io_import_images_as_planes.py

# SPDX-License-Identifier: GPL-2.0-or-later

bl_info = {
    "name": "Import Images as Planes",
    "author": "Florian Meyer (tstscr), mont29, matali, Ted Schundler (SpkyElctrc), mrbimax",
    "version": (3, 5, 0),
    "blender": (2, 91, 0),
    "location": "File > Import > Images as Planes or Add > Image > Images as Planes",
    "description": "Imports images and creates planes with the appropriate aspect ratio. "
                   "The images are mapped to the planes.",
    "warning": "",
    "doc_url": "{BLENDER_MANUAL_URL}/addons/import_export/images_as_planes.html",
    "support": 'OFFICIAL',
    "category": "Import-Export",
}

import os
import warnings
import re
from itertools import count, repeat
from collections import namedtuple
from math import pi

import bpy
from bpy.types import Operator
from bpy.app.translations import pgettext_tip as tip_
from mathutils import Vector

from bpy.props import (
    StringProperty,
    BoolProperty,
    EnumProperty,
    FloatProperty,
    CollectionProperty,
)

from bpy_extras.object_utils import (
    AddObjectHelper,
    world_to_camera_view,
)

from bpy_extras.image_utils import load_image

# -----------------------------------------------------------------------------
# Module-level Shared State

watched_objects = {}  # used to trigger compositor updates on scene updates


# -----------------------------------------------------------------------------
# Misc utils.

def add_driver_prop(driver, name, type, id, path):
    """Configure a new driver variable."""
    dv = driver.variables.new()
    dv.name = name
    dv.type = 'SINGLE_PROP'
    target = dv.targets[0]
    target.id_type = type
    target.id = id
    target.data_path = path


# -----------------------------------------------------------------------------
# Image loading

ImageSpec = namedtuple(
    'ImageSpec',
    ['image', 'size', 'frame_start', 'frame_offset', 'frame_duration'])

num_regex = re.compile('[0-9]')  # Find a single number
nums_regex = re.compile('[0-9]+')  # Find a set of numbers


def find_image_sequences(files):
    """From a group of files, detect image sequences.

    This returns a generator of tuples, which contain the filename,
    start frame, and length of the detected sequence

    >>> list(find_image_sequences([
    ...     "test2-001.jp2", "test2-002.jp2",
    ...     "test3-003.jp2", "test3-004.jp2", "test3-005.jp2", "test3-006.jp2",
    ...     "blaah"]))
    [('blaah', 1, 1), ('test2-001.jp2', 1, 2), ('test3-003.jp2', 3, 4)]

    """
    files = iter(sorted(files))
    prev_file = None
    pattern = ""
    matches = []
    segment = None
    length = 1
    for filename in files:
        new_pattern = num_regex.sub('#', filename)
        new_matches = list(map(int, nums_regex.findall(filename)))
        if new_pattern == pattern:
            # this file looks like it may be in sequence from the previous

            # if there are multiple sets of numbers, figure out what changed
            if segment is None:
                for i, prev, cur in zip(count(), matches, new_matches):
                    if prev != cur:
                        segment = i
                        break

            # did it only change by one?
            for i, prev, cur in zip(count(), matches, new_matches):
                if i == segment:
                    # We expect this to increment
                    prev = prev + length
                if prev != cur:
                    break

            # All good!
            else:
                length += 1
                continue

        # No continuation -> spit out what we found and reset counters
        if prev_file:
            if length > 1:
                yield prev_file, matches[segment], length
            else:
                yield prev_file, 1, 1

        prev_file = filename
        matches = new_matches
        pattern = new_pattern
        segment = None
        length = 1

    if prev_file:
        if length > 1:
            yield prev_file, matches[segment], length
        else:
            yield prev_file, 1, 1


def load_images(filenames, directory, force_reload=False, frame_start=1, find_sequences=False):
    """Wrapper for bpy's load_image

    Loads a set of images, movies, or even image sequences
    Returns a generator of ImageSpec wrapper objects later used for texture setup
    """
    if find_sequences:  # if finding sequences, we need some pre-processing first
        file_iter = find_image_sequences(filenames)
    else:
        file_iter = zip(filenames, repeat(1), repeat(1))

    for filename, offset, frames in file_iter:
        image = load_image(filename, directory, check_existing=True, force_reload=force_reload)

        # Size is unavailable for sequences, so we grab it early
        size = tuple(image.size)

        if image.source == 'MOVIE':
            # Blender BPY BUG!
            # This number is only valid when read a second time in 2.77
            # This repeated line is not a mistake
            frames = image.frame_duration
            frames = image.frame_duration

        elif frames > 1:  # Not movie, but multiple frames -> image sequence
            image.source = 'SEQUENCE'

        yield ImageSpec(image, size, frame_start, offset - 1, frames)


# -----------------------------------------------------------------------------
# Position & Size Helpers

def offset_planes(planes, gap, axis):
    """Offset planes from each other by `gap` amount along a _local_ vector `axis`

    For example, offset_planes([obj1, obj2], 0.5, Vector(0, 0, 1)) will place
    obj2 0.5 blender units away from obj1 along the local positive Z axis.

