bugfix/Fix UnboundLocalError in TiffGlobReader (#449)

* Fix UnboundLocalError in TiffGlobReader

* ensure indexer and fileseries indexes align

* Make TIffGlobReader glob_in types more permissive

* Update aicsimageio/tests/readers/test_glob_reader.py

Co-authored-by: Madison Swain-Bowden <[email protected]>

* get typing correct

* apply black + fix test typing

Co-authored-by: Madison Swain-Bowden <[email protected]>

Author: ianhi

aicsimageio/readers/tiff_glob_reader.py

@@ -5,7 +5,7 @@
 import re
 from collections import OrderedDict
 from pathlib import Path
-from typing import Any, Callable, Dict, List, Optional, Sequence, Tuple, Union
+from typing import Any, Callable, Dict, Iterable, List, Optional, Sequence, Tuple, Union
 
 import dask.array as da
 import numpy as np
@@ -37,8 +37,8 @@ class TiffGlobReader(Reader):
 
     Parameters
     ----------
-    glob_in: Union[PathLike, List[PathLike]]
-        Glob string that identifies all files to be loaded or a list
+    glob_in: Union[PathLike, Iterable[PathLike]]
+        Glob string that identifies all files to be loaded or an iterable
         of paths to the files as returned by glob.
     indexer: Union[Callable, pandas.DataFrame]
         If callable, should consume each filename and return a pd.Series with series
@@ -125,7 +125,7 @@ def _is_supported_image(
 
     def __init__(
         self,
-        glob_in: Union[types.PathLike, List[types.PathLike]],
+        glob_in: Union[types.PathLike, Iterable[types.PathLike]],
         indexer: Union[pd.DataFrame, Callable] = None,
         scene_glob_character: str = "S",
         chunk_dims: Union[str, List[str]] = DEFAULT_CHUNK_DIMS,
@@ -143,10 +143,15 @@ def __init__(
         # Assemble glob list if given a string
         if isinstance(glob_in, str):
             file_series = pd.Series(glob.glob(glob_in))
-        elif isinstance(glob_in, list):
-            file_series = pd.Series(glob_in)
         elif isinstance(glob_in, Path) and "*" in str(glob_in):
             file_series = pd.Series(glob.glob(str(glob_in)))
+        elif isinstance(glob_in, pd.Series):
+            # Ensure all of our indices line up
+            file_series = glob_in.reset_index(drop=True, inplace=False)
+        elif isinstance(glob_in, Iterable):
+            file_series = pd.Series(glob_in)
+        else:
+            raise TypeError(f"Invalid type glob_in - got type {type(glob_in)}")
 
         if len(file_series) == 0:
             raise ValueError("No files found matching glob pattern")
@@ -174,7 +179,10 @@ def indexer(x: str) -> pd.Series:
             self._all_files = file_series.apply(indexer)
             self._all_files["filename"] = file_series
         elif isinstance(indexer, pd.DataFrame):
-            self._all_files = indexer
+            # make a copy of the indexing dataframe and reset it index
+            # to ensure that we don't generate nans when aligning with
+            # file_series.
+            self._all_files = indexer.reset_index(drop=True, inplace=False)
             self._all_files["filename"] = file_series
 
         # If a dim doesn't exist on the file set the column value for that dim to zero.

aicsimageio/tests/readers/test_glob_reader.py

@@ -2,8 +2,11 @@
 import os
 from itertools import product
 from pathlib import Path
+from typing import Any
 
 import numpy as np
+import pandas as pd
+import pytest
 import tifffile as tiff
 import xarray as xr
 
@@ -68,6 +71,49 @@ def test_glob_reader_2d(tmp_path: Path) -> None:
     check_values(gr, reference)
 
 
+def test_index_alignment(tmp_path: Path) -> None:
+    # Testing case where user has passed in a list of files
+    # and a dataframe with non-continuous index
+
+    # use as_mm to have an easily available Indexer function
+    _ = make_fake_data_2d(tmp_path, as_mm=True)
+    filenames = np.array(list((tmp_path / "2d_images").glob("*.tif")))
+    # print(filenames)
+    indexer = pd.Series(filenames).apply(TiffGlobReader.MicroManagerIndexer)
+
+    # Keep only some of the Z
+    # more realistic case is eliminating everything after a given time point
+    # but this garuntees that our eliminated images will be embedded all through the
+    # order rather than just at the end
+    keep = indexer.Z < 5
+
+    indexer = indexer.loc[keep]
+
+    reader = TiffGlobReader(filenames[keep], indexer)
+
+    # check that there are no nans
+    # nans are a symptom of index misalignment
+    assert not reader._all_files.isnull().any().any()
+
+
+@pytest.mark.parametrize(
+    "type_",
+    [
+        list,
+        pd.Series,
+        np.array,
+        # should throw a TypeError instead of an unboundlocal error
+        pytest.param(bytes, marks=pytest.mark.raises(exception=TypeError)),
+    ],
+)
+def test_glob_types(type_: Any, tmp_path: Path) -> None:
+    reference = make_fake_data_2d(tmp_path)
+    filenames = list((tmp_path / "2d_images").glob("*.tif"))
+
+    gr = TiffGlobReader(type_(filenames))
+    check_values(gr, reference)
+
+
 def test_mm_indexer(tmp_path: Path) -> None:
     _ = make_fake_data_2d(tmp_path, True)
     gr = aicsimageio.readers.TiffGlobReader(