Cura/cura/Arranging/GridArrange.py
2023-10-10 11:15:57 +02:00

348 lines
17 KiB
Python

import math
from typing import List, TYPE_CHECKING, Tuple, Set, Union
if TYPE_CHECKING:
from UM.Scene.SceneNode import SceneNode
from cura.BuildVolume import BuildVolume
from UM.Application import Application
from UM.Math.AxisAlignedBox import AxisAlignedBox
from UM.Math.Polygon import Polygon
from UM.Math.Vector import Vector
from UM.Operations.AddSceneNodeOperation import AddSceneNodeOperation
from UM.Operations.GroupedOperation import GroupedOperation
from UM.Operations.TranslateOperation import TranslateOperation
from cura.Arranging.Arranger import Arranger
class GridArrange(Arranger):
def __init__(self, nodes_to_arrange: List["SceneNode"], build_volume: "BuildVolume", fixed_nodes: List["SceneNode"] = None):
if fixed_nodes is None:
fixed_nodes = []
self._nodes_to_arrange = nodes_to_arrange
self._build_volume = build_volume
self._build_volume_bounding_box = build_volume.getBoundingBox()
self._fixed_nodes = fixed_nodes
self._margin_x: float = 1
self._margin_y: float = 1
self._grid_width = 0
self._grid_height = 0
for node in self._nodes_to_arrange:
bounding_box = node.getBoundingBox()
self._grid_width = max(self._grid_width, bounding_box.width)
self._grid_height = max(self._grid_height, bounding_box.depth)
self._grid_width += self._margin_x
self._grid_height += self._margin_y
# Round up the grid size to the nearest cm, this assures that new objects will
# be placed on integer offsets from each other
grid_precision = 10 # 1cm
rounded_grid_width = math.ceil(self._grid_width / grid_precision) * grid_precision
rounded_grid_height = math.ceil(self._grid_height / grid_precision) * grid_precision
# The space added by the "grid precision rounding up" of the grid size
self._grid_round_margin_x = rounded_grid_width - self._grid_width
self._grid_round_margin_y = rounded_grid_height - self._grid_height
self._grid_width = rounded_grid_width
self._grid_height = rounded_grid_height
self._offset_x = 0
self._offset_y = 0
self._findOptimalGridOffset()
coord_initial_leftover_x = self._build_volume_bounding_box.right + 2 * self._grid_width
coord_initial_leftover_y = (self._build_volume_bounding_box.back + self._build_volume_bounding_box.front) * 0.5
self._initial_leftover_grid_x, self._initial_leftover_grid_y = self._coordSpaceToGridSpace(
coord_initial_leftover_x, coord_initial_leftover_y)
self._initial_leftover_grid_x = math.floor(self._initial_leftover_grid_x)
self._initial_leftover_grid_y = math.floor(self._initial_leftover_grid_y)
# Find grid indexes that intersect with fixed objects
self._fixed_nodes_grid_ids = set()
for node in self._fixed_nodes:
self._fixed_nodes_grid_ids = self._fixed_nodes_grid_ids.union(
self._intersectingGridIdxInclusive(node.getBoundingBox()))