    This is in local space, not world space, so all planes should share
    a common scale and rotation.
    """
    prior = planes[0]
    offset = Vector()
    for current in planes[1:]:
        local_offset = abs((prior.dimensions + current.dimensions).dot(axis)) / 2.0 + gap

        offset += local_offset * axis
        current.location = current.matrix_world @ offset

        prior = current


def compute_camera_size(context, center, fill_mode, aspect):
    """Determine how large an object needs to be to fit or fill the camera's field of view."""
    scene = context.scene
    camera = scene.camera
    view_frame = camera.data.view_frame(scene=scene)
    frame_size = \
        Vector([max(v[i] for v in view_frame) for i in range(3)]) - \
        Vector([min(v[i] for v in view_frame) for i in range(3)])
    camera_aspect = frame_size.x / frame_size.y

    # Convert the frame size to the correct sizing at a given distance
    if camera.type == 'ORTHO':
        frame_size = frame_size.xy
    else:
        # Perspective transform
        distance = world_to_camera_view(scene, camera, center).z
        frame_size = distance * frame_size.xy / (-view_frame[0].z)

    # Determine what axis to match to the camera
    match_axis = 0  # match the Y axis size
    match_aspect = aspect
    if (fill_mode == 'FILL' and aspect > camera_aspect) or \
            (fill_mode == 'FIT' and aspect < camera_aspect):
        match_axis = 1  # match the X axis size
        match_aspect = 1.0 / aspect

    # scale the other axis to the correct aspect
    frame_size[1 - match_axis] = frame_size[match_axis] / match_aspect

    return frame_size


def center_in_camera(scene, camera, obj, axis=(1, 1)):
    """Center object along specified axis of the camera"""
    camera_matrix_col = camera.matrix_world.col
    location = obj.location

    # Vector from the camera's world coordinate center to the object's center
    delta = camera_matrix_col[3].xyz - location

    # How far off center we are along the camera's local X
    camera_x_mag = delta.dot(camera_matrix_col[0].xyz) * axis[0]
    # How far off center we are along the camera's local Y
    camera_y_mag = delta.dot(camera_matrix_col[1].xyz) * axis[1]

    # Now offset only along camera local axis
    offset = camera_matrix_col[0].xyz * camera_x_mag + \
        camera_matrix_col[1].xyz * camera_y_mag

    obj.location = location + offset


# -----------------------------------------------------------------------------
# Cycles/Eevee utils

def get_input_nodes(node, links):
    """Get nodes that are a inputs to the given node"""
    # Get all links going to node.
    input_links = {lnk for lnk in links if lnk.to_node == node}
    # Sort those links, get their input nodes (and avoid doubles!).
    sorted_nodes = []
    done_nodes = set()
    for socket in node.inputs:
        done_links = set()
        for link in input_links:
            nd = link.from_node
            if nd in done_nodes:
                # Node already treated!
                done_links.add(link)
            elif link.to_socket == socket:
                sorted_nodes.append(nd)
                done_links.add(link)
                done_nodes.add(nd)
        input_links -= done_links
    return sorted_nodes


def auto_align_nodes(node_tree):
    """Given a shader node tree, arrange nodes neatly relative to the output node."""
    x_gap = 200
    y_gap = 180
    nodes = node_tree.nodes
    links = node_tree.links
    output_node = None
    for node in nodes:
        if node.type == 'OUTPUT_MATERIAL' or node.type == 'GROUP_OUTPUT':
            output_node = node
            break

    else:  # Just in case there is no output
        return

    def align(to_node):
        from_nodes = get_input_nodes(to_node, links)
        for i, node in enumerate(from_nodes):
            node.location.x = min(node.location.x, to_node.location.x - x_gap)
            node.location.y = to_node.location.y
            node.location.y -= i * y_gap
            node.location.y += (len(from_nodes) - 1) * y_gap / (len(from_nodes))
            align(node)

    align(output_node)


def clean_node_tree(node_tree):
    """Clear all nodes in a shader node tree except the output.

    Returns the output node
    """
    nodes = node_tree.nodes
    for node in list(nodes):  # copy to avoid altering the loop's data source
        if not node.type == 'OUTPUT_MATERIAL':
            nodes.remove(node)

    return node_tree.nodes[0]


def get_shadeless_node(dest_node_tree):
    """Return a "shadless" cycles/eevee node, creating a node group if nonexistent"""
    try:
        node_tree = bpy.data.node_groups['IAP_SHADELESS']

    except KeyError:
        # need to build node shadeless node group
        node_tree = bpy.data.node_groups.new('IAP_SHADELESS', 'ShaderNodeTree')
        output_node = node_tree.nodes.new('NodeGroupOutput')
        input_node = node_tree.nodes.new('NodeGroupInput')

        node_tree.outputs.new('NodeSocketShader', 'Shader')
        node_tree.inputs.new('NodeSocketColor', 'Color')

        # This could be faster as a transparent shader, but then no ambient occlusion
        diffuse_shader = node_tree.nodes.new('ShaderNodeBsdfDiffuse')
        node_tree.links.new(diffuse_shader.inputs[0], input_node.outputs[0])

        emission_shader = node_tree.nodes.new('ShaderNodeEmission')
        node_tree.links.new(emission_shader.inputs[0], input_node.outputs[0])

        light_path = node_tree.nodes.new('ShaderNodeLightPath')
        is_glossy_ray = light_path.outputs['Is Glossy Ray']
        is_shadow_ray = light_path.outputs['Is Shadow Ray']
        ray_depth = light_path.outputs['Ray Depth']
        transmission_depth = light_path.outputs['Transmission Depth']

        unrefracted_depth = node_tree.nodes.new('ShaderNodeMath')
        unrefracted_depth.operation = 'SUBTRACT'
        unrefracted_depth.label = 'Bounce Count'
        node_tree.links.new(unrefracted_depth.inputs[0], ray_depth)
        node_tree.links.new(unrefracted_depth.inputs[1], transmission_depth)

        refracted = node_tree.nodes.new('ShaderNodeMath')
        refracted.operation = 'SUBTRACT'
        refracted.label = 'Camera or Refracted'
        refracted.inputs[0].default_value = 1.0
        node_tree.links.new(refracted.inputs[1], unrefracted_depth.outputs[0])

