chore(xml_validation): turn into pytest (#815)

src/dsp_tools/cli/call_action.py

@@ -21,7 +21,7 @@
 from dsp_tools.commands.xmlupload.upload_config import DiagnosticsConfig, UploadConfig
 from dsp_tools.commands.xmlupload.xmlupload import xmlupload
 from dsp_tools.utils.create_logger import get_logger
-from dsp_tools.utils.validate_data_xml import validate_xml
+from dsp_tools.utils.xml_validation import validate_xml
 
 logger = get_logger(__name__)
 

src/dsp_tools/commands/excel2xml/excel2xml_lib.py

@@ -19,7 +19,7 @@
 from dsp_tools.utils.date_util import is_full_date
 from dsp_tools.utils.shared import check_notna, simplify_name
 from dsp_tools.utils.uri_util import is_uri
-from dsp_tools.utils.validate_data_xml import validate_xml
+from dsp_tools.utils.xml_validation import validate_xml
 
 # ruff: noqa: E501 (line-too-long)
 

src/dsp_tools/commands/xmlupload/read_validate_xml_file.py

@@ -9,8 +9,8 @@
 from dsp_tools.models.exceptions import UserError
 from dsp_tools.utils.create_logger import get_logger
 from dsp_tools.utils.iri_util import is_resource_iri
-from dsp_tools.utils.validate_data_xml import validate_xml
 from dsp_tools.utils.xml_utils import parse_and_clean_xml_file
+from dsp_tools.utils.xml_validation import validate_xml
 
 logger = get_logger(__name__)
 

test/unittests/utils/test_xml_validation_high_level.py

@@ -0,0 +1,71 @@
+import pytest
+from lxml import etree
+
+from dsp_tools.models.exceptions import InputError
+from dsp_tools.utils.xml_validation import validate_xml
+
+
+def test_validate_xml_data_systematic() -> None:
+    assert validate_xml(input_file="testdata/xml-data/test-data-systematic.xml") is True
+
+
+def test_validate_xml_data_minimal() -> None:
+    assert validate_xml(input_file=etree.parse(source="testdata/xml-data/test-data-minimal.xml")) is True
+
+
+def test_validate_xml_invalid_resource_tag_line_twelve() -> None:
+    with pytest.raises(
+        InputError,
+        match=(
+            r"""The XML file cannot be uploaded due to the following validation error\(s\)\:
+    Line 12\: Element 'resource', attribute 'invalidtag'\: The attribute 'invalidtag' is not allowed\."""
+        ),
+    ):
+        validate_xml(input_file="testdata/invalid-testdata/xml-data/invalid-resource-tag.xml")
+
+
+def test_validate_xml_invalid_resource_tag_problem() -> None:
+    with pytest.raises(
+        InputError,
+        match=r"XML-tags are not allowed in text properties with encoding=utf8\.\n"
+        r"The following resources of your XML file violate this rule:\n"
+        r"    - line 13: resource 'the_only_resource', property ':test'\n"
+        r"    - line 14: resource 'the_only_resource', property ':test'\n"
+        r"    - line 15: resource 'the_only_resource', property ':test'\n"
+        r"    - line 16: resource 'the_only_resource', property ':test'",
+    ):
+        validate_xml(input_file="testdata/invalid-testdata/xml-data/utf8-text-with-xml-tags.xml")
+
+
+def test_validate_xml_data_duplicate_iri() -> None:
+    with pytest.raises(
+        InputError,
+        match=r"The XML file cannot be uploaded due to the following validation error\(s\)\:\n"
+        r"    Line 19\: Element 'resource'\: Duplicate key-sequence \['http://rdfh.ch/4123/54SYvWF0QUW6a'\] "
+        r"in unique identity-constraint 'IRI_attribute_of_resource_must_be_unique'\.",
+    ):
+        validate_xml(input_file="testdata/invalid-testdata/xml-data/duplicate-iri.xml")
+
+
+def test_validate_xml_duplicate_ark() -> None:
+    with pytest.raises(
+        InputError,
+        match=r"The XML file cannot be uploaded due to the following validation error\(s\)\:\n"
+        r"    Line 19\: Element 'resource'\: Duplicate key-sequence \['ark\:/72163/4123-31ec6eab334-a.2022829'\] "
+        r"in unique identity-constraint 'ARK_attribute_of_resource_must_be_unique'\.",
+    ):
+        validate_xml(input_file="testdata/invalid-testdata/xml-data/duplicate-ark.xml")
+
+
+def test_validate_xml_empty_label() -> None:
+    with pytest.raises(
+        InputError,
+        match=r"The XML file cannot be uploaded due to the following validation error\(s\)\:\n"
+        r"    Line 11\: Element 'resource', attribute 'label'\: \[facet 'minLength'\] "
+        r"The value '' has a length of '0'; this underruns the allowed minimum length of '1'\.",
+    ):
+        validate_xml(input_file="testdata/invalid-testdata/xml-data/empty-label.xml")
+
+
+if __name__ == "__main__":
+    pytest.main([__file__])

