End-user oriented frontpage, footer and person listings

[namedgraph]

Makes the deployed dataspace read as a public data portal rather than a developer demo, and gives every person listing usable filter and sort columns.

Frontpage (app/root.ttl)

The intro dropped its RDF vocabulary — knowledge graph, global URIs, Wikidata/EU vocabulary alignment, the SPARQL spec link and the GitHub EXAMPLES.md link — for what the site is and what you can do with it: official registry data in one place, linked so you can walk from a county to its municipalities or from an MP to their committees and party. dct:description got the same treatment (it renders directly above the intro), the browse prompt lost LDH's "konteineriai" jargon, and the factions chart no longer explains itself in terms of the temporal model ("narystė dar neturi pabaigos datos"). The Turtle comments explaining the DESCRIBE/GridMode constraints stay — they never reach the page.

Footer (files/layout.xsl)

Replaces the stock LinkedDataHub footer with an LTLOD one: own wordmark, dataset shortcuts, source attribution (Registrų centras, get.data.gov.lt, Seimo kanceliarija, Wikidata), licence, and one column of developer entry points — SPARQL, example queries, dataset downloads, {base}ns — which is where they belong instead of the frontpage.

It keeps the stock markup contract, because app.css styles the footer structurally: .ldh-footer > .cols is a 1.4fr repeat(4, 1fr) grid, so exactly one .brand-col plus four .col children, each a .ftitle followed by bare <a>s, then .legal with two space-between spans. The only deviation is the wordmark .mark, restyled inline to the Lithuanian tricolour with the class geometry kept.

Footer-only in layout.xsl rather than overrides.xsl: bs2:Footer is applied once server-side and never re-rendered during CSR, so it needs no client-side twin and no SEF rebuild.

Person listings (app/persons.ttl, app/ns.ttl, files/client.xsl)

The persons/ container swaps the stock ldh:ChildrenView for its own ldh:View + sp:Select in ac:GridMode. LDH derives a view's facet pills and sort options from the BGP triples whose subject is the query's first projected variable and whose object is a variable, so the focus has to be the person entity (?person = {doc}#this) rather than the dh:Item document — that is where foaf:givenName / foaf:familyName / foaf:gender / ltlod:nominatedBy sit. The stock ChildrenView projects the document and offers no person columns at all. The same columns go into the person-valued 1:N views (:CurrentMembers, :FormerMembers, :NominatedMembers, :MembersElectedHere) so every person listing filters and sorts alike, and the faceted predicates get rdfs:labels in {base}ns (an unlabelled predicate degrades to its local name in the pill).

That needed a performance fix to be usable. To label a filter pill's values LDH appends an OPTIONAL with a 7-way alternative property path, and ARQ does not resolve an alternative path through the quad indexes inside a GRAPH block — it walks all ~117k named graphs, ~104 s per pill, a 500 in the browser. The ldh:bgp-value-counts override emits the equivalent variable predicate + FILTER (?labelProp IN (…)) — same predicates in the same order, so SAMPLE(?label) is unchanged — at ~0.1 s per pill.

Two rules this surfaced (CLAUDE.md)

• Declare xmlns="http://www.w3.org/1999/xhtml" on xsl:stylesheet in any override that emits literal result elements. LDH's own layout.xsl declares it, so its bare <div>s are XHTML; without it ours land in no namespace and the serialiser emits a stray xmlns="" on the block. HTML parsers ignore the attribute, but namespace-sensitive XPath/XSLT matching on xhtml:* then misses those nodes. Verified on the running server before and after.
• docker compose restart linkeddatahub does not pick up an edited mounted stylesheet — it keeps serving the one compiled at the previous start. docker compose up -d --force-recreate linkeddatahub does, followed by an nginx/varnish restart per the 502 gotcha.

Notes

• docker-compose.override.yml pins the runtime image to LDH e38f693d; the design system the footer targets only exists there, and the SEF and layout override are compiled and mounted against it.
• .gitignore now lists docker-compose.override.yml, which is inert while the file stays tracked. Untracking it would also drop the tdb-loader service and the TZ/signup settings that make load and make up rely on, so it stays committed for now — worth a separate decision.

🤖 Generated with Claude Code

https://claude.ai/code/session_01PoiMWyMGtWZos9wWoVUCwg