Fix GConstruct file parsing when input is a list with a single wildca…

…rd path (#1090)

*Issue #, if available:*

* Reported by customer

*Description of changes:*

* Previously if a user passed in a list of input files with a single
wildcard path the path was treated as a regular (non-wildcard) path,
leading to file not found error.
* We make the file parsing consistent by using the `expand_wildcard` and
always passing a list of str to it.
* Add a test case where we used to fail

By submitting this pull request, I confirm that you can use, modify,
copy, and redistribute this contribution, under the terms of your
choice.

python/graphstorm/gconstruct/file_io.py

@@ -21,6 +21,7 @@
 import json
 import os
 import logging
+from typing import Union, List
 
 import pyarrow.parquet as pq
 import pyarrow as pa
@@ -76,9 +77,9 @@ def read_index(split_info):
     return res[0], res[1], res[2]
 
 
-def expand_wildcard(data_files):
+def expand_wildcard(data_files: List[str]) -> List[str]:
     """
-    Expand the wildcard to the actual file lists.
+    Expand a list of paths that can contain wildcards to the actual file lists.
 
     Parameters
     ----------
@@ -490,7 +491,7 @@ def _parse_file_format(conf, is_node, in_mem):
 parse_node_file_format = partial(_parse_file_format, is_node=True)
 parse_edge_file_format = partial(_parse_file_format, is_node=False)
 
-def get_in_files(in_files):
+def get_in_files(in_files: Union[List[str], str]) -> List[str]:
     """ Get the input files.
 
     The input file string may contains a wildcard. This function
@@ -505,16 +506,12 @@ def get_in_files(in_files):
     -------
     a list of str : the full name of input files.
     """
-    # If the input file has a wildcard, get all files that matches the input file name.
-    if '*' in in_files:
-        in_file_list = sorted(glob.glob(in_files))
-        assert len(in_file_list) > 0, \
-            f"There is no file matching {in_files} pattern"
-        in_files = in_file_list
-    # This is a single file.
-    elif not isinstance(in_files, list):
+    # Convert single str to list of str if needed
+    if isinstance(in_files, str):
         in_files = [in_files]
 
+    in_files = expand_wildcard(in_files)
+
     # Verify the existence of the input files.
     for in_file in in_files:
         assert os.path.isfile(in_file), \

tests/unit-tests/gconstruct/test_gconstruct_utils.py

@@ -19,10 +19,11 @@
 
 import dgl
 import numpy as np
-import torch as th
+import pytest
 import pandas as pd
 import pyarrow.parquet as pq
 import pyarrow as pa
+import torch as th
 
 from numpy.testing import assert_almost_equal
 
@@ -292,22 +293,25 @@ def test_get_in_files():
         for i in range(10):
             data = {"test": np.random.rand(10)}
             write_data_parquet(data, files[i])
+        files.sort()
 
+        # Test single string wildcard path
         in_files = get_in_files(os.path.join(tmpdirname,"*.parquet"))
         assert len(in_files) == 10
-        files.sort()
+
+        assert files == in_files
+
+        # Test list of wildcard paths with a single element
+        in_files = get_in_files([os.path.join(tmpdirname, "*.parquet")])
+        assert len(in_files) == 10
         assert files == in_files
 
         in_files = get_in_files(os.path.join(tmpdirname,"test9.parquet"))
         assert len(in_files) == 1
         assert os.path.join(tmpdirname,"test9.parquet") == in_files[0]
 
-        pass_test = False
-        try:
+        with pytest.raises(AssertionError):
             in_files = get_in_files(os.path.join(tmpdirname,"test10.parquet"))
-        except:
-            pass_test = True
-        assert pass_test
 
 def test_get_hard_edge_negs_feats():
     hard_trans0 = HardEdgeDstNegativeTransform("hard_neg", "hard_neg")