        reflection_limit = node_tree.nodes.new('ShaderNodeMath')
        reflection_limit.operation = 'SUBTRACT'
        reflection_limit.label = 'Limit Reflections'
        reflection_limit.inputs[0].default_value = 2.0
        node_tree.links.new(reflection_limit.inputs[1], ray_depth)

        camera_reflected = node_tree.nodes.new('ShaderNodeMath')
        camera_reflected.operation = 'MULTIPLY'
        camera_reflected.label = 'Camera Ray to Glossy'
        node_tree.links.new(camera_reflected.inputs[0], reflection_limit.outputs[0])
        node_tree.links.new(camera_reflected.inputs[1], is_glossy_ray)

        shadow_or_reflect = node_tree.nodes.new('ShaderNodeMath')
        shadow_or_reflect.operation = 'MAXIMUM'
        shadow_or_reflect.label = 'Shadow or Reflection?'
        node_tree.links.new(shadow_or_reflect.inputs[0], camera_reflected.outputs[0])
        node_tree.links.new(shadow_or_reflect.inputs[1], is_shadow_ray)

        shadow_or_reflect_or_refract = node_tree.nodes.new('ShaderNodeMath')
        shadow_or_reflect_or_refract.operation = 'MAXIMUM'
        shadow_or_reflect_or_refract.label = 'Shadow, Reflect or Refract?'
        node_tree.links.new(shadow_or_reflect_or_refract.inputs[0], shadow_or_reflect.outputs[0])
        node_tree.links.new(shadow_or_reflect_or_refract.inputs[1], refracted.outputs[0])

        mix_shader = node_tree.nodes.new('ShaderNodeMixShader')
        node_tree.links.new(mix_shader.inputs[0], shadow_or_reflect_or_refract.outputs[0])
        node_tree.links.new(mix_shader.inputs[1], diffuse_shader.outputs[0])
        node_tree.links.new(mix_shader.inputs[2], emission_shader.outputs[0])

        node_tree.links.new(output_node.inputs[0], mix_shader.outputs[0])

        auto_align_nodes(node_tree)

    group_node = dest_node_tree.nodes.new("ShaderNodeGroup")
    group_node.node_tree = node_tree

    return group_node


# -----------------------------------------------------------------------------
# Corner Pin Driver Helpers

@bpy.app.handlers.persistent
def check_drivers(*args, **kwargs):
    """Check if watched objects in a scene have changed and trigger compositor update

    This is part of a hack to ensure the compositor updates
    itself when the objects used for drivers change.

    It only triggers if transformation matricies change to avoid
    a cyclic loop of updates.
    """
    if not watched_objects:
        # if there is nothing to watch, don't bother running this
        bpy.app.handlers.depsgraph_update_post.remove(check_drivers)
        return

    update = False
    for name, matrix in list(watched_objects.items()):
        try:
            obj = bpy.data.objects[name]
        except KeyError:
            # The user must have removed this object
            del watched_objects[name]
        else:
            new_matrix = tuple(map(tuple, obj.matrix_world)).__hash__()
            if new_matrix != matrix:
                watched_objects[name] = new_matrix
                update = True

    if update:
        # Trick to re-evaluate drivers
        bpy.context.scene.frame_current = bpy.context.scene.frame_current


def register_watched_object(obj):
    """Register an object to be monitored for transformation changes"""
    name = obj.name

    # known object? -> we're done
    if name in watched_objects:
        return

    if not watched_objects:
        # make sure check_drivers is active
        bpy.app.handlers.depsgraph_update_post.append(check_drivers)

    watched_objects[name] = None


def find_plane_corner(object_name, x, y, axis, camera=None, *args, **kwargs):
    """Find the location in camera space of a plane's corner"""
    if args or kwargs:
        # I've added args / kwargs as a compatibility measure with future versions
        warnings.warn("Unknown Parameters Passed to \"Images as Planes\".  Maybe you need to upgrade?")

    plane = bpy.data.objects[object_name]

    # Passing in camera doesn't work before 2.78, so we use the current one
    camera = camera or bpy.context.scene.camera

    # Hack to ensure compositor updates on future changes
    register_watched_object(camera)
    register_watched_object(plane)

    scale = plane.scale * 2.0
    v = plane.dimensions.copy()
    v.x *= x / scale.x
    v.y *= y / scale.y
    v = plane.matrix_world @ v

    camera_vertex = world_to_camera_view(
        bpy.context.scene, camera, v)

    return camera_vertex[axis]


@bpy.app.handlers.persistent
def register_driver(*args, **kwargs):
    """Register the find_plane_corner function for use with drivers"""
    bpy.app.driver_namespace['import_image__find_plane_corner'] = find_plane_corner


# -----------------------------------------------------------------------------
# Compositing Helpers

def group_in_frame(node_tree, name, nodes):
    frame_node = node_tree.nodes.new("NodeFrame")
    frame_node.label = name
    frame_node.name = name + "_frame"

    min_pos = Vector(nodes[0].location)
    max_pos = min_pos.copy()

    for node in nodes:
        top_left = node.location
        bottom_right = top_left + Vector((node.width, -node.height))

        for i in (0, 1):
            min_pos[i] = min(min_pos[i], top_left[i], bottom_right[i])
            max_pos[i] = max(max_pos[i], top_left[i], bottom_right[i])

        node.parent = frame_node

    frame_node.width = max_pos[0] - min_pos[0] + 50
    frame_node.height = max(max_pos[1] - min_pos[1] + 50, 450)
    frame_node.shrink = True

    return frame_node


def position_frame_bottom_left(node_tree, frame_node):
    newpos = Vector((100000, 100000))  # start reasonably far top / right

