Files
io_scene_psk_psa/io_scene_psk_psa/psa/builder.py
2022-08-06 23:52:18 -07:00

284 lines
12 KiB
Python

from typing import Dict
from bpy.types import Action
from .data import *
from ..helpers import *
class PsaBuildOptions(object):
def __init__(self):
self.should_override_animation_data = False
self.animation_data_override = None
self.fps_source = 'SCENE'
self.fps_custom = 30.0
self.sequence_source = 'ACTIONS'
self.actions = []
self.marker_names = []
self.bone_filter_mode = 'ALL'
self.bone_group_indices = []
self.should_use_original_sequence_names = False
self.should_trim_timeline_marker_sequences = True
self.sequence_name_prefix = ''
self.sequence_name_suffix = ''
self.root_motion = False
def get_sequence_fps(context, options: PsaBuildOptions, actions: Iterable[Action]) -> float:
if options.fps_source == 'SCENE':
return context.scene.render.fps
if options.fps_source == 'CUSTOM':
return options.fps_custom
elif options.fps_source == 'ACTION_METADATA':
# Get the minimum value of action metadata FPS values.
fps_list = []
for action in filter(lambda x: 'psa_sequence_fps' in x, actions):
fps = action['psa_sequence_fps']
if type(fps) == int or type(fps) == float:
fps_list.append(fps)
if len(fps_list) > 0:
return min(fps_list)
else:
# No valid action metadata to use, fallback to scene FPS
return context.scene.render.fps
else:
raise RuntimeError(f'Invalid FPS source "{options.fps_source}"')
def get_timeline_marker_sequence_frame_ranges(animation_data, context, options: PsaBuildOptions) -> Dict:
# Timeline markers need to be sorted so that we can determine the sequence start and end positions.
sequence_frame_ranges = dict()
sorted_timeline_markers = list(sorted(context.scene.timeline_markers, key=lambda x: x.frame))
sorted_timeline_marker_names = list(map(lambda x: x.name, sorted_timeline_markers))
for marker_name in options.marker_names:
marker = context.scene.timeline_markers[marker_name]
frame_min = marker.frame
# Determine the final frame of the sequence based on the next marker.
# If no subsequent marker exists, use the maximum frame_end from all NLA strips.
marker_index = sorted_timeline_marker_names.index(marker_name)
next_marker_index = marker_index + 1
frame_max = 0
if next_marker_index < len(sorted_timeline_markers):
# There is a next marker. Use that next marker's frame position as the last frame of this sequence.
frame_max = sorted_timeline_markers[next_marker_index].frame
if options.should_trim_timeline_marker_sequences:
nla_strips = get_nla_strips_in_timeframe(animation_data, marker.frame, frame_max)
if len(nla_strips) > 0:
frame_max = min(frame_max, max(map(lambda nla_strip: nla_strip.frame_end, nla_strips)))
frame_min = max(frame_min, min(map(lambda nla_strip: nla_strip.frame_start, nla_strips)))
else:
# No strips in between this marker and the next, just export this as a one-frame animation.
frame_max = frame_min
else:
# There is no next marker.
# Find the final frame of all the NLA strips and use that as the last frame of this sequence.
for nla_track in animation_data.nla_tracks:
if nla_track.mute:
continue
for strip in nla_track.strips:
frame_max = max(frame_max, strip.frame_end)
if frame_min > frame_max:
continue
sequence_frame_ranges[marker_name] = int(frame_min), int(frame_max)
return sequence_frame_ranges
def build_psa(context, options: PsaBuildOptions) -> Psa:
active_object = context.view_layer.objects.active
if active_object.type != 'ARMATURE':
raise RuntimeError('Selected object must be an Armature')
if options.should_override_animation_data:
animation_data_object = options.animation_data_override
else:
animation_data_object = active_object
animation_data = animation_data_object.animation_data
if animation_data is None:
raise RuntimeError(f'No animation data for object \'{animation_data_object.name}\'')
# Ensure that we actually have items that we are going to be exporting.
if options.sequence_source == 'ACTIONS' and len(options.actions) == 0:
raise RuntimeError('No actions were selected for export')
elif options.sequence_source == 'TIMELINE_MARKERS' and len(options.marker_names) == 0:
raise RuntimeError('No timeline markers were selected for export')
psa = Psa()
armature = active_object
bones = list(armature.data.bones)
