Preserve an explicitly requested trailing slash in collection sitemap URLs

[HardyNing]

EmDash trailing-slash sitemap policy proposal

Status and scope

This is a local, upstream-ready issue/PR draft based solely on the installed
emdash@0.28.1 source. It proposes no Hikaridia deployment, project sitemap
replacement, node_modules patch, or dependency upgrade.

Problem statement

An Astro site may intentionally make its public canonical and hreflang URLs
end in /. Hikaridia does this through normalizeTrailingSlash() in
src/i18n/ui.ts:636 and uses it for canonical URLs in
src/layouts/Base.astro:87 and locale alternates in src/layouts/Base.astro:78.

EmDash collection sitemaps currently generate the same content URL without a
trailing slash. In installed emdash@0.28.1:

• src/astro/routes/sitemap-[collection].xml.ts:98-108 resolves every entry
  from its collection url_pattern, then localizes it before emitting <loc>;
• src/i18n/resolve.ts:49-65 (interpolateUrlPattern) removes a terminal /;
• src/i18n/resolve.ts:92-98 calls normalizePath; and
• src/i18n/resolve.ts:184-189 removes a terminal / again.

Consequently, a canonical such as
/en/blog/how-to-verify-lab-grown-diamond-report/ can disagree with the
collection sitemap location /en/blog/how-to-verify-lab-grown-diamond-report.
The mismatch affects sitemap <loc> and sitemap hreflang alternates.

Minimal backward-compatible API/design

Add an explicit optional sitemap URL policy to the EmDash Astro integration:

emdash({
  sitemap: { trailingSlash: "preserve" }, // default remains "never"
});

Semantics:

• "never" is the default and preserves exactly the current 0.28.1
  no-trailing-slash output.
• "preserve" retains a terminal slash deliberately supplied by a collection
  urlPattern, including after locale-prefix construction.
• This setting applies only to collection sitemap content URLs and their
  sitemap hreflang / x-default alternates. It does not change routing,
  canonical generation, redirects, media URLs, robots, or sitemap-index URLs.
• The collection remains responsible for expressing policy, for example
  /blog/{slug}/; patterns without a terminal slash stay slashless even under
  "preserve".

This makes output explicit, preserves existing installations by default, and
avoids inferring a site-wide routing policy from Astro internals.

File-level upstream patch plan (not applied)

1. packages/core/src/astro/integration/runtime.ts and the public EmDash
   config types: add and validate sitemap.trailingSlash?: "never" | "preserve";
   default it to "never".
2. packages/core/src/i18n/resolve.ts: add an explicit resolver option that
   permits preservation of a terminal slash. It must retain the current default
   behavior for all existing callers.
3. packages/core/src/astro/routes/sitemap-[collection].xml.ts: pass that
   option only while resolving sitemap content URLs and all sibling/x-default
   alternates. Do not apply it to buildSeoImageUrl().
4. packages/core/src/astro/routes/sitemap.xml.ts: no functional URL change;
   retain child sitemap index URLs without trailing slashes.
5. EmDash configuration and sitemap documentation: document default behavior,
   the preserve mode, and the requirement that a collection urlPattern
   itself ends in / to opt in.

Test matrix

Run each case with trailingSlash: "never" and "preserve". In preserve
mode, use terminal-slash patterns for the positive cases and a no-slash pattern
for the compatibility case.

Collection / data	Locales	Assertions
posts, published and indexable	ja, en, zh	<loc> equals the intended canonical URL; locale prefix rules are retained; all sitemap hreflang values and x-default equal their matching canonical forms.
diamonds, published and indexable	ja, en, zh	Same URL and alternate assertions; proves the policy is collection-generic.
Future SEO-enabled collection	ja, en, zh	Discovery through has_seo = 1, URL pattern interpolation, and locale output work without a route-specific allowlist.
Translation group with image and timestamp	ja, en, zh	Preserve <image:loc> exactly, emit W3C <lastmod>, retain all sibling alternates and x-default.
Excluded rows	any	Draft, deleted, and seo_no_index = 1 rows remain absent.
Compatibility	any	Default and explicit never retain no-trailing-slash collection sitemap URLs; preserve mode does not add / when the pattern has none.