    # Align with the furthest left
    for node in node_tree.nodes.values():
        if node != frame_node and node.parent != frame_node:
            newpos.x = min(newpos.x, node.location.x + 30)

    # As high as we can get without overlapping anything to the right
    for node in node_tree.nodes.values():
        if node != frame_node and not node.parent:
            if node.location.x < newpos.x + frame_node.width:
                print("Below", node.name, node.location, node.height, node.dimensions)
                newpos.y = min(newpos.y, node.location.y - max(node.dimensions.y, node.height) - 20)

    frame_node.location = newpos


def setup_compositing(context, plane, img_spec):
    # Node Groups only work with "new" dependency graph and even
    # then it has some problems with not updating the first time
    # So instead this groups with a node frame, which works reliably

    scene = context.scene
    scene.use_nodes = True
    node_tree = scene.node_tree
    name = plane.name

    image_node = node_tree.nodes.new("CompositorNodeImage")
    image_node.name = name + "_image"
    image_node.image = img_spec.image
    image_node.location = Vector((0, 0))
    image_node.frame_start = img_spec.frame_start
    image_node.frame_offset = img_spec.frame_offset
    image_node.frame_duration = img_spec.frame_duration

    scale_node = node_tree.nodes.new("CompositorNodeScale")
    scale_node.name = name + "_scale"
    scale_node.space = 'RENDER_SIZE'
    scale_node.location = image_node.location + \
        Vector((image_node.width + 20, 0))
    scale_node.show_options = False

    cornerpin_node = node_tree.nodes.new("CompositorNodeCornerPin")
    cornerpin_node.name = name + "_cornerpin"
    cornerpin_node.location = scale_node.location + \
        Vector((0, -scale_node.height))

    node_tree.links.new(scale_node.inputs[0], image_node.outputs[0])
    node_tree.links.new(cornerpin_node.inputs[0], scale_node.outputs[0])

    # Put all the nodes in a frame for organization
    frame_node = group_in_frame(
        node_tree, name,
        (image_node, scale_node, cornerpin_node)
    )

    # Position frame at bottom / left
    position_frame_bottom_left(node_tree, frame_node)

    # Configure Drivers
    for corner in cornerpin_node.inputs[1:]:
        id = corner.identifier
        x = -1 if 'Left' in id else 1
        y = -1 if 'Lower' in id else 1
        drivers = corner.driver_add('default_value')
        for i, axis_fcurve in enumerate(drivers):
            driver = axis_fcurve.driver
            # Always use the current camera
            add_driver_prop(driver, 'camera', 'SCENE', scene, 'camera')
            # Track camera location to ensure Deps Graph triggers (not used in the call)
            add_driver_prop(driver, 'cam_loc_x', 'OBJECT', scene.camera, 'location[0]')
            # Don't break if the name changes
            add_driver_prop(driver, 'name', 'OBJECT', plane, 'name')
            driver.expression = "import_image__find_plane_corner(name or %s, %d, %d, %d, camera=camera)" % (
                repr(plane.name),
                x, y, i
            )
            driver.type = 'SCRIPTED'
            driver.is_valid = True
            axis_fcurve.is_valid = True
            driver.expression = "%s" % driver.expression

    context.view_layer.update()


# -----------------------------------------------------------------------------
# Operator

class IMPORT_IMAGE_OT_to_plane(Operator, AddObjectHelper):
    """Create mesh plane(s) from image files with the appropriate aspect ratio"""

    bl_idname = "import_image.to_plane"
    bl_label = "Import Images as Planes"
    bl_options = {'REGISTER', 'PRESET', 'UNDO'}

    # ----------------------
    # File dialog properties
    files: CollectionProperty(type=bpy.types.OperatorFileListElement, options={'HIDDEN', 'SKIP_SAVE'})

    directory: StringProperty(maxlen=1024, subtype='FILE_PATH', options={'HIDDEN', 'SKIP_SAVE'})

    filter_image: BoolProperty(default=True, options={'HIDDEN', 'SKIP_SAVE'})
    filter_movie: BoolProperty(default=True, options={'HIDDEN', 'SKIP_SAVE'})
    filter_folder: BoolProperty(default=True, options={'HIDDEN', 'SKIP_SAVE'})

    # ----------------------
    # Properties - Importing
    force_reload: BoolProperty(
        name="Force Reload", default=False,
        description="Force reloading of the image if already opened elsewhere in Blender"
    )

    image_sequence: BoolProperty(
        name="Animate Image Sequences", default=False,
        description="Import sequentially numbered images as an animated "
                    "image sequence instead of separate planes"
    )

    # -------------------------------------
    # Properties - Position and Orientation
    axis_id_to_vector = {
        'X+': Vector(( 1,  0,  0)),
        'Y+': Vector(( 0,  1,  0)),
        'Z+': Vector(( 0,  0,  1)),
        'X-': Vector((-1,  0,  0)),
        'Y-': Vector(( 0, -1,  0)),
        'Z-': Vector(( 0,  0, -1)),
    }

    offset: BoolProperty(name="Offset Planes", default=True, description="Offset Planes From Each Other")

    OFFSET_MODES = (
        ('X+', "X+", "Side by Side to the Left"),
        ('Y+', "Y+", "Side by Side, Downward"),
        ('Z+', "Z+", "Stacked Above"),
        ('X-', "X-", "Side by Side to the Right"),
        ('Y-', "Y-", "Side by Side, Upward"),
        ('Z-', "Z-", "Stacked Below"),
    )
    offset_axis: EnumProperty(
        name="Orientation", default='X+', items=OFFSET_MODES,
        description="How planes are oriented relative to each others' local axis"
    )

    offset_amount: FloatProperty(
        name="Offset", soft_min=0, default=0.1, description="Space between planes",
        subtype='DISTANCE', unit='LENGTH'
    )

