import os
import copy
import sys
import importlib
import argparse
import pandas as pd
from easydict import EasyDict as edict
from functools import partial
import numpy as np
import open3d_pycg as o3d
import utils3d


def _voxelize(file, sha256, output_dir):
    mesh = o3d.io.read_triangle_mesh(os.path.join(output_dir, 'renders', sha256, 'mesh.ply'))
    # clamp vertices to the range [-0.5, 0.5]
    vertices = np.clip(np.asarray(mesh.vertices), -0.5 + 1e-6, 0.5 - 1e-6)
    mesh.vertices = o3d.utility.Vector3dVector(vertices)
    voxel_grid = o3d.geometry.VoxelGrid.create_from_triangle_mesh_within_bounds(mesh, voxel_size=1/64, min_bound=(-0.5, -0.5, -0.5), max_bound=(0.5, 0.5, 0.5))
    vertices = np.array([voxel.grid_index for voxel in voxel_grid.get_voxels()])
    assert np.all(vertices >= 0) and np.all(vertices < 64), "Some vertices are out of bounds"
    vertices = (vertices + 0.5) / 64 - 0.5
    utils3d.io.write_ply(os.path.join(output_dir, 'voxels', f'{sha256}.ply'), vertices)
    return {'sha256': sha256, 'voxelized': True, 'num_voxels': len(vertices)}


if __name__ == '__main__':
    dataset_utils = importlib.import_module(f'datasets.{sys.argv[1]}')

    parser = argparse.ArgumentParser()
    parser.add_argument('--output_dir', type=str, required=True,
                        help='Directory to save the metadata')
    parser.add_argument('--filter_low_aesthetic_score', type=float, default=None,
                        help='Filter objects with aesthetic score lower than this value')
    parser.add_argument('--instances', type=str, default=None,
                        help='Instances to process')
    parser.add_argument('--num_views', type=int, default=150,
                        help='Number of views to render')
    dataset_utils.add_args(parser)
    parser.add_argument('--rank', type=int, default=0)
    parser.add_argument('--world_size', type=int, default=1)
    parser.add_argument('--max_workers', type=int, default=None)
    opt = parser.parse_args(sys.argv[2:])
    opt = edict(vars(opt))

    os.makedirs(os.path.join(opt.output_dir, 'voxels'), exist_ok=True)

    # get file list
    if not os.path.exists(os.path.join(opt.output_dir, 'metadata.csv')):
        raise ValueError('metadata.csv not found')
    metadata = pd.read_csv(os.path.join(opt.output_dir, 'metadata.csv'))
    if opt.instances is None:
        if opt.filter_low_aesthetic_score is not None:
            metadata = metadata[metadata['aesthetic_score'] >= opt.filter_low_aesthetic_score]
        if 'rendered' not in metadata.columns:
            raise ValueError('metadata.csv does not have "rendered" column, please run "build_metadata.py" first')
        metadata = metadata[metadata['rendered'] == True]
        if 'voxelized' in metadata.columns:
            metadata = metadata[metadata['voxelized'] == False]
    else:
        if os.path.exists(opt.instances):
            with open(opt.instances, 'r') as f:
                instances = f.read().splitlines()
        else:
            instances = opt.instances.split(',')
        metadata = metadata[metadata['sha256'].isin(instances)]

    start = len(metadata) * opt.rank // opt.world_size
    end = len(metadata) * (opt.rank + 1) // opt.world_size
    metadata = metadata[start:end]
    records = []

    # filter out objects that are already processed
    for sha256 in copy.copy(metadata['sha256'].values):
        if os.path.exists(os.path.join(opt.output_dir, 'voxels', f'{sha256}.ply')):
            pts = utils3d.io.read_ply(os.path.join(opt.output_dir, 'voxels', f'{sha256}.ply'))[0]
            records.append({'sha256': sha256, 'voxelized': True, 'num_voxels': len(pts)})
            metadata = metadata[metadata['sha256'] != sha256]
                
    print(f'Processing {len(metadata)} objects...')

    # process objects
    func = partial(_voxelize, output_dir=opt.output_dir)
    voxelized = dataset_utils.foreach_instance(metadata, opt.output_dir, func, max_workers=opt.max_workers, desc='Voxelizing')
    voxelized = pd.concat([voxelized, pd.DataFrame.from_records(records)])
    voxelized.to_csv(os.path.join(opt.output_dir, f'voxelized_{opt.rank}.csv'), index=False)