# The order of the armature bones and the pose bones is not guaranteed to be the same.
# As a result, we need to reconstruct the list of pose bones in the same order as the
# armature bones.
bone_names = [x.name for x in bones]
pose_bones = [(bone_names.index(bone.name), bone) for bone in armature.pose.bones]
pose_bones.sort(key=lambda x: x[0])
pose_bones = [x[1] for x in pose_bones]
# Get a list of all the bone indices and instigator bones for the bone filter settings.
export_bone_names = get_export_bone_names(armature, options.bone_filter_mode, options.bone_group_indices)
bone_indices = [bone_names.index(x) for x in export_bone_names]
# Make the bone lists contain only the bones that are going to be exported.
bones = [bones[bone_index] for bone_index in bone_indices]
pose_bones = [pose_bones[bone_index] for bone_index in bone_indices]
# No bones are going to be exported.
if len(bones) == 0:
raise RuntimeError('No bones available for export')
# Check that all bone names are valid.
check_bone_names(map(lambda bone: bone.name, bones))
# Build list of PSA bones.
for bone in bones:
psa_bone = Psa.Bone()
psa_bone.name = bytes(bone.name, encoding='windows-1252')
try:
parent_index = bones.index(bone.parent)
psa_bone.parent_index = parent_index
psa.bones[parent_index].children_count += 1
except ValueError:
psa_bone.parent_index = -1
if bone.parent is not None:
rotation = bone.matrix.to_quaternion().conjugated()
inverse_parent_rotation = bone.parent.matrix.to_quaternion().inverted()
parent_head = inverse_parent_rotation @ bone.parent.head
parent_tail = inverse_parent_rotation @ bone.parent.tail
location = (parent_tail - parent_head) + bone.head
else:
location = armature.matrix_local @ bone.head
rot_matrix = bone.matrix @ armature.matrix_local.to_3x3()
rotation = rot_matrix.to_quaternion()
psa_bone.location.x = location.x
psa_bone.location.y = location.y
psa_bone.location.z = location.z
psa_bone.rotation.x = rotation.x
psa_bone.rotation.y = rotation.y
psa_bone.rotation.z = rotation.z
psa_bone.rotation.w = rotation.w
psa.bones.append(psa_bone)
# Populate the export sequence list.
class NlaState:
def __init__(self):
self.frame_min = 0
self.frame_max = 0
self.action = None
class ExportSequence:
def __init__(self):
self.name = ''
self.nla_state = NlaState()
self.fps = 30.0
export_sequences = []
if options.sequence_source == 'ACTIONS':
for action in options.actions:
if len(action.fcurves) == 0:
continue
export_sequence = ExportSequence()
export_sequence.nla_state.action = action
export_sequence.name = get_psa_sequence_name(action, options.should_use_original_sequence_names)
frame_min, frame_max = [int(x) for x in action.frame_range]
export_sequence.nla_state.frame_min = frame_min
export_sequence.nla_state.frame_max = frame_max
export_sequence.fps = get_sequence_fps(context, options, [action])
export_sequences.append(export_sequence)
pass
elif options.sequence_source == 'TIMELINE_MARKERS':
sequence_frame_ranges = get_timeline_marker_sequence_frame_ranges(animation_data, context, options)
for name, (frame_min, frame_max) in sequence_frame_ranges.items():
export_sequence = ExportSequence()
export_sequence.name = name
export_sequence.nla_state.action = None
export_sequence.nla_state.frame_min = frame_min
export_sequence.nla_state.frame_max = frame_max
nla_strips_actions = set(
map(lambda x: x.action, get_nla_strips_in_timeframe(animation_data, frame_min, frame_max)))
export_sequence.fps = get_sequence_fps(context, options, nla_strips_actions)
export_sequences.append(export_sequence)
else:
raise ValueError(f'Unhandled sequence source: {options.sequence_source}')
# Add prefixes and suffices to the names of the export sequences and strip whitespace.
for export_sequence in export_sequences:
export_sequence.name = f'{options.sequence_name_prefix}{export_sequence.name}{options.sequence_name_suffix}'
export_sequence.name = export_sequence.name.strip()
# Now build the PSA sequences.
# We actually alter the timeline frame and simply record the resultant pose bone matrices.
frame_start_index = 0
for export_sequence in export_sequences:
# Link the action to the animation data and update view layer.
animation_data.action = export_sequence.nla_state.action
context.view_layer.update()
frame_min = export_sequence.nla_state.frame_min
frame_max = export_sequence.nla_state.frame_max
frame_count = frame_max - frame_min + 1
psa_sequence = Psa.Sequence()
psa_sequence.name = bytes(export_sequence.name, encoding='windows-1252')
psa_sequence.frame_count = frame_count
psa_sequence.frame_start_index = frame_start_index
psa_sequence.fps = export_sequence.fps
for frame in range(frame_count):
context.scene.frame_set(frame_min + frame)
for pose_bone in pose_bones:
key = Psa.Key()
if pose_bone.parent is not None:
pose_bone_matrix = pose_bone.matrix
pose_bone_parent_matrix = pose_bone.parent.matrix
pose_bone_matrix = pose_bone_parent_matrix.inverted() @ pose_bone_matrix
else:
if options.root_motion:
# Export root motion
pose_bone_matrix = armature.matrix_world @ pose_bone.matrix
else:
pose_bone_matrix = pose_bone.matrix
location = pose_bone_matrix.to_translation()
rotation = pose_bone_matrix.to_quaternion().normalized()
if pose_bone.parent is not None:
rotation.conjugate()
key.location.x = location.x
key.location.y = location.y
key.location.z = location.z
key.rotation.x = rotation.x
key.rotation.y = rotation.y
key.rotation.z = rotation.z
key.rotation.w = rotation.w
key.time = 1.0 / psa_sequence.fps
psa.keys.append(key)
psa_sequence.bone_count = len(pose_bones)
psa_sequence.track_time = frame_count
frame_start_index += frame_count
psa.sequences[export_sequence.name] = psa_sequence
return psa