    AXIS_MODES = (
        ('X+', "X+", "Facing Positive X"),
        ('Y+', "Y+", "Facing Positive Y"),
        ('Z+', "Z+ (Up)", "Facing Positive Z"),
        ('X-', "X-", "Facing Negative X"),
        ('Y-', "Y-", "Facing Negative Y"),
        ('Z-', "Z- (Down)", "Facing Negative Z"),
        ('CAM', "Face Camera", "Facing Camera"),
        ('CAM_AX', "Main Axis", "Facing the Camera's dominant axis"),
    )
    align_axis: EnumProperty(
        name="Align", default='CAM_AX', items=AXIS_MODES,
        description="How to align the planes"
    )
    # prev_align_axis is used only by update_size_model
    prev_align_axis: EnumProperty(
        items=AXIS_MODES + (('NONE', '', ''),), default='NONE', options={'HIDDEN', 'SKIP_SAVE'})
    align_track: BoolProperty(
        name="Track Camera", default=False, description="Always face the camera"
    )

    # -----------------
    # Properties - Size
    def update_size_mode(self, context):
        """If sizing relative to the camera, always face the camera"""
        if self.size_mode == 'CAMERA':
            self.prev_align_axis = self.align_axis
            self.align_axis = 'CAM'
        else:
            # if a different alignment was set revert to that when
            # size mode is changed
            if self.prev_align_axis != 'NONE':
                self.align_axis = self.prev_align_axis
                self._prev_align_axis = 'NONE'

    SIZE_MODES = (
        ('ABSOLUTE', "Absolute", "Use absolute size"),
        ('CAMERA', "Camera Relative", "Scale to the camera frame"),
        ('DPI', "Dpi", "Use definition of the image as dots per inch"),
        ('DPBU', "Dots/BU", "Use definition of the image as dots per Blender Unit"),
    )
    size_mode: EnumProperty(
        name="Size Mode", default='ABSOLUTE', items=SIZE_MODES,
        update=update_size_mode,
        description="How the size of the plane is computed")

    FILL_MODES = (
        ('FILL', "Fill", "Fill camera frame, spilling outside the frame"),
        ('FIT', "Fit", "Fit entire image within the camera frame"),
    )
    fill_mode: EnumProperty(name="Scale", default='FILL', items=FILL_MODES,
                             description="How large in the camera frame is the plane")

    height: FloatProperty(name="Height", description="Height of the created plane",
                           default=1.0, min=0.001, soft_min=0.001, subtype='DISTANCE', unit='LENGTH')

    factor: FloatProperty(name="Definition", min=1.0, default=600.0,
                           description="Number of pixels per inch or Blender Unit")

    # ------------------------------
    # Properties - Material / Shader
    SHADERS = (
        ('PRINCIPLED',"Principled","Principled Shader"),
        ('SHADELESS', "Shadeless", "Only visible to camera and reflections"),
        ('EMISSION', "Emit", "Emission Shader"),
    )
    shader: EnumProperty(name="Shader", items=SHADERS, default='PRINCIPLED', description="Node shader to use")

    emit_strength: FloatProperty(
        name="Strength", min=0.0, default=1.0, soft_max=10.0,
        step=100, description="Brightness of Emission Texture")

    use_transparency: BoolProperty(
        name="Use Alpha", default=True,
        description="Use alpha channel for transparency")

    BLEND_METHODS = (
        ('BLEND',"Blend","Render polygon transparent, depending on alpha channel of the texture"),
        ('CLIP', "Clip","Use the alpha threshold to clip the visibility (binary visibility)"),
        ('HASHED', "Hashed","Use noise to dither the binary visibility (works well with multi-samples)"),
        ('OPAQUE', "Opaque","Render surface without transparency"),
    )
    blend_method: EnumProperty(name="Blend Mode", items=BLEND_METHODS, default='BLEND', description="Blend Mode for Transparent Faces")

    SHADOW_METHODS = (
        ('CLIP', "Clip","Use the alpha threshold to clip the visibility (binary visibility)"),
        ('HASHED', "Hashed","Use noise to dither the binary visibility (works well with multi-samples)"),
        ('OPAQUE',"Opaque","Material will cast shadows without transparency"),
        ('NONE',"None","Material will cast no shadow"),
    )
    shadow_method: EnumProperty(name="Shadow Mode", items=SHADOW_METHODS, default='CLIP', description="Shadow mapping method")

    use_backface_culling: BoolProperty(
        name="Backface Culling", default=False,
        description="Use back face culling to hide the back side of faces")

    show_transparent_back: BoolProperty(
        name="Show Backface", default=True,
        description="Render multiple transparent layers (may introduce transparency sorting problems)")

    overwrite_material: BoolProperty(
        name="Overwrite Material", default=True,
        description="Overwrite existing Material (based on material name)")

    compositing_nodes: BoolProperty(
        name="Setup Corner Pin", default=False,
        description="Build Compositor Nodes to reference this image "
                    "without re-rendering")

    # ------------------
    # Properties - Image
    INTERPOLATION_MODES = (
        ('Linear', "Linear", "Linear interpolation"),
        ('Closest', "Closest", "No interpolation (sample closest texel)"),
        ('Cubic', "Cubic", "Cubic interpolation"),
        ('Smart', "Smart", "Bicubic when magnifying, else bilinear (OSL only)"),
    )
    interpolation: EnumProperty(name="Interpolation", items=INTERPOLATION_MODES, default='Linear', description="Texture interpolation")

