feat: scaffold plain .html pages that need no server config

mews/quickstart.py

@@ -41,13 +41,16 @@ DEFAULT_LANG = "en"
 THEME = "theme"
 
 # The plugin reads the allowed elements and attributes from the Mews DTD and
-# brings every page inside that subset. It only touches files ending .xhtml,
-# so everything has to be written out under that suffix.
+# brings every page inside that subset.
 PLUGINS = ["pelican.plugins.mews"]
 
-ARTICLE_URL = ARTICLE_SAVE_AS = "{{slug}}.xhtml"
-PAGE_URL = PAGE_SAVE_AS = "{{slug}}.xhtml"
-INDEX_SAVE_AS = "index.xhtml"
+# This whole site is Mews, so its pages are plain .html: a server sends those
+# as text/html, which section 7.1 asks for so a small markup error still shows
+# as a page, and index.html is the directory index everywhere. The plugin
+# defaults to .xhtml instead, for sites that are only partly Mews.
+MEWS_SUFFIX = ".html"
+ARTICLE_URL = ARTICLE_SAVE_AS = "{{slug}}.html"
+PAGE_URL = PAGE_SAVE_AS = "{{slug}}.html"
 
 # Markdown has to produce XML for the plugin to parse. Tables and definition
 # lists are both in the Mews subset but need their extensions turned on.
@@ -59,7 +62,11 @@ MARKDOWN = {{
     }},
 }}
 
-# Mews pages carry no author styling, so there is nothing for these to show.
+# Every page written under MEWS_SUFFIX goes through the plugin, and Pelican
+# falls back to its own HTML templates for these when the theme has none. They
+# have nothing to show on a site like this, so they stay off; give them
+# templates in theme/templates/ if you want them.
+DIRECT_TEMPLATES = ["index"]
 AUTHOR_SAVE_AS = CATEGORY_SAVE_AS = TAG_SAVE_AS = ""
 AUTHORS_SAVE_AS = CATEGORIES_SAVE_AS = TAGS_SAVE_AS = ARCHIVES_SAVE_AS = ""
 # Pelican writes several feeds by default and they are all broken until
@@ -115,7 +122,7 @@ ARTICLE = """\
 {% block content %}
     <h1>{{ article.title }}</h1>
 {{ article.content }}
-    <p><a href="index.xhtml">All posts</a></p>
+    <p><a href="index.html">All posts</a></p>
 {% endblock %}
 """
 
@@ -125,7 +132,7 @@ PAGE = """\
 {% block content %}
     <h1>{{ page.title }}</h1>
 {{ page.content }}
-    <p><a href="index.xhtml">All posts</a></p>
+    <p><a href="index.html">All posts</a></p>
 {% endblock %}
 """
 
@@ -157,7 +164,8 @@ A Mews site. The pages follow [Mews Profile 0.1](https://mews.page/spec/0.1).
 uv run pelican content
 ```
 
-The finished pages land in `output/`. To work on it with a reload on save:
+The finished pages land in `output/`, as plain `.html` that any web server
+serves correctly with no configuration. To work on it with a reload on save:
 
 ```shell
 uv run pelican --autoreload --listen content
@@ -165,12 +173,7 @@ uv run pelican --autoreload --listen content
 
 ## Publish it
 
-Copy `output/` to any web server. Two things are worth setting:
-
-- Serve `.xhtml` files as `text/html`. Section 7.1 of the spec asks for that,
-  so a browser still shows a page carrying a small markup error.
-- Make `index.xhtml` the directory index, since the plugin only processes
-  files under that suffix. In Caddy: `file_server {{ index index.xhtml }}`.
+Copy `output/` to any web server. Nothing to configure.
 
 ## Subscribe to it
 

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."""

tests/test_quickstart.py

@@ -35,9 +35,15 @@ def test_the_settings_are_valid_python(tmp_path):
     )
     assert settings["SITENAME"] == "Signal Bars"
     assert settings["PLUGINS"] == ["pelican.plugins.mews"]
-    # The plugin only touches this suffix, so everything has to be saved under it.
-    assert settings["ARTICLE_SAVE_AS"].endswith(".xhtml")
-    assert settings["INDEX_SAVE_AS"].endswith(".xhtml")
+    # A site that is Mews throughout writes plain .html: servers send that as
+    # text/html (section 7.1) and index.html is the directory index everywhere.
+    # The suffix the plugin acts on has to agree with where pages are saved.
+    assert settings["MEWS_SUFFIX"] == ".html"
+    assert settings["ARTICLE_SAVE_AS"].endswith(".html")
+    assert settings["PAGE_SAVE_AS"].endswith(".html")
+    # Pelican falls back to its own HTML for any direct template the theme
+    # lacks, and with .html the plugin would choke on it.
+    assert settings["DIRECT_TEMPLATES"] == ["index"]
     assert settings["MARKDOWN"]["output_format"] == "xhtml"
 
 
@@ -62,11 +68,11 @@ def test_the_scaffolded_site_builds_conforming_pages(tmp_path):
         check=False,
     )
     assert built.returncode == 0, built.stderr
-    pages = sorted((site / "output").glob("*.xhtml"))
-    assert [p.name for p in pages] == ["first-post.xhtml", "index.xhtml"]
+    pages = sorted((site / "output").glob("*.html"))
+    assert [p.name for p in pages] == ["first-post.html", "index.html"]
     for page in pages:
         report = validate_file(str(page))
         assert report.conforms, "\n".join(str(f) for f in report.failures)
     # Section 6.1: the index has to be subscribable.
-    index = (site / "output" / "index.xhtml").read_text()
-    assert '<li><a href="first-post.xhtml">' in index
+    index = (site / "output" / "index.html").read_text()
+    assert '<li><a href="first-post.html">' in index