Update XML script + add regression test for XML module

autocorpus/IAO_dicts/IAO_FINAL_MAPPING.txt

@@ -252,3 +252,4 @@ version changes	version number section
 visual abstract	graphical abstract section
 web resources	references section
 web site references	references section
+document title document title

autocorpus/IAO_dicts/IAO_term_to_ID.txt

@@ -39,5 +39,6 @@ requirements section	IAO:0000641
 statistical analysis section	IAO:0000644
 tables section	IAO:0000645
 descriptive data section	IAO:0000701
+document title	IAO:0000305
 disclosure section	IAO:CUIless
 highlights section	IAO:CUIless

autocorpus/file_processing.py

@@ -8,6 +8,7 @@
 from .autocorpus import Autocorpus
 from .file_type import FileType, check_file_type
 from .html import process_html_article
+from .parse_xml import convert_xml_to_json
 
 
 def process_file(
@@ -35,10 +36,9 @@
                 file_path, *process_html_article(config, file_path, linked_tables)
             )
         case FileType.XML:
-            raise NotImplementedError(
-                f"Could not process file {file_path}. Process XML files by running:\n\t"
-                f"python -m autocorpus.parse_xml {file_path}"
-            )
+            main_text = convert_xml_to_json(file_path)
+
+            return Autocorpus(file_path, main_text, dict(), dict())
         case FileType.PDF:
             try:
                 from .pdf import extract_pdf_content

pyproject.toml

@@ -77,7 +77,7 @@ disallow_any_generics = true
 warn_unreachable = true
 warn_unused_ignores = true
 # disallow_untyped_defs = true
-exclude = [".venv/", "docs/", "autocorpus/parse_xml.py", "site/"]
+exclude = [".venv/", "docs/", "site/"]
 
 [[tool.mypy.overrides]]
 module = "tests.*"

tests/test_regression.py

@@ -98,6 +98,26 @@ def _run_html_regression_test(
         assert not expected_tables
 
 
+@pytest.mark.parametrize("input_file", [("PMC/xml/PMC8885717.xml")])
+def test_xml(data_path: Path, input_file: str) -> None:
+    """A regression test for the xml autoCORPus function."""
+    from autocorpus.parse_xml import convert_xml_to_json
+
+    pmc_example_path = data_path / input_file
+    with open(
+        str(pmc_example_path).replace(".xml", "_bioc.json"),
+        encoding="utf-8",
+    ) as f:
+        expected_bioc = json.load(f)
+
+    bioc = convert_xml_to_json(pmc_example_path)
+    # Only remove date because XML doesn't have an "inputfile" field.
+    bioc.pop("date")
+    expected_bioc.pop("date")
+
+    assert bioc == expected_bioc
+
+
 @pytest.mark.skip_ci_macos
 @pytest.mark.parametrize(
     "input_file, config",