refactor and enable jsonl output.

test/wikiq_test_utils.py

@@ -42,8 +42,20 @@ class WikiqTester:
             else:
                 shutil.rmtree(self.output)
 
-        if out_format == "parquet":
-            os.makedirs(self.output, exist_ok=True)
+        # Also clean up resume-related files
+        for suffix in [".resume_temp", ".checkpoint", ".merged"]:
+            temp_path = self.output + suffix
+            if os.path.exists(temp_path):
+                if os.path.isfile(temp_path):
+                    os.remove(temp_path)
+                else:
+                    shutil.rmtree(temp_path)
+
+        # For JSONL and Parquet, self.output is a file path. Create parent directory if needed.
+        if out_format in ("jsonl", "parquet"):
+            parent_dir = os.path.dirname(self.output)
+            if parent_dir:
+                os.makedirs(parent_dir, exist_ok=True)
 
         if suffix is None:
             self.wikiq_baseline_name = "{0}.{1}".format(wiki, baseline_format)