Properly raise FileNotFound even if the dataset is private (#4536)

* don't use token when not specified, and properly raise filenotfound

* re-enable the tests on windows

* revert unwanted change

* no double backticks

Author: lhoestq

src/datasets/arrow_dataset.py

@@ -4073,7 +4073,7 @@ def _push_parquet_shards_to_hub(
 
         if token is None:
             raise OSError(
-                "You need to provide a `token` or be logged in to Hugging Face with " "`huggingface-cli login`."
+                "You need to provide a `token` or be logged in to Hugging Face with `huggingface-cli login`."
             )
 
         if split is None:

src/datasets/load.py

@@ -783,7 +783,7 @@ def get_module(self) -> DatasetModule:
         hfh_dataset_info = HfApi(config.HF_ENDPOINT).dataset_info(
             self.name,
             revision=self.revision,
-            token=token,
+            token=token if token else "no-token",
             timeout=100.0,
         )
         patterns = (
@@ -1140,7 +1140,7 @@ def dataset_module_factory(
                     dataset_info = hf_api.dataset_info(
                         repo_id=path,
                         revision=revision,
-                        token=token,
+                        token=token if token else "no-token",
                         timeout=100.0,
                     )
                 except Exception as e:  # noqa: catch any exception of hf_hub and consider that the dataset doesn't exist
@@ -1159,7 +1159,10 @@ def dataset_module_factory(
                     elif "401" in str(e):
                         msg = f"Dataset '{path}' doesn't exist on the Hub"
                         msg = msg + f" at revision '{revision}'" if revision else msg
-                        raise FileNotFoundError(msg + ". If the repo is private, make sure you are authenticated.")
+                        raise FileNotFoundError(
+                            msg
+                            + ". If the repo is private, make sure you are authenticated with `use_auth_token=True` after logging in with `huggingface-cli login`."
+                        )
                     else:
                         raise e
                 if filename in [sibling.rfilename for sibling in dataset_info.siblings]:

src/datasets/utils/file_utils.py

@@ -509,7 +509,7 @@ def get_from_cache(
                 logger.info(f"Couldn't get ETag version for url {url}")
             elif response.status_code == 401 and config.HF_ENDPOINT in url and use_auth_token is None:
                 raise ConnectionError(
-                    f"Unauthorized for URL {url}. Please use the parameter ``use_auth_token=True`` after logging in with ``huggingface-cli login``"
+                    f"Unauthorized for URL {url}. Please use the parameter `use_auth_token=True` after logging in with `huggingface-cli login`"
                 )
         except (OSError, requests.exceptions.Timeout) as e:
             # not connected

tests/test_load.py

@@ -781,19 +781,13 @@ def assert_auth(url, *args, headers, **kwargs):
         mock_head.assert_called()
 
 
-@pytest.mark.skipif(
-    os.name == "nt", reason="skip on windows because of SSL issues with moon-staging.huggingface.co:443"
-)
 def test_load_streaming_private_dataset(hf_token, hf_private_dataset_repo_txt_data):
     with pytest.raises(FileNotFoundError):
         load_dataset(hf_private_dataset_repo_txt_data, streaming=True)
     ds = load_dataset(hf_private_dataset_repo_txt_data, streaming=True, use_auth_token=hf_token)
     assert next(iter(ds)) is not None
 
 
-@pytest.mark.skipif(
-    os.name == "nt", reason="skip on windows because of SSL issues with moon-staging.huggingface.co:443"
-)
 def test_load_streaming_private_dataset_with_zipped_data(hf_token, hf_private_dataset_repo_zipped_txt_data):
     with pytest.raises(FileNotFoundError):
         load_dataset(hf_private_dataset_repo_zipped_txt_data, streaming=True)