fix: ship the real DTD in the Pelican plugin

pelican-mews/tests/conftest.py

@@ -21,7 +21,15 @@ THEME = TESTS / "theme"
 
 BASE_SETTINGS = {
     "ARTICLE_PATHS": [],
-    "MARKDOWN": {"output_format": "xhtml"},
+    # Tables and definition lists are both in the Mews subset, and both need
+    # an extension turned on before python-markdown will emit them.
+    "MARKDOWN": {
+        "output_format": "xhtml",
+        "extension_configs": {
+            "markdown.extensions.tables": {},
+            "markdown.extensions.def_list": {},
+        },
+    },
     "PAGE_PATHS": ["pages"],
     "PAGE_SAVE_AS": "{slug}.xhtml",
     "PAGE_URL": "{slug}.xhtml",

pelican-mews/tests/fixtures/structure/expected.xhtml

@@ -0,0 +1,35 @@
+<?xml version='1.0' encoding='UTF-8'?>
+<!DOCTYPE html PUBLIC "-//WAPFORUM//DTD XHTML Mobile 1.2//EN" "http://www.openmobilealliance.org/tech/DTD/xhtml-mobile12.dtd">
+<html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en">
+<head>
+<title>Structure</title>
+    <meta name="mews-profile" content="0.1"/>
+    <meta name="viewport" content="width=device-width"/>
+    <link rel="stylesheet" type="text/css" href="css/mews-0.1.css"/></head>
+<body>
+<p>A table, an ordered list and a definition list: everything the placeholder
+DTD used to throw away.</p>
+<table>
+<tr>
+
+<th>Network</th>
+<th>Band</th>
+</tr>
+
+<tr>
+
+<td>GSM 900</td>
+<td>900 MHz</td>
+</tr>
+
+</table>
+<ol>
+<li>First</li>
+<li>Second</li>
+</ol>
+<dl>
+<dt>Term</dt>
+<dd>Its meaning</dd>
+</dl>
+</body>
+</html>

pelican-mews/tests/fixtures/structure/input.md

@@ -0,0 +1,16 @@
+Title: Structure
+Slug: page
+Date: 2026-10-11
+
+A table, an ordered list and a definition list: everything the placeholder
+DTD used to throw away.
+
+| Network | Band |
+| --- | --- |
+| GSM 900 | 900 MHz |
+
+1. First
+2. Second
+
+Term
+:   Its meaning

pelican-mews/tests/test_mews.py

@@ -18,7 +18,7 @@ PACKAGE = "pelican.plugins.mews"
 STRICT_OVERRIDES = {"MEWS_STRICT": True}
 
 
-@pytest.mark.parametrize("case", ["basic", "strip", "cssdir"])
+@pytest.mark.parametrize("case", ["basic", "strip", "cssdir", "structure"])
 def test_written_page_matches_expected(tmp_path, case):
     """The written page matches the fixture, whatever was stripped."""
     overrides = {"MEWS_CSS_DIR": "style"} if case == "cssdir" else {}
@@ -28,7 +28,7 @@ def test_written_page_matches_expected(tmp_path, case):
     assert normalize(page) == normalize(expected)
 
 
-@pytest.mark.parametrize("case", ["basic", "strip", "cssdir"])
+@pytest.mark.parametrize("case", ["basic", "strip", "cssdir", "structure"])
 def test_css_is_copied(tmp_path, case):
     """The bundled stylesheet lands under the configured directory."""
     overrides = {"MEWS_CSS_DIR": "style"} if case == "cssdir" else {}
@@ -81,13 +81,28 @@ def test_parallel_output_folder(tmp_path):
     assert not (output / "css").exists()
 
 
+# Section 4.1 of the specification keeps this many elements. The plugin lists
+# none of them itself; the DTD is the only place they are written down.
+SECTION_4_1_ELEMENTS = 51
+
+
 def test_allowlist_is_read_from_the_dtd():
     """The subset comes from the DTD, not from code."""
     dtd = files(PACKAGE).joinpath("dtd", "mews.dtd").read_text(encoding="utf-8")
     allowlist = parse_allowlist(dtd)
-    assert allowlist["p"] == {"id", "title", "xml:lang"}
-    assert allowlist["html"] == {"dir", "id", "lang", "title", "xml:lang", "xmlns"}
+    assert len(allowlist) == SECTION_4_1_ELEMENTS
+    # The common attributes, including the two the profile adds to XHTML-MP.
+    assert allowlist["p"] == {"dir", "id", "lang", "title", "xml:lang"}
+    # html takes no id or title; section 4.1 gives it these four and no more.
+    assert allowlist["html"] == {"dir", "lang", "xml:lang", "xmlns"}
     assert "href" in allowlist["a"]
+    # class is permitted on body and nowhere else (section 5.3).
+    assert "class" in allowlist["body"]
+    assert not [
+        name for name, attrs in allowlist.items() if name != "body" and "class" in attrs
+    ]
+    # The parts the placeholder DTD lacked, which the plugin was stripping.
+    assert {"table", "td", "ol", "dl", "form", "input", "address"} <= set(allowlist)
 
 
 @pytest.mark.skipif(shutil.which("uv") is None, reason="uv is not installed")