Merge #503

503: Monkey patch sphinx SortIDs transform to sort our own IDs r=pietroalbini a=Veykril

This fixes some anchoring inconsistencies and also makes link check correctly work for section id <-> other id collisions, see 3cf79ed which CI did not catch so far, now it will. cc ferrocene/ferrocene#699 which broke because of the collision

Co-authored-by: Lukas Wirth <[email protected]>

Author: bors-ferrocene[bot]

exts/ferrocene_spec/__init__.py

@@ -3,6 +3,7 @@
 
 from . import definitions, informational, syntax_directive, std_role, paragraph_ids
 from . import items_with_rubric
+from .utils import FlsSortIds
 from sphinx.domains import Domain
 
 
@@ -37,6 +38,7 @@ def is_empty(data):
 
 def setup(app):
     app.add_domain(SpecDomain)
+    app.add_transform(FlsSortIds)
     definitions.setup(app)
     paragraph_ids.setup(app)
     informational.setup(app)

exts/ferrocene_spec/utils.py

@@ -2,6 +2,7 @@
 # SPDX-FileCopyrightText: The Ferrocene Developers
 
 from docutils import nodes
+from sphinx import transforms
 
 
 def section_id_and_anchor(section):
@@ -23,3 +24,25 @@ def section_id_and_anchor(section):
 
 class NoSectionIdError(RuntimeError):
     pass
+
+
+# Sphinx by default sorts all ids of the form `id[0-9]+` to the end.
+# Our IDs are section name and fls_ id pairs, so in some cases this transform
+# will instead sort the section name to the back, but not always!
+# So we overwrite the transform instead so that our fls ids are sorted to the back.
+# In addition to that we normalize them, as sphinx turns the `_` in `fls_{id}`
+# into `fls-{id}` which can break the link check from working correctly.
+class FlsSortIds(transforms.SphinxTransform):
+    # Run this step after sphinx sorted.
+    default_priority = transforms.SortIds.default_priority + 1
+
+    def apply(self, **kwargs):
+        from docutils import nodes
+
+        for node in self.document.findall(nodes.section):
+            for n, id in enumerate(node["ids"]):
+                if id.startswith("fls-"):
+                    node["ids"][n] = id[:3] + "_" + id[4:]
+            # sort the fls id to the back
+            if len(node["ids"]) > 1 and node["ids"][0].startswith("fls_"):
+                node["ids"] = node["ids"][1:] + [node["ids"][0]]