hdf5_data/hdf5_merger.py

import argparse
import json
import os
import time

import cv2
import h5py
import numpy as np
import tqdm

def sort_key_hdf5(name):
    # extract number after 'demo_'
    number = int(name.split('_')[-1].split('.')[0])
    return number

def sort_key_metainfo(name):
    # extract number after 'demo_'
    number = int(name.split('_')[-2].split('.')[0])
    return number

def recursive_merge(dest, src):
    for key, value in src.items():
        if key in dest and isinstance(dest[key], dict) and isinstance(value, dict):
            recursive_merge(dest[key], value)
        else:
            dest[key] = value

def recursive_copy(src, dest):
    for key in src.keys():
        if isinstance(src[key], h5py.Group):
            new_grp = dest.create_group(key)
            recursive_copy(src[key], new_grp)
        elif isinstance(src[key], h5py.Dataset):
            src.copy(key, dest)
    for attr_key in src.attrs:
        dest.attrs[attr_key] = src.attrs[attr_key]

def main(args):

    # Prepare JSON file to record success/false and initial states per episode
    metainfo_json_dict = {}
    metainfo_json_out_path = os.path.join(args.out_dir, f"./metainfo.json")
    with open(metainfo_json_out_path, "w") as f:
        # Just test that we can write to this file (we overwrite it later)
        json.dump(metainfo_json_dict, f)

    # Get task suite
    task_suite = ['task1', 'task2', 'task3']
    num_tasks_in_suite = 3
    
    # Iterate through the task suites
    for task_id in tqdm.tqdm(range(num_tasks_in_suite)):
        # Get task in suite
        task = task_suite[task_id]
        data_dir = os.path.join('./', task)
        data_files = os.listdir(data_dir)
        
        hdf5_files = [_file for _file in data_files if '.hdf5' in _file]
        hdf5_files = sorted(hdf5_files, key=sort_key_hdf5)
        meta_files = [_file for _file in data_files if '_metainfo.json' in _file]
        meta_files = sorted(meta_files, key=sort_key_metainfo)

        # Create new HDF5 file for regenerated demos
        new_data_path = os.path.join(args.out_dir, f"{task}_demo.hdf5")
        new_data_file = h5py.File(new_data_path, "w")
        grp = new_data_file.create_group("data")

        for idx, hdf5_name in tqdm.tqdm(enumerate(hdf5_files)):
            hdf5_name = os.path.join(data_dir, hdf5_name)
            traj_data_file = h5py.File(hdf5_name, "r")
            traj_data = traj_data_file["data"]

            # Copy trajectory data
            for ep_key in traj_data.keys():
                src_grp = traj_data[ep_key]
                dest_grp = grp.create_group(ep_key)
                recursive_copy(src_grp, dest_grp)
            
            traj_data_file.close()

            meta_name = os.path.join(data_dir, meta_files[idx])
            with open(meta_name, "r") as f:
                # Just test that we can write to this file (we overwrite it later)
                meta_data = json.load(f)
                meta_data_key = list(meta_data.keys())[0]
                demo_data_key = list(meta_data[meta_data_key].keys())[0]
                indexed_meta_data = meta_data[meta_data_key][demo_data_key]

                # Recursively merge the meta data
                recursive_merge(metainfo_json_dict, meta_data)

                # Write metainfo dict to JSON file
                # (We repeatedly overwrite, rather than doing this once at the end, just in case the script crashes midway)
                with open(metainfo_json_out_path, "w") as f:
                    json.dump(metainfo_json_dict, f, indent=2)

            # if idx > 1:
            #     break
        
        new_data_file.close()

if __name__ == '__main__':
    # Parse command-line arguments
    parser = argparse.ArgumentParser()
    parser.add_argument("--in_dir", default='./')
    parser.add_argument("--out_dir", default='./')
    args = parser.parse_args()

    main(args)