Set back feature to create a repo when using clone_from (#1187)

* Set back feature to create a repo when using clone_from

* fix repo_id vs repo_name

* expect deprecation ++ run dataset repository tests separately

Author: Wauplin

.github/workflows/python-tests.yml

@@ -75,11 +75,11 @@ jobs:
           case "${{ matrix.test_name }}" in
 
             "Repository only")
-              eval "$PYTEST ../tests -k 'RepositoryTest'"
+              eval "$PYTEST ../tests -k 'RepositoryTest or RepositoryDatasetTest'"
               ;;
 
             "Everything else")
-              eval "$PYTEST ../tests -k 'not RepositoryTest'"
+              eval "$PYTEST ../tests -k 'not RepositoryTest and not RepositoryDatasetTest'"
               ;;
 
             lfs)

src/huggingface_hub/repository.py

@@ -5,6 +5,7 @@
 import tempfile
 import threading
 import time
+import warnings
 from contextlib import contextmanager
 from pathlib import Path
 from typing import Callable, Dict, Iterator, List, Optional, Tuple, Union
@@ -15,7 +16,14 @@
 
 from .hf_api import HfApi, repo_type_and_id_from_hf_id
 from .lfs import LFS_MULTIPART_UPLOAD_COMMAND
-from .utils import HfFolder, logging, run_subprocess, tqdm, validate_hf_hub_args
+from .utils import (
+    HfFolder,
+    RepositoryNotFoundError,
+    logging,
+    run_subprocess,
+    tqdm,
+    validate_hf_hub_args,
+)
 from .utils._deprecation import _deprecate_arguments, _deprecate_method
 from .utils._typing import TypedDict
 
@@ -684,9 +692,10 @@ def clone_from(self, repo_url: str, token: Union[bool, str, None] = None):
         if hub_url in repo_url or (
             "http" not in repo_url and len(repo_url.split("/")) <= 2
         ):
-            repo_type, namespace, repo_id = repo_type_and_id_from_hf_id(
+            repo_type, namespace, repo_name = repo_type_and_id_from_hf_id(
                 repo_url, hub_url=hub_url
             )
+            repo_id = f"{namespace}/{repo_name}" if namespace is not None else repo_name
 
             if repo_type is not None:
                 self._repo_type = repo_type
@@ -701,10 +710,32 @@ def clone_from(self, repo_url: str, token: Union[bool, str, None] = None):
                 scheme = urlparse(repo_url).scheme
                 repo_url = repo_url.replace(f"{scheme}://", f"{scheme}://user:{token}@")
 
-            if namespace is not None:
-                repo_url += f"{namespace}/"
             repo_url += repo_id
 
+            # To be removed: check if repo exists. If not, create it first.
+            try:
+                HfApi().repo_info(f"{repo_id}", repo_type=self._repo_type, token=token)
+            except RepositoryNotFoundError:
+                if self._repo_type == "space":
+                    raise ValueError(
+                        "Creating a Space through passing Space link to clone_from is"
+                        " not allowed. Make sure the Space exists on Hugging Face Hub."
+                    )
+                else:
+                    warnings.warn(
+                        "Creating a repository through 'clone_from' is deprecated and"
+                        " will be removed in v0.12. Please create the repository first"
+                        " using `create_repo(..., exists_ok=True)`.",
+                        FutureWarning,
+                    )
+                    self.client.create_repo(
+                        repo_id=repo_id,
+                        token=token,
+                        repo_type=self._repo_type,
+                        exist_ok=True,
+                        private=self._private,
+                    )
+
         # For error messages, it's cleaner to show the repo url without the token.
         clean_repo_url = re.sub(r"(https?)://.*@", r"\1://", repo_url)
         try:

tests/test_repository.py

@@ -22,8 +22,6 @@
 import uuid
 from io import BytesIO
 
-import pytest
-
 import requests
 from huggingface_hub._login import _currently_setup_credential_helpers
 from huggingface_hub.hf_api import HfApi
@@ -118,17 +116,16 @@ def test_init_clone_from(self):
             repo_id=f"{USER}/{self.REPO_NAME}-temp", repo_type="space"
         )
 
+    @expect_deprecation("clone_from")
     def test_clone_from_missing_repo(self):
         """
-        If the repo does not exist, it is not created automatically. This behavior
-        was possible before but has been deprecated and removed.
+        If the repo does not exist, it is created automatically.
+
+        This behavior is deprecated and will be removed in v0.12.
         """
-        with pytest.raises(EnvironmentError, match=".*remote: Repository not found.*"):
-            Repository(
-                WORKING_REPO_DIR,
-                clone_from=f"{USER}/{uuid.uuid4()}",
-                use_auth_token=self._token,
-            )
+        Repository(
+            WORKING_REPO_DIR, clone_from=f"{USER}/{uuid.uuid4()}", token=self._token
+        )
 
     def test_clone_from_model(self):
         temp_repo_url = self._api.create_repo(
@@ -484,6 +481,7 @@ def test_clone_with_repo_name_and_user_namespace(self):
         self.assertTrue("model.bin" in files)
 
     @retry_endpoint
+    @expect_deprecation("clone_from")
     def test_clone_with_repo_name_and_no_namespace(self):
         with self.assertRaises(EnvironmentError):
             Repository(
@@ -1745,6 +1743,7 @@ def test_clone_with_repo_name_and_user_namespace(self):
         self.assertTrue("test.py" in files)
 
     @retry_endpoint
+    @expect_deprecation("clone_from")
     def test_clone_with_repo_name_and_no_namespace(self):
         self.assertRaises(
             OSError,