Merge pull request #1013 from jeromedockes/better_err_msg_for_files_with_wrong_gz_extension

ENH: Provide clear error message when files with zip extensions don't match file contents

Author: effigies

nibabel/loadsave.py

@@ -20,6 +20,52 @@
 from .deprecated import deprecate_with_version
 
 
+def _signature_matches_extension(filename, sniff):
+    """Check if signature aka magic number matches filename extension.
+
+    Parameters
+    ----------
+    filename : str or os.PathLike
+        Path to the file to check
+
+    sniff : bytes or None
+        First bytes of the file. If not `None` and long enough to contain the
+        signature, avoids having to read the start of the file.
+
+    Returns
+    -------
+    matches : bool
+       - `True` if the filename extension is not recognized (not .gz nor .bz2)
+       - `True` if the magic number was successfully read and corresponds to
+         the format indicated by the extension.
+       - `False` otherwise.
+    error_message : str
+       An error message if opening the file failed or a mismatch is detected;
+       the empty string otherwise.
+
+    """
+    signatures = {
+        ".gz": {"signature": b"\x1f\x8b", "format_name": "gzip"},
+        ".bz2": {"signature": b"BZh", "format_name": "bzip2"}
+    }
+    filename = _stringify_path(filename)
+    *_, ext = splitext_addext(filename)
+    ext = ext.lower()
+    if ext not in signatures:
+        return True, ""
+    expected_signature = signatures[ext]["signature"]
+    if sniff is None or len(sniff) < len(expected_signature):
+        try:
+            with open(filename, "rb") as fh:
+                sniff = fh.read(len(expected_signature))
+        except OSError:
+            return False, f"Could not read file: {filename}"
+    if sniff.startswith(expected_signature):
+        return True, ""
+    format_name = signatures[ext]["format_name"]
+    return False, f"File {filename} is not a {format_name} file"
+
+
 def load(filename, **kwargs):
     r""" Load file given filename, guessing at file type
 
@@ -52,6 +98,10 @@ def load(filename, **kwargs):
             img = image_klass.from_filename(filename, **kwargs)
             return img
 
+    matches, msg = _signature_matches_extension(filename, sniff)
+    if not matches:
+        raise ImageFileError(msg)
+
     raise ImageFileError(f'Cannot work out file type of "{filename}"')
 
 

nibabel/tests/test_loadsave.py

@@ -10,7 +10,7 @@
 from .. import (Spm99AnalyzeImage, Spm2AnalyzeImage,
                 Nifti1Pair, Nifti1Image,
                 Nifti2Pair, Nifti2Image)
-from ..loadsave import load, read_img_data
+from ..loadsave import load, read_img_data, _signature_matches_extension
 from ..filebasedimages import ImageFileError
 from ..tmpdirs import InTemporaryDirectory, TemporaryDirectory
 
@@ -76,6 +76,45 @@ def test_load_empty_image():
     assert str(err.value).startswith('Empty file: ')
 
 
+@pytest.mark.parametrize("extension", [".gz", ".bz2"])
+def test_load_bad_compressed_extension(tmp_path, extension):
+    file_path = tmp_path / f"img.nii{extension}"
+    file_path.write_bytes(b"bad")
+    with pytest.raises(ImageFileError, match=".*is not a .* file"):
+        load(file_path)
+
+
+def test_signature_matches_extension(tmp_path):
+    gz_signature = b"\x1f\x8b"
+    good_file = tmp_path / "good.gz"
+    good_file.write_bytes(gz_signature)
+    bad_file = tmp_path / "bad.gz"
+    bad_file.write_bytes(b"bad")
+    matches, msg = _signature_matches_extension(
+        tmp_path / "uncompressed.nii", None)
+    assert matches
+    assert msg == ""
+    matches, msg = _signature_matches_extension(tmp_path / "missing.gz", None)
+    assert not matches
+    assert msg.startswith("Could not read")
+    matches, msg = _signature_matches_extension(bad_file, None)
+    assert not matches
+    assert "is not a" in msg
+    matches, msg = _signature_matches_extension(bad_file, gz_signature + b"abc")
+    assert matches
+    assert msg == ""
+    matches, msg = _signature_matches_extension(
+        good_file, gz_signature + b"abc")
+    assert matches
+    assert msg == ""
+    matches, msg = _signature_matches_extension(good_file, gz_signature[:1])
+    assert matches
+    assert msg == ""
+    matches, msg = _signature_matches_extension(good_file, None)
+    assert matches
+    assert msg == ""
+
+
 def test_read_img_data_nifti():
     shape = (2, 3, 4)
     data = np.random.normal(size=shape)