fix jsonl.d output.

test/test_resume.py

@@ -668,6 +668,88 @@ def test_resume_page_boundary():
     assert_frame_equal(df_full, df_resumed)
 
 
+def test_jsonl_dir_output():
+    """Test that .jsonl.d output creates files named after input files.
+
+    When output is a .jsonl.d directory, each input file should write to
+    a separate JSONL file named after the input (e.g., sailormoon.jsonl),
+    not a generic data.jsonl.
+    """
+    import pandas as pd
+    from pandas.testing import assert_frame_equal
+
+    output_dir = os.path.join(TEST_OUTPUT_DIR, "jsonl_dir_test.jsonl.d")
+    input_file = os.path.join(TEST_DIR, "dumps", f"{SAILORMOON}.xml.7z")
+
+    if os.path.exists(output_dir):
+        shutil.rmtree(output_dir)
+
+    # Run wikiq with .jsonl.d output
+    cmd = f"{WIKIQ} {input_file} -o {output_dir} --fandom-2020 --batch-size 10"
+    try:
+        subprocess.check_output(cmd, stderr=subprocess.PIPE, shell=True)
+    except subprocess.CalledProcessError as exc:
+        pytest.fail(exc.stderr.decode("utf8"))
+
+    # Verify output file is named after input, not "data.jsonl"
+    expected_output = os.path.join(output_dir, f"{SAILORMOON}.jsonl")
+    wrong_output = os.path.join(output_dir, "data.jsonl")
+
+    assert os.path.exists(expected_output), \
+        f"Expected {expected_output} to exist, but it doesn't. Directory contents: {os.listdir(output_dir)}"
+    assert not os.path.exists(wrong_output), \
+        f"Expected {wrong_output} NOT to exist (should be named after input file)"
+
+    # Verify output has data
+    rows = read_jsonl(expected_output)
+    assert len(rows) > 0, "Output file should have data"
+
+
+def test_jsonl_dir_resume():
+    """Test that resume works correctly with .jsonl.d directory output.
+
+    The resume logic must derive the same filename from the input file
+    as the write logic does.
+    """
+    import pandas as pd
+    from pandas.testing import assert_frame_equal
+
+    output_dir = os.path.join(TEST_OUTPUT_DIR, "jsonl_dir_resume.jsonl.d")
+    input_file = os.path.join(TEST_DIR, "dumps", f"{SAILORMOON}.xml.7z")
+
+    if os.path.exists(output_dir):
+        shutil.rmtree(output_dir)
+
+    # First run: complete
+    cmd_full = f"{WIKIQ} {input_file} -o {output_dir} --fandom-2020 --batch-size 10"
+    try:
+        subprocess.check_output(cmd_full, stderr=subprocess.PIPE, shell=True)
+    except subprocess.CalledProcessError as exc:
+        pytest.fail(exc.stderr.decode("utf8"))
+
+    expected_output = os.path.join(output_dir, f"{SAILORMOON}.jsonl")
+    full_rows = read_jsonl(expected_output)
+
+    # Truncate to partial
+    partial_idx = len(full_rows) // 2
+    with open(expected_output, 'w') as f:
+        for row in full_rows[:partial_idx]:
+            f.write(json.dumps(row) + "\n")
+
+    # Resume
+    cmd_resume = f"{WIKIQ} {input_file} -o {output_dir} --fandom-2020 --batch-size 10 --resume"
+    try:
+        subprocess.check_output(cmd_resume, stderr=subprocess.PIPE, shell=True)
+    except subprocess.CalledProcessError as exc:
+        pytest.fail(exc.stderr.decode("utf8"))
+
+    resumed_rows = read_jsonl(expected_output)
+
+    df_full = pd.DataFrame(full_rows)
+    df_resumed = pd.DataFrame(resumed_rows)
+    assert_frame_equal(df_full, df_resumed)
+
+
 def test_resume_revert_detection():
     """Test that revert detection works correctly after resume.