    EXTENSION_MODES = (
        ('CLIP', "Clip", "Clip to image size and set exterior pixels as transparent"),
        ('EXTEND', "Extend", "Extend by repeating edge pixels of the image"),
        ('REPEAT', "Repeat", "Cause the image to repeat horizontally and vertically"),
    )
    extension: EnumProperty(name="Extension", items=EXTENSION_MODES, default='CLIP', description="How the image is extrapolated past its original bounds")

    t = bpy.types.Image.bl_rna.properties["alpha_mode"]
    alpha_mode_items = tuple((e.identifier, e.name, e.description) for e in t.enum_items)
    alpha_mode: EnumProperty(
        name=t.name, items=alpha_mode_items, default=t.default,
        description=t.description)

    t = bpy.types.ImageUser.bl_rna.properties["use_auto_refresh"]
    use_auto_refresh: BoolProperty(name=t.name, default=True, description=t.description)

    relative: BoolProperty(name="Relative Paths", default=True, description="Use relative file paths")

    # -------
    # Draw UI
    def draw_import_config(self, context):
        # --- Import Options --- #
        layout = self.layout
        box = layout.box()

        box.label(text="Import Options:", icon='IMPORT')
        row = box.row()
        row.active = bpy.data.is_saved
        row.prop(self, "relative")

        box.prop(self, "force_reload")
        box.prop(self, "image_sequence")

    def draw_material_config(self, context):
        # --- Material / Rendering Properties --- #
        layout = self.layout
        box = layout.box()

        box.label(text="Compositing Nodes:", icon='RENDERLAYERS')
        box.prop(self, "compositing_nodes")
        layout = self.layout
        box = layout.box()
        box.label(text="Material Settings:", icon='MATERIAL')

        box.label(text="Material Type")
        row = box.row()
        row.prop(self, 'shader', expand=True)
        if self.shader == 'EMISSION':
            box.prop(self, "emit_strength")

        box.label(text="Blend Mode")
        row = box.row()
        row.prop(self, 'blend_method', expand=True)
        if self.use_transparency and self.alpha_mode != "NONE" and self.blend_method == "OPAQUE":
            box.label(text="'Opaque' does not support alpha", icon="ERROR")
        if self.blend_method == 'BLEND':
            row = box.row()
            row.prop(self, "show_transparent_back")

        box.label(text="Shadow Mode")
        row = box.row()
        row.prop(self, 'shadow_method', expand=True)

        row = box.row()
        row.prop(self, "use_backface_culling")

        engine = context.scene.render.engine
        if engine not in ('CYCLES', 'BLENDER_EEVEE', 'BLENDER_WORKBENCH'):
            box.label(text=tip_("%s is not supported") % engine, icon='ERROR')

        box.prop(self, "overwrite_material")
        layout = self.layout
        box = layout.box()
        box.label(text="Texture Settings:", icon='TEXTURE')
        box.label(text="Interpolation")
        row = box.row()
        row.prop(self, 'interpolation', expand=True)
        box.label(text="Extension")
        row = box.row()
        row.prop(self, 'extension', expand=True)
        row = box.row()
        row.prop(self, "use_transparency")
        if self.use_transparency:
            sub = row.row()
            sub.prop(self, "alpha_mode", text="")
        row = box.row()
        row.prop(self, "use_auto_refresh")

    def draw_spatial_config(self, context):
        # --- Spatial Properties: Position, Size and Orientation --- #
        layout = self.layout
        box = layout.box()

        box.label(text="Position:", icon='SNAP_GRID')
        box.prop(self, "offset")
        col = box.column()
        row = col.row()
        row.prop(self, "offset_axis", expand=True)
        row = col.row()
        row.prop(self, "offset_amount")
        col.enabled = self.offset

        box.label(text="Plane dimensions:", icon='ARROW_LEFTRIGHT')
        row = box.row()
        row.prop(self, "size_mode", expand=True)
        if self.size_mode == 'ABSOLUTE':
            box.prop(self, "height")
        elif self.size_mode == 'CAMERA':
            row = box.row()
            row.prop(self, "fill_mode", expand=True)
        else:
            box.prop(self, "factor")

        box.label(text="Orientation:")
        row = box.row()
        row.enabled = 'CAM' not in self.size_mode
        row.prop(self, "align_axis")
        row = box.row()
        row.enabled = 'CAM' in self.align_axis
        row.alignment = 'RIGHT'
        row.prop(self, "align_track")

    def draw(self, context):

        # Draw configuration sections
        self.draw_import_config(context)
        self.draw_material_config(context)
        self.draw_spatial_config(context)

    # -------------------------------------------------------------------------
    # Core functionality
    def invoke(self, context, event):
        engine = context.scene.render.engine
        if engine not in {'CYCLES', 'BLENDER_EEVEE'}:
            if engine != 'BLENDER_WORKBENCH':
                self.report({'ERROR'}, tip_("Cannot generate materials for unknown %s render engine") % engine)
                return {'CANCELLED'}
            else:
                self.report({'WARNING'},
                            tip_("Generating Cycles/EEVEE compatible material, but won't be visible with %s engine") % engine)

        # Open file browser
        context.window_manager.fileselect_add(self)
        return {'RUNNING_MODAL'}

    def execute(self, context):
        if not bpy.data.is_saved:
            self.relative = False

        # this won't work in edit mode
        editmode = context.preferences.edit.use_enter_edit_mode
        context.preferences.edit.use_enter_edit_mode = False
        if context.active_object and context.active_object.mode != 'OBJECT':
            bpy.ops.object.mode_set(mode='OBJECT')

        self.import_images(context)

        context.preferences.edit.use_enter_edit_mode = editmode

        return {'FINISHED'}

    def import_images(self, context):