# grid indexes that are in disallowed area
for polygon in self._build_volume.getDisallowedAreas():
self._fixed_nodes_grid_ids = self._fixed_nodes_grid_ids.union(self._intersectingGridIdxInclusive(polygon))
self._build_plate_grid_ids = self._intersectingGridIdxExclusive(self._build_volume_bounding_box)
# Filter out the corner grid squares if the build plate shape is elliptic
if self._build_volume.getShape() == "elliptic":
self._build_plate_grid_ids = set(
filter(lambda grid_id: self._checkGridUnderDiscSpace(grid_id[0], grid_id[1]),
self._build_plate_grid_ids))
self._allowed_grid_idx = self._build_plate_grid_ids.difference(self._fixed_nodes_grid_ids)
def createGroupOperationForArrange(self, add_new_nodes_in_scene: bool = False) -> Tuple[GroupedOperation, int]:
# Find the sequence in which items are placed
coord_build_plate_center_x = self._build_volume_bounding_box.width * 0.5 + self._build_volume_bounding_box.left
coord_build_plate_center_y = self._build_volume_bounding_box.depth * 0.5 + self._build_volume_bounding_box.back
grid_build_plate_center_x, grid_build_plate_center_y = self._coordSpaceToGridSpace(coord_build_plate_center_x,
coord_build_plate_center_y)
sequence: List[Tuple[int, int]] = list(self._allowed_grid_idx)
sequence.sort(key=lambda grid_id: (grid_build_plate_center_x - grid_id[0]) ** 2 + (
grid_build_plate_center_y - grid_id[1]) ** 2)
scene_root = Application.getInstance().getController().getScene().getRoot()
grouped_operation = GroupedOperation()
for grid_id, node in zip(sequence, self._nodes_to_arrange):
if add_new_nodes_in_scene:
grouped_operation.addOperation(AddSceneNodeOperation(node, scene_root))
grid_x, grid_y = grid_id
operation = self._moveNodeOnGrid(node, grid_x, grid_y)
grouped_operation.addOperation(operation)
leftover_nodes = self._nodes_to_arrange[len(sequence):]
left_over_grid_y = self._initial_leftover_grid_y
for node in leftover_nodes:
if add_new_nodes_in_scene:
grouped_operation.addOperation(AddSceneNodeOperation(node, scene_root))
# find the first next grid position that isn't occupied by a fixed node
while (self._initial_leftover_grid_x, left_over_grid_y) in self._fixed_nodes_grid_ids:
left_over_grid_y = left_over_grid_y - 1
operation = self._moveNodeOnGrid(node, self._initial_leftover_grid_x, left_over_grid_y)
grouped_operation.addOperation(operation)
left_over_grid_y = left_over_grid_y - 1
return grouped_operation, len(leftover_nodes)
def _findOptimalGridOffset(self):
if len(self._fixed_nodes) == 0:
self._offset_x = 0
self._offset_y = 0
return
if len(self._fixed_nodes) == 1:
center_grid_x = 0.5 * self._grid_width + self._build_volume_bounding_box.left
center_grid_y = 0.5 * self._grid_height + self._build_volume_bounding_box.back
bounding_box = self._fixed_nodes[0].getBoundingBox()
