Do not filter out .zip extensions from no-script datasets (#6208)

* Rename zip_csv_path fixture dirname and filename

* Test load no-script dataset with ZIP file

* Fix style

* Avoid filtering out .zip extension

Author: albertvillanova

src/datasets/packaged_modules/__init__.py

@@ -60,5 +60,5 @@ def _hash_python_lines(lines: List[str]) -> str:
 for _ext, (_module, _) in _EXTENSION_TO_MODULE.items():
     _MODULE_TO_EXTENSIONS.setdefault(_module, []).append(_ext)
 
-_MODULE_TO_EXTENSIONS["imagefolder"].append(".zip")
-_MODULE_TO_EXTENSIONS["audiofolder"].append(".zip")
+for _module in _MODULE_TO_EXTENSIONS:
+    _MODULE_TO_EXTENSIONS[_module].append(".zip")

tests/fixtures/files.py

@@ -289,7 +289,7 @@ def bz2_csv_path(csv_path, tmp_path_factory):
 
 @pytest.fixture(scope="session")
 def zip_csv_path(csv_path, csv2_path, tmp_path_factory):
-    path = tmp_path_factory.mktemp("data") / "dataset.csv.zip"
+    path = tmp_path_factory.mktemp("zip_csv_path") / "csv-dataset.zip"
     with zipfile.ZipFile(path, "w") as f:
         f.write(csv_path, arcname=os.path.basename(csv_path))
         f.write(csv2_path, arcname=os.path.basename(csv2_path))

tests/test_load.py

@@ -1458,3 +1458,12 @@ def test_load_dataset_with_storage_options_with_decoding(mockfs, image_file):
     ds = load_dataset("imagefolder", data_files=data_files, storage_options=mockfs.storage_options)
     assert len(ds["train"]) == 1
     assert isinstance(ds["train"][0]["image"], PIL.Image.Image)
+
+
+def test_load_dataset_without_script_with_zip(zip_csv_path):
+    path = str(zip_csv_path.parent)
+    ds = load_dataset(path)
+    assert list(ds.keys()) == ["train"]
+    assert ds["train"].column_names == ["col_1", "col_2", "col_3"]
+    assert ds["train"].num_rows == 8
+    assert ds["train"][0] == {"col_1": 0, "col_2": 0, "col_3": 0.0}