mirror of
https://github.com/3b1b/manim.git
synced 2025-07-31 22:13:30 +08:00
Added some notes to light.py
This commit is contained in:
@ -42,14 +42,16 @@ inverse_quadratic = lambda maxint,scale,cutoff: inverse_power_law(maxint,scale,c
|
||||
|
||||
|
||||
|
||||
# Note: Overall, this class seems perfectly reasonable to me, the main
|
||||
# thing to be wary of is that calling self.add(submob) puts that submob
|
||||
# at the end of the submobjects list, and hence on top of everything else
|
||||
# which is why the shadow might sometimes end up behind the spotlight
|
||||
class LightSource(VMobject):
|
||||
|
||||
# combines:
|
||||
# a lighthouse
|
||||
# an ambient light
|
||||
# a spotlight
|
||||
# and a shadow
|
||||
|
||||
CONFIG = {
|
||||
"source_point": ORIGIN,
|
||||
"color": LIGHT_COLOR,
|
||||
@ -115,6 +117,8 @@ class LightSource(VMobject):
|
||||
if self.has_screen():
|
||||
self.spotlight.screen = new_screen
|
||||
else:
|
||||
# Note: See below
|
||||
# index = self.submobjects.index(self.spotlight)
|
||||
self.remove(self.spotlight)
|
||||
self.spotlight = Spotlight(
|
||||
source_point = self.source_point,
|
||||
@ -124,6 +128,13 @@ class LightSource(VMobject):
|
||||
screen = new_screen
|
||||
)
|
||||
self.spotlight.move_source_to(self.source_point)
|
||||
|
||||
# Note: This line will make spotlight show up at the end
|
||||
# of the submojects list, which can make it show up on
|
||||
# top of the shadow. To make it show up in the
|
||||
# same spot, you could try the following line,
|
||||
# where "index" is what I defined above:
|
||||
# self.submobjects.insert(index, self.spotlight)
|
||||
self.add(self.spotlight)
|
||||
|
||||
# in any case
|
||||
@ -133,9 +144,12 @@ class LightSource(VMobject):
|
||||
|
||||
|
||||
def move_source_to(self,point):
|
||||
|
||||
apoint = np.array(point)
|
||||
v = apoint - self.source_point
|
||||
# Note: As discussed, things stand to behave better if source
|
||||
# point is a submobject, so that it automatically interpolates
|
||||
# during an animation, and other updates can be defined wrt
|
||||
# that source point's location
|
||||
self.source_point = apoint
|
||||
self.lighthouse.next_to(apoint,DOWN,buff = 0)
|
||||
self.ambient_light.move_source_to(apoint)
|
||||
@ -143,12 +157,6 @@ class LightSource(VMobject):
|
||||
self.spotlight.move_source_to(apoint)
|
||||
self.update()
|
||||
return self
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
def set_radius(self,new_radius):
|
||||
self.radius = new_radius
|
||||
@ -169,14 +177,15 @@ class LightSource(VMobject):
|
||||
for point in self.screen.get_anchors():
|
||||
projected_screen_points.append(self.spotlight.project(point))
|
||||
|
||||
print "projected", self.screen.get_anchors(), "onto", projected_screen_points
|
||||
|
||||
|
||||
# print "projected", self.screen.get_anchors(), "onto", projected_screen_points
|
||||
|
||||
projected_source = project_along_vector(self.source_point,self.spotlight.projection_direction())
|
||||
|
||||
projected_point_cloud_3d = np.append(projected_screen_points,
|
||||
np.reshape(projected_source,(1,3)),axis = 0)
|
||||
projected_point_cloud_3d = np.append(
|
||||
projected_screen_points,
|
||||
np.reshape(projected_source,(1,3)),
|
||||
axis = 0
|
||||
)
|
||||
rotation_matrix = z_to_vector(self.spotlight.projection_direction())
|
||||
back_rotation_matrix = rotation_matrix.T # i. e. its inverse
|
||||
|
||||
@ -223,6 +232,9 @@ class LightSource(VMobject):
|
||||
new_anchors = np.append(new_anchors,anchors[index:],axis = 0)
|
||||
self.shadow.set_points_as_corners(new_anchors)
|
||||
|
||||
# Note: Theoretically this should not be necessary as long as we make
|
||||
# sure the shadow shows up after the spotlight in the submobjects list.
|
||||
#
|
||||
# shift it closer to the camera so it is in front of the spotlight
|
||||
self.shadow.shift(1e-5*self.spotlight.projection_direction())
|
||||
self.shadow.mark_paths_closed = True
|
||||
@ -293,6 +305,10 @@ class AmbientLight(VMobject):
|
||||
|
||||
# in theory, this method is only called once, right?
|
||||
# so removing submobs shd not be necessary
|
||||
#
|
||||
# Note: Usually, yes, it is only called within Mobject.__init__,
|
||||
# but there is no strong guarantee of that, and you may want certain
|
||||
# update functions to regenerate points here and there.
|
||||
for submob in self.submobjects:
|
||||
self.remove(submob)
|
||||
|
||||
@ -313,7 +329,7 @@ class AmbientLight(VMobject):
|
||||
|
||||
|
||||
def move_source_to(self,point):
|
||||
|
||||
# Note: Best to rewrite in terms of VectorizedPoint source_point
|
||||
v = np.array(point) - self.source_point
|
||||
self.source_point = np.array(point)
|
||||
self.shift(v)
|
||||
@ -348,6 +364,10 @@ class Spotlight(VMobject):
|
||||
}
|
||||
|
||||
def projection_direction(self):
|
||||
# Note: This seems reasonable, though for it to work you'd
|
||||
# need to be sure that any 3d scene including a spotlight
|
||||
# somewhere assigns that spotlights "camera" attribute
|
||||
# to be the camera associated with that scene.
|
||||
if self.camera == None:
|
||||
return OUT
|
||||
else:
|
||||
@ -376,6 +396,10 @@ class Spotlight(VMobject):
|
||||
|
||||
|
||||
def new_sector(self,r,dr,lower_angle,upper_angle):
|
||||
# Note: I'm not looking _too_ closely at the implementation
|
||||
# of these updates based on viewing angles and such. It seems to
|
||||
# behave as intended, but let me know if you'd like more thorough
|
||||
# scrutiny
|
||||
|
||||
alpha = self.max_opacity * self.opacity_function(r)
|
||||
annular_sector = AnnularSector(
|
||||
@ -467,13 +491,12 @@ class Spotlight(VMobject):
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
|
||||
def dimming(self,new_alpha):
|
||||
old_alpha = self.max_opacity
|
||||
self.max_opacity = new_alpha
|
||||
for submob in self.submobjects:
|
||||
# Note: Maybe it'd be best to have a Shadow class so that the
|
||||
# type can be checked directly?
|
||||
if type(submob) != AnnularSector:
|
||||
# it's the shadow, don't dim it
|
||||
continue
|
||||
@ -498,6 +521,14 @@ class Spotlight(VMobject):
|
||||
submob.set_fill(opacity = alpha)
|
||||
|
||||
|
||||
# Note: Stylistically, I typically keep all of the implementation for an
|
||||
# update inside the relevant Animation or ContinualAnimation, rather than
|
||||
# in the mobject. Things are fine the way you've done it, but sometimes
|
||||
# I personally like a nice conceptual divide between all the things that
|
||||
# determine how the mobject is displayed in a single moment (implement in
|
||||
# the mobject's class itself) and all the things determining how that changes.
|
||||
#
|
||||
# Up to you though.
|
||||
|
||||
class ScreenTracker(ContinualAnimation):
|
||||
|
||||
|
Reference in New Issue
Block a user