#!/usr/bin/env python3
"""Load one public eight-frame simulation observation, without model/GPU usage."""
import argparse,hashlib,io,json,tarfile
from pathlib import Path
import numpy as np
from PIL import Image
from schema import validate

def load(root,row):
 validate(row);root=Path(root)
 with tarfile.open(root/row['media_archive'],'r:') as archive:
  member=archive.getmember(row['media_member'])
  if not member.isfile() or member.name!=row['media_member']:raise ValueError('Unsafe tar member')
  data=archive.extractfile(member).read()
 if hashlib.sha256(data).hexdigest()!=row['media_sha256']:raise ValueError('Media SHA256 mismatch')
 with np.load(io.BytesIO(data),allow_pickle=False) as arrays:
  if set(arrays.files)!={f'f{i}' for i in range(8)}:raise ValueError('Unknown NPZ fields')
  frames=[]
  for i in range(8):
   a=arrays[f'f{i}']
   if a.dtype!=np.uint8 or a.ndim!=1:raise ValueError('Unexpected encoded frame')
   im=Image.open(io.BytesIO(a.tobytes()));im.load()
   if list(reversed(im.size))!=row['hw']:raise ValueError('Frame dimensions differ')
   frames.append(im.convert('RGB'))
 return frames,row

def main():
 p=argparse.ArgumentParser(description=__doc__);p.add_argument('--root',type=Path,default=Path(__file__).resolve().parents[1]);p.add_argument('--split',choices=('train','heldout'),default='train');p.add_argument('--index',type=int,default=0);p.add_argument('--gif',type=Path);a=p.parse_args()
 with (a.root/'data'/(a.split+'.jsonl')).open() as f:
  row=next(json.loads(line) for n,line in enumerate(f) if n==a.index)
 frames,row=load(a.root,row)
 if a.gif:
  if a.gif.exists():raise ValueError('Will not overwrite an existing user file')
  frames[0].save(a.gif,save_all=True,append_images=frames[1:],duration=180,loop=0)
 print(json.dumps({'sample_id':row['sample_id'],'frames':len(frames),'source':row['source_dataset'],'task':row['task'],'image_size':frames[0].size,'model_loaded':False},indent=2))
if __name__=='__main__':main()