modules.patch

diff --git a/documentation/doxygen.py b/documentation/doxygen.py
index 008c197..c850f25 100755
--- a/documentation/doxygen.py
+++ b/documentation/doxygen.py
@@ -70,6 +70,14 @@ import latex2svg
 import latex2svgextra
 import ansilexer
 
+# C++ module interface unit extensions. .ixx is MSVC, .cppm/.ccm/.cxxm
+# are Clang, .mxx is the other widely used spelling. Implementation units keep
+# ordinary .cpp/.cc/.cxx suffixes and stay plain files.
+cpp_module_extensions = ('.ixx', '.mxx', '.cppm', '.ccm', '.cxxm')
+
+def is_cpp_module_filename(filename: str) -> bool:
+    return filename.endswith(cpp_module_extensions)
+
 class EntryType(enum.Enum):
     # Order must match the search_type_map below; first value is reserved for
     # ResultFlag.ALIAS
@@ -1335,6 +1343,13 @@ def parse_desc_internal(state: State, element: ET.Element, immediate_parent: ET.
             # Custom mapping of filenames to languages
             mapping = [('.h', 'c++'),
                        ('.h.cmake', 'c++'),
+                       # C++ module interface units. Pygments does not know
+                       # these suffixes, so map them explicitly.
+                       ('.cppm', 'c++'),
+                       ('.cxxm', 'c++'),
+                       ('.ixx', 'c++'),
+                       ('.mxx', 'c++'),
+                       ('.ccm', 'c++'),
                        # Pygments knows only .vert, .frag, .geo
                        ('.glsl', 'glsl'),
                        ('.conf', 'ini'),
@@ -2473,7 +2488,8 @@ def extract_metadata(state: State, xml):
         return
     assert len([i for i in root]) == 1
 
-    if compounddef.attrib['kind'] not in ['namespace', 'group', 'class', 'struct', 'union', 'dir', 'file', 'page']:
+    # 'module' is a C++ module (export module), not a Doxygen group.
+    if compounddef.attrib['kind'] not in ['namespace', 'group', 'class', 'struct', 'union', 'dir', 'file', 'page', 'module']:
         logging.debug("No useful info in {}, skipping".format(state.current))
         return
 
@@ -2503,7 +2519,7 @@ def extract_metadata(state: State, xml):
     # ElementTree deprecated the __bool__ conversion of Element, so I now have
     # to check that `detaileddescription` actually has any children. Checking
     # against None is not enough as it could be present but be empty.
-    compound.has_details = compound.kind == 'group' or compound.brief or len(compounddef.find('detaileddescription')) or (compound.kind == 'page' and not is_a_stupid_empty_markdown_page(compounddef))
+    compound.has_details = compound.kind in ['group', 'module'] or compound.brief or len(compounddef.find('detaileddescription')) or (compound.kind == 'page' and not is_a_stupid_empty_markdown_page(compounddef))
     compound.children = []
 
     # Version badges, deprecation status. If @since is followed by
@@ -2572,6 +2588,10 @@ def extract_metadata(state: State, xml):
     elif compounddef.attrib['kind'] == 'group':
         for i in compounddef.findall('innergroup'):
             compound.children += [i.attrib['refid']]
+    elif compounddef.attrib['kind'] == 'module':
+        # Partitions and implementation units Doxygen nests under the module.
+        for i in compounddef.findall('innermodule'):
+            compound.children += [i.attrib['refid']]
 
     state.compounds[compound.id] = compound
 
@@ -2584,7 +2604,7 @@ def postprocess_state(state: State):
 
     # Strip name of parent symbols from names to get leaf names
     for _, compound in state.compounds.items():
-        if not compound.parent or compound.kind in ['file', 'page', 'group']:
+        if not compound.parent or compound.kind in ['file', 'page', 'group', 'module']:
             compound.leaf_name = compound.name
             continue
 
@@ -2826,6 +2846,9 @@ def parse_xml(state: State, xml: str):
     # This is similar to compound.url handling in extract_metadata() above.
     compound.url_base = ('index' if compound.id == 'indexpage' else compound.id)
     compound.url = compound.url_base + '.html'
+    # Interface unit vs implementation unit is an extension convention; Doxygen
+    # still emits kind="file" for the source file itself.
+    compound.is_module_interface = compound.kind == 'file' and is_cpp_module_filename(compound.name)
     # Save current compound URL for search data building and ID extraction,
     # save current URL prefix for extract_id_hash() (these are the same for
     # top-level desc, but usually not for definitions, as an enum can be both
@@ -2878,7 +2901,7 @@ def parse_xml(state: State, xml: str):
 
     # Build breadcrumb. Breadcrumb for example pages is built after everything
     # is parsed.
-    if compound.kind in ['namespace', 'group', 'struct', 'class', 'union', 'file', 'dir', 'page']:
+    if compound.kind in ['namespace', 'group', 'struct', 'class', 'union', 'file', 'dir', 'page', 'module']:
         # Gather parent compounds
         path_reverse = [compound.id]
         while path_reverse[-1] in state.compounds and state.compounds[path_reverse[-1]].parent:
@@ -3089,7 +3112,7 @@ def parse_xml(state: State, xml: str):
                             assert compounddef_child.attrib['prot'] == 'protected'
                             compound.protected_types += [('class', class_)]
                     else:
-                        assert compound.kind in ['namespace', 'group', 'file']
+                        assert compound.kind in ['namespace', 'group', 'file', 'module']
                         compound.classes += [class_]
 