test/unittests/utils/test_xml_validation_low_level.py

@@ -0,0 +1,78 @@
+import pytest
+from lxml import etree
+
+from dsp_tools.utils.xml_validation import _find_xml_tags_in_simple_text_elements
+
+
+class TestFindXMLTagsInUTF8:
+    def test_find_xml_tags_in_simple_text_elements_all_good(self) -> None:
+        allowed_html_escapes = [
+            "(<2cm) (>10cm)",
+            "text < text/>",
+            "text < text> & text",
+            "text <text text > text",
+            'text < text text="text"> text',
+            'text <text text="text" > text',
+        ]
+        utf8_texts_with_allowed_html_escapes = [
+            f"""
+            <knora shortcode="4123" default-ontology="testonto">
+                <resource label="label" restype=":restype" id="id">
+                    <text-prop name=":name">
+                        <text encoding="utf8">{txt}</text>
+                    </text-prop>
+                </resource>
+            </knora>
+            """
+            for txt in allowed_html_escapes
+        ]
+        for xml in utf8_texts_with_allowed_html_escapes:
+            all_good, msg = _find_xml_tags_in_simple_text_elements(etree.fromstring(xml))
+            assert all_good is True
+            assert msg == ""
+
+    def test_find_xml_tags_in_simple_text_elements_forbidden_escapes(self) -> None:
+        test_ele = etree.fromstring(
+            """
+            <knora shortcode="4123" default-ontology="testonto">
+                <resource label="label" restype=":restype" id="id">
+                    <text-prop name=":name">
+                        <text encoding="utf8">&lt;tag s="t"&gt;</text>
+                    </text-prop>
+                </resource>
+            </knora>
+            """
+        )
+        expected_msg = (
+            "XML-tags are not allowed in text properties with encoding=utf8.\n"
+            "The following resources of your XML file violate this rule:\n"
+            "    - line 5: resource 'id', property ':name'"
+        )
+        all_good, res_msg = _find_xml_tags_in_simple_text_elements(test_ele)
+        assert all_good is False
+        assert res_msg == expected_msg
+
+    def test_find_xml_tags_in_simple_text_elements_forbidden_escapes_two(self) -> None:
+        test_ele = etree.fromstring(
+            """
+            <knora shortcode="4123" default-ontology="testonto">
+                <resource label="label" restype=":restype" id="id">
+                    <text-prop name=":propName">
+                        <text encoding="utf8">&lt;em&gt;text&lt;/em&gt;</text>
+                    </text-prop>
+                </resource>
+            </knora>
+            """
+        )
+        expected_msg = (
+            "XML-tags are not allowed in text properties with encoding=utf8.\n"
+            "The following resources of your XML file violate this rule:\n"
+            "    - line 5: resource 'id', property ':propName'"
+        )
+        all_good, res_msg = _find_xml_tags_in_simple_text_elements(test_ele)
+        assert all_good is False
+        assert res_msg == expected_msg
+
+
+if __name__ == "__main__":
+    pytest.main([__file__])