Several datasets not loading, with PermissionError when running PyTest, caused by `with_open` on folder?

cleong110 commented 4 months ago

SignBank and Sign2Mint are having loading issues in https://github.com/sign-language-processing/datasets/blob/master/examples/load.ipynb, perhaps this is why.

Running PyTest as noted in #53, I find that some datasets have PermissionErrors like the one below. https://stackoverflow.com/questions/36434764/permissionerror-errno-13-permission-denied suggests this is caused suggests that this might be caused by attempting to use with open on a folder instead of a file.

Example:

self = <datasets.sign_language_datasets.datasets.signbank.signbank.SignBank object at 0x000001A8C5C903D0>
dl_manager = <tensorflow_datasets.core.download.download_manager.DownloadManager object at 0x000001A8C5C93790>

    def _split_generators(self, dl_manager: tfds.download.DownloadManager):
        """Returns SplitGenerators."""
        dataset_warning(self)

        index = dl_manager.download("http://signbank.org/signpuddle2.0/data/spml/")
        regex = r"\"sgn[\d]+.spml\""
>       with open(index, "r", encoding="utf-8") as f:
E       PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\signbank\\dummy_data'

sign_language_datasets\datasets\signbank\signbank.py:218: PermissionError

Affected Datasets

Datasets that may be affected due to this:

Sign2MINT

```python self = dl_manager = def _split_generators(self, dl_manager: tfds.download.DownloadManager): dataset_warning(self) annotations_path = dl_manager.download("https://sign2mint.de/api/entries/all/") local_videos = {} if self._builder_config.include_video and self._builder_config.process_video: > with open(annotations_path, "r", encoding="utf-8") as f: E PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\sign2mint\\dummy_data' sign_language_datasets\datasets\sign2mint\sign2mint.py:88: PermissionError ``` ![image](https://github.com/sign-language-processing/datasets/assets/122366389/99f58b4c-0f22-49d2-98d9-0f2d4e2ccb62)

SignBank

```python self = dl_manager = def _split_generators(self, dl_manager: tfds.download.DownloadManager): """Returns SplitGenerators.""" dataset_warning(self) index = dl_manager.download("http://signbank.org/signpuddle2.0/data/spml/") regex = r"\"sgn[\d]+.spml\"" > with open(index, "r", encoding="utf-8") as f: E PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\signbank\\dummy_data' sign_language_datasets\datasets\signbank\signbank.py:218: PermissionError ``` ![image](https://github.com/sign-language-processing/datasets/assets/122366389/3f65a9f9-d461-4e98-887a-d6604f1a1bdb) ![image](https://github.com/sign-language-processing/datasets/assets/122366389/bfd865eb-80c6-42eb-ae74-b069e0b56e57)

ASL Lex

``` self = csv_path = WindowsGPath('C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\asl_lex\\dummy_data') def _generate_examples(self, csv_path: str): """Yields examples.""" > with open(csv_path, "r", encoding="ISO-8859-1") as csvfile: E PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\asl_lex\\dummy_data' sign_language_datasets\datasets\asl_lex\asl_lex.py:72: PermissionError ```

DGS Corpus

``` filepath_in = 'C:\\Users\\Colin\\AppData\\Local\\Temp\\tmp6sc7ayxv', filepath_out = 'C:\\Users\\Colin\\AppData\\Local\\Temp\\tmp6sc7ayxv.gz' def _gzip_file(filepath_in: str, filepath_out: str) -> None: > with open(filepath_in, "rb") as filehandle_in: E PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\AppData\\Local\\Temp\\tmp6sc7ayxv' sign_language_datasets\datasets\dgs_corpus\dgs_corpus_test.py:75: PermissionError ```

Dicta-Sign

``` self = dl_manager = def _split_generators(self, dl_manager: tfds.download.DownloadManager): dataset_warning(self) concepts_path = dl_manager.download("https://www.sign-lang.uni-hamburg.de/dicta-sign/portal/concepts/concepts_eng.html") if self._builder_config.include_pose is not None: poses_path = dl_manager.download_and_extract(_POSE_URLS[self._builder_config.include_pose]) else: poses_path = None regex = r" with open(concepts_path, "r", encoding="utf-8") as concepts_f: E PermissionError: [Errno 13] Permission denied: 'C:\\Users\\Colin\\projects\\sign-language\\colin_pull_requesting\\datasets\\sign_language_datasets\\datasets\\dicta_sign\\dummy_data' sign_language_datasets\datasets\dicta_sign\dicta_sign.py:92: PermissionError ```

I tried these in The Colab Loading Script, and SignBank and Sign2Mint actually are crashing.

For SignBank the issue might be at https://github.com/sign-language-processing/datasets/blob/3aa515c0da9f3c5f43db5a8cc407a7abbe083db0/sign_language_datasets/datasets/signbank/signbank.py#L216

For Sign2Mint the issue might be at https://github.com/sign-language-processing/datasets/blob/3aa515c0da9f3c5f43db5a8cc407a7abbe083db0/sign_language_datasets/datasets/sign2mint/sign2mint.py#L84

cleong110 commented 4 months ago

Below is the FULL PYTEST OUTPUT, very long!

Full PyTest output - Click to expand

AmitMY commented 4 months ago

Just dropping notes here, seems like:

we need to use tf.io.gfile in more places
for signbank, we are opening a file, which is fine, but for sign2mint, the index URL we use (https://sign2mint.de/api/entries/all/) is no longer valid
Most importantly, DL_EXTRACT_RESULT is missing in many test files, which should point to dummy_data/some file

sign-language-processing / datasets

Several datasets not loading, with PermissionError when running PyTest, caused by `with_open` on folder? #55

Affected Datasets