        # load images / sequences
        images = tuple(load_images(
            (fn.name for fn in self.files),
            self.directory,
            force_reload=self.force_reload,
            find_sequences=self.image_sequence
        ))

        # Create individual planes
        planes = [self.single_image_spec_to_plane(context, img_spec) for img_spec in images]

        context.view_layer.update()

        # Align planes relative to each other
        if self.offset:
            offset_axis = self.axis_id_to_vector[self.offset_axis]
            offset_planes(planes, self.offset_amount, offset_axis)

            if self.size_mode == 'CAMERA' and offset_axis.z:
                for plane in planes:
                    x, y = compute_camera_size(
                        context, plane.location,
                        self.fill_mode, plane.dimensions.x / plane.dimensions.y)
                    plane.dimensions = x, y, 0.0

        # setup new selection
        for plane in planes:
            plane.select_set(True)

        # all done!
        self.report({'INFO'}, "Added {} Image Plane(s)".format(len(planes)))

    # operate on a single image
    def single_image_spec_to_plane(self, context, img_spec):

        # Configure image
        self.apply_image_options(img_spec.image)

        # Configure material
        engine = context.scene.render.engine
        if engine in {'CYCLES', 'BLENDER_EEVEE', 'BLENDER_WORKBENCH'}:
            material = self.create_cycles_material(context, img_spec)

        # Create and position plane object
        plane = self.create_image_plane(context, material.name, img_spec)

        # Assign Material
        plane.data.materials.append(material)

        # If applicable, setup Corner Pin node
        if self.compositing_nodes:
            setup_compositing(context, plane, img_spec)

        return plane

    def apply_image_options(self, image):
        if self.use_transparency == False:
            image.alpha_mode = 'NONE'
        else:
            image.alpha_mode = self.alpha_mode

        if self.relative:
            try:  # can't always find the relative path (between drive letters on windows)
                image.filepath = bpy.path.relpath(image.filepath)
            except ValueError:
                pass

    def apply_texture_options(self, texture, img_spec):
        # Shared by both Cycles and Blender Internal
        image_user = texture.image_user
        image_user.use_auto_refresh = self.use_auto_refresh
        image_user.frame_start = img_spec.frame_start
        image_user.frame_offset = img_spec.frame_offset
        image_user.frame_duration = img_spec.frame_duration

        # Image sequences need auto refresh to display reliably
        if img_spec.image.source == 'SEQUENCE':
            image_user.use_auto_refresh = True

        texture.extension = 'CLIP'  # Default of "Repeat" can cause artifacts

    def apply_material_options(self, material, slot):
        shader = self.shader

        if self.use_transparency:
            material.alpha = 0.0
            material.specular_alpha = 0.0
            slot.use_map_alpha = True
        else:
            material.alpha = 1.0
            material.specular_alpha = 1.0
            slot.use_map_alpha = False

        material.specular_intensity = 0
        material.diffuse_intensity = 1.0
        material.use_transparency = self.use_transparency
        material.transparency_method = 'Z_TRANSPARENCY'
        material.use_shadeless = (shader == 'SHADELESS')
        material.use_transparent_shadows = (shader == 'DIFFUSE')
        material.emit = self.emit_strength if shader == 'EMISSION' else 0.0

    # -------------------------------------------------------------------------
    # Cycles/Eevee
    def create_cycles_texnode(self, context, node_tree, img_spec):
        tex_image = node_tree.nodes.new('ShaderNodeTexImage')
        tex_image.image = img_spec.image
        tex_image.show_texture = True
        tex_image.interpolation = self.interpolation
        tex_image.extension = self.extension
        self.apply_texture_options(tex_image, img_spec)
        return tex_image

    def create_cycles_material(self, context, img_spec):
        image = img_spec.image
        name_compat = bpy.path.display_name_from_filepath(image.filepath)
        material = None
        if self.overwrite_material:
            for mat in bpy.data.materials:
                if mat.name == name_compat:
                    material = mat
        if not material:
            material = bpy.data.materials.new(name=name_compat)

        material.use_nodes = True

        material.blend_method = self.blend_method
        material.shadow_method = self.shadow_method

        material.use_backface_culling = self.use_backface_culling
        material.show_transparent_back = self.show_transparent_back

        node_tree = material.node_tree
        out_node = clean_node_tree(node_tree)

        tex_image = self.create_cycles_texnode(context, node_tree, img_spec)

        if self.shader == 'PRINCIPLED':
            core_shader = node_tree.nodes.new('ShaderNodeBsdfPrincipled')
        elif self.shader == 'SHADELESS':
            core_shader = get_shadeless_node(node_tree)
        elif self.shader == 'EMISSION':
            core_shader = node_tree.nodes.new('ShaderNodeBsdfPrincipled')
            core_shader.inputs['Emission Strength'].default_value = self.emit_strength
            core_shader.inputs['Base Color'].default_value = (0.0, 0.0, 0.0, 1.0)
            core_shader.inputs['Specular'].default_value = 0.0

