects:
|
|
obj.select = False
|
|
|
|
mdl = MDL()
|
|
if not mdl.read(filepath):
|
|
operator.report({'ERROR'},
|
|
"Unrecognized format: %s %d" % (mdl.ident, mdl.version))
|
|
return {'CANCELLED'}
|
|
faces, uvs = make_faces (mdl)
|
|
verts = make_verts (mdl, 0)
|
|
mdl.mesh = bpy.data.meshes.new(mdl.name)
|
|
mdl.mesh.from_pydata(verts, [], faces)
|
|
mdl.obj = bpy.data.objects.new(mdl.name, mdl.mesh)
|
|
bpy.context.scene.objects.link(mdl.obj)
|
|
bpy.context.scene.objects.active = mdl.obj
|
|
mdl.obj.select = True
|
|
setup_skins (mdl, uvs)
|
|
if len(mdl.frames) > 1 or mdl.frames[0].type:
|
|
build_shape_keys(mdl)
|
|
merge_frames(mdl)
|
|
build_actions(mdl)
|
|
write_text(mdl)
|
|
set_properties(mdl)
|
|
|
|
mdl.mesh.update()
|
|
|
|
bpy.context.user_preferences.edit.use_global_undo = True
|
|
return {'FINISHED'}
|