         # Base class (if it links to anywhere)
@@ -3152,6 +3175,20 @@ def parse_xml(state: State, xml: str):
 
                     compound.derived_classes += [class_]
 
+        # C++ module partition (not a Doxygen group)
+        elif compounddef_child.tag == 'innermodule':
+            assert compound.kind == 'module'
+            id = compounddef_child.attrib['refid']
+            if id in state.compounds and state.compounds[id].has_details:
+                symbol = state.compounds[id]
+                module = Empty()
+                module.url = symbol.url
+                module.name = symbol.leaf_name
+                module.brief = symbol.brief
+                module.deprecated = symbol.deprecated
+                module.since = symbol.since
+                compound.modules += [module]
+
         # Module (*not* member group)
         elif compounddef_child.tag == 'innergroup':
             assert compound.kind == 'group'
@@ -3628,6 +3665,8 @@ def parse_xml(state: State, xml: str):
             kind = EntryType.DIR
         elif compound.kind == 'file':
             kind = EntryType.FILE
+        elif compound.kind == 'module':
+            kind = EntryType.FILE
         elif compound.kind == 'page':
             kind = EntryType.PAGE
         elif compound.kind == 'group':
@@ -3719,6 +3758,10 @@ def parse_index_xml(state: State, xml):
                 top_level_dirs += [entry]
             elif compound.kind == 'file':
                 top_level_files += [entry]
+            elif compound.kind == 'module':
+                # Listed with files so they are reachable without a new navbar
+                # slot. Doxygen groups keep the existing "modules" index.
+                top_level_files += [entry]
             else:
                 assert compound.kind == 'page'
                 # Ignore index page in page listing, add it later only if it
@@ -3736,7 +3779,7 @@ def parse_index_xml(state: State, xml):
                     orphan_pages[compound.parent] = {}
                 orphan_pages[compound.parent][entry.id] = entry
             else:
-                assert compound.kind in ['class', 'struct', 'union', 'file']
+                assert compound.kind in ['class', 'struct', 'union', 'file', 'module']
                 if not compound.parent in orphans:
                     orphans[compound.parent] = []
                 orphans[compound.parent] += [entry]
diff --git a/documentation/templates/doxygen/file.html b/documentation/templates/doxygen/file.html
index 2427983..6fbb032 100644
--- a/documentation/templates/doxygen/file.html
+++ b/documentation/templates/doxygen/file.html
@@ -1,10 +1,10 @@
 {% extends 'base-reference.html' %}
 
-{% block title %}{% set j = joiner('/') %}{% for name, _ in compound.breadcrumb %}{{ j() }}{{ name }}{% endfor %} file | {{ super() }}{% endblock %}
+{% block title %}{% set j = joiner('/') %}{% for name, _ in compound.breadcrumb %}{{ j() }}{{ name }}{% endfor %} {% if compound.is_module_interface %}module{% else %}file{% endif %} | {{ super() }}{% endblock %}
 
 {% block header %}
         <h1>
-          {%+ for name, target in compound.breadcrumb[:-1] %}<span class="m-breadcrumb"><a href="{{ target }}">{{ name }}</a>/</span>{% endfor %}{{ compound.breadcrumb[-1][0] }} <span class="m-thin">file</span>{% if compound.since %} {{ compound.since }}{% endif %}
+          {%+ for name, target in compound.breadcrumb[:-1] %}<span class="m-breadcrumb"><a href="{{ target }}">{{ name }}</a>/</span>{% endfor %}{{ compound.breadcrumb[-1][0] }} <span class="m-thin">{% if compound.is_module_interface %}module{% else %}file{% endif %}</span>{% if compound.since %} {{ compound.since }}{% endif %}
           {# need an explicit space here otherwise the newline gets removed #}
 
         </h1>
diff --git a/documentation/templates/doxygen/module.html b/documentation/templates/doxygen/module.html
new file mode 100644
index 0000000..cacd35d
--- /dev/null
+++ b/documentation/templates/doxygen/module.html
@@ -0,0 +1,11 @@
+{% extends 'base-reference.html' %}
+
+{% block title %}{% set j = joiner('/') %}{% for name, _ in compound.breadcrumb %}{{ j() }}{{ name }}{% endfor %} module | {{ super() }}{% endblock %}
+
+{% block header %}
+        <h1>
+          {%+ for name, target in compound.breadcrumb[:-1] %}<span class="m-breadcrumb"><a href="{{ target }}">{{ name }}</a>/</span>{% endfor %}{{ compound.breadcrumb[-1][0] }} <span class="m-thin">module</span>{% if compound.since %} {{ compound.since }}{% endif %}
+          {# need an explicit space here otherwise the newline gets removed #}
+
+        </h1>
+{% endblock %}