        # Connect color from texture
        if self.shader in {'PRINCIPLED', 'SHADELESS'}:
            node_tree.links.new(core_shader.inputs[0], tex_image.outputs['Color'])
        elif self.shader == 'EMISSION':
            node_tree.links.new(core_shader.inputs['Emission'], tex_image.outputs['Color'])

        if self.use_transparency:
            if self.shader in {'PRINCIPLED', 'EMISSION'}:
                node_tree.links.new(core_shader.inputs['Alpha'], tex_image.outputs['Alpha'])
            else:
                bsdf_transparent = node_tree.nodes.new('ShaderNodeBsdfTransparent')

                mix_shader = node_tree.nodes.new('ShaderNodeMixShader')
                node_tree.links.new(mix_shader.inputs['Fac'], tex_image.outputs['Alpha'])
                node_tree.links.new(mix_shader.inputs[1], bsdf_transparent.outputs['BSDF'])
                node_tree.links.new(mix_shader.inputs[2], core_shader.outputs[0])
                core_shader = mix_shader

        node_tree.links.new(out_node.inputs['Surface'], core_shader.outputs[0])

        auto_align_nodes(node_tree)
        return material

    # -------------------------------------------------------------------------
    # Geometry Creation
    def create_image_plane(self, context, name, img_spec):

        width, height = self.compute_plane_size(context, img_spec)

        # Create new mesh
        bpy.ops.mesh.primitive_plane_add('INVOKE_REGION_WIN')
        plane = context.active_object
        # Why does mesh.primitive_plane_add leave the object in edit mode???
        if plane.mode != 'OBJECT':
            bpy.ops.object.mode_set(mode='OBJECT')
        plane.dimensions = width, height, 0.0
        plane.data.name = plane.name = name
        bpy.ops.object.transform_apply(location=False, rotation=False, scale=True)

        # If sizing for camera, also insert into the camera's field of view
        if self.size_mode == 'CAMERA':
            offset_axis = self.axis_id_to_vector[self.offset_axis]
            translate_axis = [0 if offset_axis[i] else 1 for i in (0, 1)]
            center_in_camera(context.scene, context.scene.camera, plane, translate_axis)

        self.align_plane(context, plane)

        return plane

    def compute_plane_size(self, context, img_spec):
        """Given the image size in pixels and location, determine size of plane"""
        px, py = img_spec.size

        # can't load data
        if px == 0 or py == 0:
            px = py = 1

        if self.size_mode == 'ABSOLUTE':
            y = self.height
            x = px / py * y

        elif self.size_mode == 'CAMERA':
            x, y = compute_camera_size(
                context, context.scene.cursor.location,
                self.fill_mode, px / py
            )

        elif self.size_mode == 'DPI':
            fact = 1 / self.factor / context.scene.unit_settings.scale_length * 0.0254
            x = px * fact
            y = py * fact

        else:  # elif self.size_mode == 'DPBU'
            fact = 1 / self.factor
            x = px * fact
            y = py * fact

        return x, y

    def align_plane(self, context, plane):
        """Pick an axis and align the plane to it"""
        if 'CAM' in self.align_axis:
            # Camera-aligned
            camera = context.scene.camera
            if (camera):
                # Find the axis that best corresponds to the camera's view direction
                axis = camera.matrix_world @ \
                    Vector((0, 0, 1)) - camera.matrix_world.col[3].xyz
                # pick the axis with the greatest magnitude
                mag = max(map(abs, axis))
                # And use that axis & direction
                axis = Vector([
                    n / mag if abs(n) == mag else 0.0
                    for n in axis
                ])
            else:
                # No camera? Just face Z axis
                axis = Vector((0, 0, 1))
                self.align_axis = 'Z+'
        else:
            # Axis-aligned
            axis = self.axis_id_to_vector[self.align_axis]

        # rotate accordingly for x/y axiis
        if not axis.z:
            plane.rotation_euler.x = pi / 2

            if axis.y > 0:
                plane.rotation_euler.z = pi
            elif axis.y < 0:
                plane.rotation_euler.z = 0
            elif axis.x > 0:
                plane.rotation_euler.z = pi / 2
            elif axis.x < 0:
                plane.rotation_euler.z = -pi / 2

        # or flip 180 degrees for negative z
        elif axis.z < 0:
            plane.rotation_euler.y = pi

        if self.align_axis == 'CAM':
            constraint = plane.constraints.new('COPY_ROTATION')
            constraint.target = camera
            constraint.use_x = constraint.use_y = constraint.use_z = True
            if not self.align_track:
                bpy.ops.object.visual_transform_apply()
                plane.constraints.clear()

        if self.align_axis == 'CAM_AX' and self.align_track:
            constraint = plane.constraints.new('LOCKED_TRACK')
            constraint.target = camera
            constraint.track_axis = 'TRACK_Z'
            constraint.lock_axis = 'LOCK_Y'


# -----------------------------------------------------------------------------
# Register

def import_images_button(self, context):
    self.layout.operator(IMPORT_IMAGE_OT_to_plane.bl_idname, text="Images as Planes", icon='TEXTURE')


classes = (
    IMPORT_IMAGE_OT_to_plane,
)


def register():
    for cls in classes:
        bpy.utils.register_class(cls)

    bpy.types.TOPBAR_MT_file_import.append(import_images_button)
    bpy.types.VIEW3D_MT_image_add.append(import_images_button)

    bpy.app.handlers.load_post.append(register_driver)
    register_driver()


def unregister():
    bpy.types.TOPBAR_MT_file_import.remove(import_images_button)
    bpy.types.VIEW3D_MT_image_add.remove(import_images_button)

    # This will only exist if drivers are active
    if check_drivers in bpy.app.handlers.depsgraph_update_post:
        bpy.app.handlers.depsgraph_update_post.remove(check_drivers)

    bpy.app.handlers.load_post.remove(register_driver)
    del bpy.app.driver_namespace['import_image__find_plane_corner']

    for cls in classes:
        bpy.utils.unregister_class(cls)


if __name__ == "__main__":
    # Run simple doc tests
    import doctest
    doctest.testmod()

    unregister()
    register()