fix: 🐛 reject blocked datasets in /rows endpoint too

also: move control on blocked datasets from /models to /routes

pyproject.toml

@@ -2,7 +2,7 @@
 authors = ["Sylvain Lesage <[email protected]>"]
 description = "API to extract rows of 🤗 datasets"
 name = "datasets-preview-backend"
-version = "0.16.4"
+version = "0.16.5"
 
 [tool.poetry.dependencies]
 Pillow = "^8.4.0"

src/datasets_preview_backend/models/dataset.py

@@ -3,7 +3,7 @@
 
 from datasets import get_dataset_config_names, get_dataset_split_names
 
-from datasets_preview_backend.constants import DATASETS_BLOCKLIST, FORCE_REDOWNLOAD
+from datasets_preview_backend.constants import FORCE_REDOWNLOAD
 from datasets_preview_backend.exceptions import Status400Error
 
 logger = logging.getLogger(__name__)
@@ -17,8 +17,6 @@ class SplitFullName(TypedDict):
 
 def get_dataset_split_full_names(dataset_name: str, hf_token: Optional[str] = None) -> List[SplitFullName]:
     logger.info(f"get dataset '{dataset_name}' split full names")
-    if dataset_name in DATASETS_BLOCKLIST:
-        raise Status400Error("this dataset is not supported for now.")
     try:
         return [
             {"dataset_name": dataset_name, "config_name": config_name, "split_name": split_name}

src/datasets_preview_backend/routes/rows.py

@@ -4,7 +4,8 @@
 from starlette.responses import Response
 
 from datasets_preview_backend.config import MAX_AGE_LONG_SECONDS
-from datasets_preview_backend.exceptions import StatusError
+from datasets_preview_backend.constants import DATASETS_BLOCKLIST
+from datasets_preview_backend.exceptions import Status400Error, StatusError
 from datasets_preview_backend.io.cache import get_rows_response
 from datasets_preview_backend.routes._utils import get_response
 
@@ -20,6 +21,8 @@ async def rows_endpoint(request: Request) -> Response:
     try:
         if not isinstance(dataset_name, str) or not isinstance(config_name, str) or not isinstance(split_name, str):
             raise StatusError("Parameters 'dataset', 'config' and 'split' are required", 400)
+        if dataset_name in DATASETS_BLOCKLIST:
+            raise Status400Error("this dataset is not supported for now.")
         rows_response, rows_error, status_code = get_rows_response(dataset_name, config_name, split_name)
         return get_response(rows_response or rows_error, status_code, MAX_AGE_LONG_SECONDS)
     except StatusError as err:

src/datasets_preview_backend/routes/splits.py

@@ -4,7 +4,8 @@
 from starlette.responses import Response
 
 from datasets_preview_backend.config import MAX_AGE_LONG_SECONDS
-from datasets_preview_backend.exceptions import StatusError
+from datasets_preview_backend.constants import DATASETS_BLOCKLIST
+from datasets_preview_backend.exceptions import Status400Error, StatusError
 from datasets_preview_backend.io.cache import get_splits_response
 from datasets_preview_backend.routes._utils import get_response
 
@@ -17,7 +18,9 @@ async def splits_endpoint(request: Request) -> Response:
 
     try:
         if not isinstance(dataset_name, str):
-            raise StatusError("Parameter 'dataset' is required", 400)
+            raise Status400Error("Parameter 'dataset' is required")
+        if dataset_name in DATASETS_BLOCKLIST:
+            raise Status400Error("this dataset is not supported for now.")
         splits_response, splits_error, status_code = get_splits_response(dataset_name)
         return get_response(splits_response or splits_error, status_code, MAX_AGE_LONG_SECONDS)
     except StatusError as err: