Truncate filenames to avoid ones that are too long (#129)

* Truncate data filenames to avoid OSError

* Add test and revert to Path / operator

Author: BenGalewsky

servicex/cache.py

@@ -7,6 +7,9 @@
 
 _ignore_cache = False
 
+# Make sure that generated download path names are below this to avoid os errors
+MAX_PATH_LEN = 235
+
 
 @contextmanager
 def ignore_cache():
@@ -175,6 +178,11 @@ def data_file_location(self, request_id: str, data_name: str) -> Path:
         '''
         Return the path to the file that should be written out for this
         data_name. This is where the output file should get stored.
+        Truncate the leftmost characters from filenames to avoid throwing a
+        OSError: [Errno 63] File name too long error Assume that the most unique part of
+        the name is the right hand side
         '''
-        (self._path / 'data' / request_id).mkdir(parents=True, exist_ok=True)
-        return self._path / 'data' / request_id / sanitize_filename(data_name)
+        parent = self._path / 'data' / request_id
+        parent.mkdir(parents=True, exist_ok=True)
+        sanitized = sanitize_filename(data_name)
+        return parent / sanitized[-1 * (MAX_PATH_LEN - len(parent.name)):]

tests/test_cache.py

@@ -1,3 +1,5 @@
+import string
+import random
 from servicex.utils import ServiceXException
 import pytest
 
@@ -148,6 +150,18 @@ def test_data_file_location(tmp_path):
     assert str(p).startswith(str(tmp_path))
 
 
+def test_data_file_location_long_path(tmp_path):
+    c = Cache(tmp_path)
+    letters = string.ascii_lowercase
+    file_significant_name = 'junk.root'
+    long_file_name = ''.join(random.choice(letters) for i in range(230))
+
+    p = c.data_file_location('123-456', long_file_name+file_significant_name)
+
+    assert(len(p.name) == 235 - len(p.parent.name))
+    assert p.name.endswith(file_significant_name)
+
+
 def test_data_file_location_twice(tmp_path):
     c = Cache(tmp_path)
     _ = c.data_file_location('123-456', 'junk1.root')