Generate a metadata dictionary using flattened metadata keys. Args: dataset: dictionary containing the dataset keys and values. Keys are flatened. prefix: common prefix of the metadata fields. Returns: Nested dictionary with the episode metadata. If the dataset contains:
(
dataset: Dict[str, Any], prefix: str
)
| 103 | |
| 104 | |
| 105 | def _get_nested_metadata( |
| 106 | dataset: Dict[str, Any], prefix: str |
| 107 | ) -> Dict[str, Any]: |
| 108 | """Generate a metadata dictionary using flattened metadata keys. |
| 109 | |
| 110 | Args: |
| 111 | dataset: dictionary containing the dataset keys and values. Keys are |
| 112 | flatened. |
| 113 | prefix: common prefix of the metadata fields. |
| 114 | |
| 115 | Returns: |
| 116 | Nested dictionary with the episode metadata. |
| 117 | |
| 118 | If the dataset contains: |
| 119 | { |
| 120 | 'metadata/v1/v2': 1, |
| 121 | 'metadata/v3': 2, |
| 122 | } |
| 123 | and prefix='metadata', it returns: |
| 124 | { |
| 125 | 'v1':{ |
| 126 | 'v2': 1, |
| 127 | } |
| 128 | 'v3': 2, |
| 129 | } |
| 130 | It assumes that the flattened metadata keys are well-formed. |
| 131 | """ |
| 132 | episode_metadata = {} |
| 133 | for k in dataset.keys(): |
| 134 | if f'{prefix}/' not in k: |
| 135 | continue |
| 136 | keys = k.split('/')[1:] |
| 137 | nested_dict = episode_metadata |
| 138 | leaf_value = dataset[k] |
| 139 | for index, nested_key in enumerate(keys): |
| 140 | if index == (len(keys) - 1): |
| 141 | nested_dict[nested_key] = leaf_value |
| 142 | else: |
| 143 | if nested_key not in nested_dict: |
| 144 | nested_dict[nested_key] = {} |
| 145 | nested_dict = nested_dict[nested_key] |
| 146 | |
| 147 | return episode_metadata |
| 148 | |
| 149 | |
| 150 | def _get_episode( |
no test coverage detected