feat: let a site choose which suffix the plugin treats as Mews

pelican-mews/README.md

@@ -52,6 +52,7 @@ Three optional settings:
 | `MEWS_STRICT` | `False` | Fail the build on content outside the subset instead of stripping it. Removals are always logged either way. |
 | `MEWS_CSS_DIR` | `"css"` | Directory below the output root where `mews-0.1.css` is copied. Themes see the resulting path as the `MEWS_CSS` global. |
 | `MEWS_OUTPUT_PATH` | unset | Put the finished pages in this folder instead of `OUTPUT_PATH`. |
+| `MEWS_SUFFIX` | `".xhtml"` | Which written files are Mews pages. Set it to `".html"` on a site that is Mews throughout, and save your pages under that suffix too: servers send `.xhtml` as `application/xhtml+xml`, which makes a browser parse the page as XML and fail outright on a small markup error, while section 7.1 asks for `text/html` so the page still shows. `index.html` is also the directory index everywhere, which `index.xhtml` is not. Note that `.html` puts every page Pelican writes through the plugin, including the index, tag and archive pages it falls back to when the theme has no template for them, so give those a Mews template or switch them off with `DIRECT_TEMPLATES` and the matching `*_SAVE_AS` settings. |
 
 ### Rendering beside an existing blog
 

pelican-mews/pelican/plugins/mews/mews.py

@@ -47,7 +47,13 @@ LOGGER = logging.getLogger(__name__)
 
 CSS_NAME = "mews-0.1.css"
 DEFAULT_CSS_DIR = "css"
-XHTML_SUFFIX = ".xhtml"
+# Which files Pelican writes are Mews pages. The default keeps the plugin off
+# everything else, so it can run on a site that also produces ordinary HTML.
+# A site that is Mews throughout wants ".html" instead: servers send .xhtml as
+# application/xhtml+xml, which makes a browser parse as XML and fail hard on a
+# small markup error, and section 7.1 asks for text/html precisely to avoid
+# that. Set MEWS_SUFFIX to choose.
+DEFAULT_SUFFIX = ".xhtml"
 
 ALLOWLIST = parse_allowlist(
     files(__package__).joinpath("dtd", "mews.dtd").read_text(encoding="utf-8")
@@ -67,6 +73,12 @@ def _on_initialized(pelican) -> None:
         overrides.append(templates)
 
 
+def _suffix() -> str:
+    """Return the file extension a Mews page is written under, with its dot."""
+    suffix = str(_settings.get("MEWS_SUFFIX", DEFAULT_SUFFIX)).strip()
+    return suffix if suffix.startswith(".") else f".{suffix}"
+
+
 def _on_generator_init(generator) -> None:
     css_dir = generator.settings.get("MEWS_CSS_DIR", DEFAULT_CSS_DIR)
     generator.env.globals["MEWS_CSS"] = f"{css_dir}/{CSS_NAME}"
@@ -74,7 +86,7 @@ def _on_generator_init(generator) -> None:
 
 def _on_content_written(path, **_) -> None:
     file = Path(path)
-    if file.suffix != XHTML_SUFFIX:
+    if file.suffix != _suffix():
         return
     try:
         tree, doctype = parse_page(file.read_bytes())

pelican-mews/tests/test_mews.py

@@ -28,6 +28,44 @@ def test_written_page_matches_expected(tmp_path, case):
     assert normalize(page) == normalize(expected)
 
 
+def test_a_site_can_choose_its_own_suffix(tmp_path):
+    """A site that is Mews throughout writes .html, so servers send text/html."""
+    output = build_case(
+        "basic",
+        tmp_path,
+        MEWS_SUFFIX=".html",
+        PAGE_SAVE_AS="{slug}.html",
+        PAGE_URL="{slug}.html",
+        # With .html the plugin sees every page Pelican writes, including the
+        # index, tags and archives it falls back to when the theme has no
+        # template for them. Those fallbacks are HTML, not XML, so a site
+        # using .html has to give every direct template a Mews template or
+        # switch the lot off, as here.
+        DIRECT_TEMPLATES=[],
+    )
+    page = (output / "page.html").read_bytes()
+    assert normalize(page) == normalize(
+        (FIXTURES / "basic" / "expected.xhtml").read_bytes()
+    )
+
+
+def test_the_default_suffix_leaves_other_html_alone(tmp_path):
+    """Left at its default the plugin must not touch a plain HTML page."""
+    output = build_case(
+        "strip",
+        tmp_path,
+        PAGE_SAVE_AS="{slug}.html",
+        PAGE_URL="{slug}.html",
+        INDEX_SAVE_AS="",
+    )
+    # The strip fixture carries a <span>, a class and a comment, all three of
+    # which the plugin removes. Untouched, every one of them survives.
+    written = (output / "page.html").read_bytes()
+    assert b"<span" in written
+    assert b'class="big"' in written
+    assert b"stray comment" in written
+
+
 @pytest.mark.parametrize("case", ["basic", "strip", "cssdir", "structure"])
 def test_css_is_copied(tmp_path, case):
     """The bundled stylesheet lands under the configured directory."""