center_node_x = (bounding_box.left + bounding_box.right) * 0.5
center_node_y = (bounding_box.back + bounding_box.front) * 0.5
self._offset_x = center_node_x - center_grid_x
self._offset_y = center_node_y - center_grid_y
return
# If there are multiple fixed nodes, an optimal solution is not always possible
# We will try to find an offset that minimizes the number of grid intersections
# with fixed nodes. The algorithm below achieves this by utilizing a scanline
# algorithm. In this algorithm each axis is solved separately as offsetting
# is completely independent in each axis. The comments explaining the algorithm
# below are for the x-axis, but the same applies for the y-axis.
#
# Each node either occupies ceil((node.right - node.right) / grid_width) or
# ceil((node.right - node.right) / grid_width) + 1 grid squares. We will call
# these the node's "footprint".
#
# ┌────────────────┐
# minimum foot-print │ NODE │
# └────────────────┘
# │ grid 1 │ grid 2 │ grid 3 │ grid 4 | grid 5 |
# ┌────────────────┐
# maximum foot-print │ NODE │
# └────────────────┘
#
# The algorithm will find the grid offset such that the number of nodes with
# a _minimal_ footprint is _maximized_.
# The scanline algorithm works as follows, we create events for both end points
# of each node's footprint. The event have two properties,
# - the coordinate: the amount the endpoint can move to the
# left before it crosses a grid line
# - the change: either +1 or -1, indicating whether crossing the grid line
# would result in a minimal footprint node becoming a maximal footprint
class Event:
def __init__(self, coord: float, change: float):
self.coord = coord
self.change = change
# create events for both the horizontal and vertical axis
events_horizontal: List[Event] = []
events_vertical: List[Event] = []
for node in self._fixed_nodes:
bounding_box = node.getBoundingBox()
left = bounding_box.left - self._build_volume_bounding_box.left
right = bounding_box.right - self._build_volume_bounding_box.left
back = bounding_box.back - self._build_volume_bounding_box.back
front = bounding_box.front - self._build_volume_bounding_box.back
value_left = math.ceil(left / self._grid_width) * self._grid_width - left
value_right = math.ceil(right / self._grid_width) * self._grid_width - right
value_back = math.ceil(back / self._grid_height) * self._grid_height - back
value_front = math.ceil(front / self._grid_height) * self._grid_height - front
# give nodes a weight according to their size. This
# weight is heuristically chosen to be proportional to
# the number of grid squares the node-boundary occupies
weight = bounding_box.width + bounding_box.depth
events_horizontal.append(Event(value_left, weight))
events_horizontal.append(Event(value_right, -weight))
events_vertical.append(Event(value_back, weight))
events_vertical.append(Event(value_front, -weight))
events_horizontal.sort(key=lambda event: event.coord)
events_vertical.sort(key=lambda event: event.coord)
def findOptimalShiftAxis(events: List[Event], interval: float) -> float:
# executing the actual scanline algorithm
# iteratively go through events (left to right) and keep track of the
# current footprint. The optimal location is the one with the minimal
# footprint. If there are multiple locations with the same minimal
# footprint, the optimal location is the one with the largest range
# between the left and right endpoint of the footprint.
prev_offset = events[-1].coord - interval
current_minimal_footprint_count = 0
best_minimal_footprint_count = float('inf')
best_offset_span = float('-inf')
best_offset = 0.0
for event in events:
offset_span = event.coord - prev_offset
if current_minimal_footprint_count < best_minimal_footprint_count or (
current_minimal_footprint_count == best_minimal_footprint_count and offset_span > best_offset_span):
best_minimal_footprint_count = current_minimal_footprint_count
best_offset_span = offset_span
best_offset = event.coord
current_minimal_footprint_count += event.change
prev_offset = event.coord
return best_offset - best_offset_span * 0.5
center_grid_x = 0.5 * self._grid_width
center_grid_y = 0.5 * self._grid_height
optimal_center_x = self._grid_width - findOptimalShiftAxis(events_horizontal, self._grid_width)
optimal_center_y = self._grid_height - findOptimalShiftAxis(events_vertical, self._grid_height)