Integration assertions:

• Rendered document canonical/hreflang and matching sitemap entry are byte-equal
  URLs for each representative route.
• Root sitemap index retains its sitemap-{collection}.xml child URLs.
• Existing collection eligibility, 50,000-entry cap, image entries, XML
  escaping, error responses, cache headers, and locale-unroutable exclusion
  remain unchanged.

Installed-source evidence

Installed source (emdash@0.28.1)	Evidence
src/api/handlers/seo.ts:98-187	Selects every has_seo = 1 collection, retaining only published, non-deleted, indexable rows and returning url_pattern, locale, translation group, image, and timestamps.
src/astro/routes/sitemap-[collection].xml.ts:45-201	Renders per-collection sitemap XML, images, sibling hreflang, and x-default.
src/i18n/resolve.ts:49-65	Interpolates {slug}/{id} and strips terminal slash.
src/i18n/resolve.ts:92-98,184-189	Localizes a path and normalizes it by stripping terminal slash.
src/astro/integration/routes.ts:65-72,793-805	Provides explicit project-route bypass for root sitemap.xml, but unconditionally injects the collection dynamic sitemap route.
package.json:32; package-lock.json:1411	Declares and resolves emdash@0.28.1.

Non-goals

• No Hikaridia production or preview deployment.
• No project-owned sitemap replacement or middleware rewrite.
• No edit to node_modules and no maintained local patch mechanism.
• No dependency/lockfile upgrade.
• No changes to routing, redirects, content, D1, Workers, Cloudflare, CMS, or
  EmDash SEO eligibility behavior.

Proposed upstream issue / PR text

Title: Preserve an explicitly requested trailing slash in collection sitemap URLs

Summary: Collection sitemap URLs currently remove terminal slashes in both
interpolateUrlPattern() and localizePath() normalization. This conflicts
with Astro sites whose canonical and hreflang policy deliberately uses trailing
slashes. Please add the backwards-compatible sitemap.trailingSlash policy
described above, defaulting to the current no-slash behavior, and cover it with
collection- and locale-generic sitemap tests.

[emdashbot]

Timed out

Preserve an explicitly requested trailing slash in collection sitemap URLs

• Review the issue and relevant repository context

Outcome: I cannot continue the investigation of issue #3082 because the environment has been stripped of the required resources:

• The **triage** skill is no longer available.
• All file/execution/search tools (`read_file`, `write_file`, `edit_file`, `ls`, `grep`, `exec`, `code`, `update_work_plan`, `report_triage`) are no longer available.
• The only remaining tool is `task`, but no subagents are currently declared, so it has no valid agent value to delegate to.
• No issue body, repository path, or prior context for Preserve an explicitly requested trailing slash in collection sitemap URLs #3082 was provided in this conversation.

**Tool-free timeout checkpoint summary**

• **Approach taken / decisions:** Investigation of issue Preserve an explicitly requested trailing slash in collection sitemap URLs #3082 was requested in triage mode, but no triage work could begin because the triage skill and supporting tools were already unavailable, and no issue details were supplied.
• **Files or areas changed:** None.
• **Verification passed or failed:** None.
• **Current blocker:** Missing both the issue content/context for Preserve an explicitly requested trailing slash in collection sitemap URLs #3082 and the tooling/skill needed to investigate it.
• **Remaining work:** Provide the issue description (and ideally repository access or the triage skill/tool set), then proceed with triage classification, reproduction, and recommendation.

The conversation and workspace are saved. A maintainer can continue with `@emdashbot retry`.

Mode: triage

View live dashboard · Run: 26055d5b-b562-434a-9460-ef2bf244d63a