Better handling of multiple matching template wildcard paths

[simonw]

I tried building this:

templates/pages/{topic}.html
templates/pages/{topic}/{slug}.html

And it didn't work - hits to /foo/bar which should have been rendered by the {slug}.html template were instead rendered by the top level {topic.html} template.

[simonw]

Relevant code:

datasette/datasette/app.py

Lines 1030 to 1037 in 14982bd

	if template is None:
	# Try for a pages/blah/{name}.html template match
	for regex, wildcard_template in self.page_routes:
	match = regex.match(request.scope["path"])
	if match is not None:
	context.update(match.groupdict())
	template = wildcard_template
	break

datasette/datasette/app.py

Lines 951 to 954 in 14982bd

	self.page_routes = [
	(route_pattern_from_filepath(filepath[len("pages/") :]), filepath)
	for filepath in pattern_templates
	]

datasette/datasette/app.py

Lines 1197 to 1207 in 14982bd

	def route_pattern_from_filepath(filepath):
	# Drop the ".html" suffix
	if filepath.endswith(".html"):
	filepath = filepath[: -len(".html")]
	re_bits = ["/"]
	for bit in _curly_re.split(filepath):
	if _curly_re.match(bit):
	re_bits.append("(?P<{}>[^/]*)".format(bit[1:-1]))
	else:
	re_bits.append(re.escape(bit))
	return re.compile("".join(re_bits))

[simonw]

This line was meant to handle this case, by only matching non-/ characters. Looks like it did not have the desired effect:

datasette/datasette/app.py

Line 1204 in 14982bd

re_bits.append("(?P<{}>[^/]*)".format(bit[1:-1]))

Maybe because the finished regex doesn't include ^ and $ anchors?

[simonw]

One option: compare the length of the matches and pick the shortest one.