self._offset_x = optimal_center_x - center_grid_x
self._offset_y = optimal_center_y - center_grid_y
def _moveNodeOnGrid(self, node: "SceneNode", grid_x: int, grid_y: int) -> "Operation.Operation":
coord_grid_x, coord_grid_y = self._gridSpaceToCoordSpace(grid_x, grid_y)
center_grid_x = coord_grid_x + (0.5 * self._grid_width)
center_grid_y = coord_grid_y + (0.5 * self._grid_height)
bounding_box = node.getBoundingBox()
center_node_x = (bounding_box.left + bounding_box.right) * 0.5
center_node_y = (bounding_box.back + bounding_box.front) * 0.5
delta_x = center_grid_x - center_node_x
delta_y = center_grid_y - center_node_y
return TranslateOperation(node, Vector(delta_x, 0, delta_y))
def _getGridCornerPoints(
self,
bounds: Union[AxisAlignedBox, Polygon],
*,
margin_x: float = 0.0,
margin_y: float = 0.0
) -> Tuple[float, float, float, float]:
if isinstance(bounds, AxisAlignedBox):
coord_x1 = bounds.left - margin_x
coord_x2 = bounds.right + margin_x
coord_y1 = bounds.back - margin_y
coord_y2 = bounds.front + margin_y
elif isinstance(bounds, Polygon):
coord_x1 = float('inf')
coord_y1 = float('inf')
coord_x2 = float('-inf')
coord_y2 = float('-inf')
for x, y in bounds.getPoints():
coord_x1 = min(coord_x1, x)
coord_y1 = min(coord_y1, y)
coord_x2 = max(coord_x2, x)
coord_y2 = max(coord_y2, y)
else:
raise TypeError("bounds must be either an AxisAlignedBox or a Polygon")
coord_x1 -= margin_x
coord_x2 += margin_x
coord_y1 -= margin_y
coord_y2 += margin_y
grid_x1, grid_y1 = self._coordSpaceToGridSpace(coord_x1, coord_y1)
grid_x2, grid_y2 = self._coordSpaceToGridSpace(coord_x2, coord_y2)
return grid_x1, grid_y1, grid_x2, grid_y2
def _intersectingGridIdxInclusive(self, bounds: Union[AxisAlignedBox, Polygon]) -> Set[Tuple[int, int]]:
grid_x1, grid_y1, grid_x2, grid_y2 = self._getGridCornerPoints(
bounds,
margin_x=-(self._margin_x + self._grid_round_margin_x) * 0.5,
margin_y=-(self._margin_y + self._grid_round_margin_y) * 0.5,
)
grid_idx = set()
for grid_x in range(math.floor(grid_x1), math.ceil(grid_x2)):
for grid_y in range(math.floor(grid_y1), math.ceil(grid_y2)):
grid_idx.add((grid_x, grid_y))
return grid_idx
def _intersectingGridIdxExclusive(self, bounds: Union[AxisAlignedBox, Polygon]) -> Set[Tuple[int, int]]:
grid_x1, grid_y1, grid_x2, grid_y2 = self._getGridCornerPoints(
bounds,
margin_x=(self._margin_x + self._grid_round_margin_x) * 0.5,
margin_y=(self._margin_y + self._grid_round_margin_y) * 0.5,
)
grid_idx = set()
for grid_x in range(math.ceil(grid_x1), math.floor(grid_x2)):
for grid_y in range(math.ceil(grid_y1), math.floor(grid_y2)):
grid_idx.add((grid_x, grid_y))
return grid_idx
def _gridSpaceToCoordSpace(self, x: float, y: float) -> Tuple[float, float]:
grid_x = x * self._grid_width + self._build_volume_bounding_box.left + self._offset_x
grid_y = y * self._grid_height + self._build_volume_bounding_box.back + self._offset_y
return grid_x, grid_y
def _coordSpaceToGridSpace(self, grid_x: float, grid_y: float) -> Tuple[float, float]:
coord_x = (grid_x - self._build_volume_bounding_box.left - self._offset_x) / self._grid_width
coord_y = (grid_y - self._build_volume_bounding_box.back - self._offset_y) / self._grid_height
return coord_x, coord_y
def _checkGridUnderDiscSpace(self, grid_x: int, grid_y: int) -> bool:
left, back = self._gridSpaceToCoordSpace(grid_x, grid_y)
right, front = self._gridSpaceToCoordSpace(grid_x + 1, grid_y + 1)
corners = [(left, back), (right, back), (right, front), (left, front)]
return all([self._checkPointUnderDiscSpace(x, y) for x, y in corners])
def _checkPointUnderDiscSpace(self, x: float, y: float) -> bool:
disc_x, disc_y = self._coordSpaceToDiscSpace(x, y)
distance_to_center_squared = disc_x ** 2 + disc_y ** 2
return distance_to_center_squared <= 1.0
def _coordSpaceToDiscSpace(self, x: float, y: float) -> Tuple[float, float]:
# Transform coordinate system to
#
# coord_build_plate_left = -1
# | coord_build_plate_right = 1
# v (0,1) v
# ┌───────┬───────┐ < coord_build_plate_back = -1
# │ │ │
# │ │(0,0) │
# (-1,0)├───────o───────┤(1,0)
# │ │ │
# │ │ │
# └───────┴───────┘ < coord_build_plate_front = +1
# (0,-1)
disc_x = ((x - self._build_volume_bounding_box.left) / self._build_volume_bounding_box.width) * 2.0 - 1.0
disc_y = ((y - self._build_volume_bounding_box.back) / self._build_volume_bounding_box.depth) * 2.0 - 1.0
return disc_x, disc_y