Data origins, validation, cache behavior, manifests, staff groups, student batches, search, prefetching, and sitemap scope.
The People loader builds URLs from NEXT_PUBLIC_PEOPLE_DATA_BASE_URL and uses
cache: no-store at the network layer. A module-level promise cache deduplicates
requests for five minutes; rejected promises are removed so a later request can
retry. Prefetch helpers intentionally ignore failures because navigation still
owns the visible error state.
| Input | Validation and missing-data behavior |
|---|---|
manifest.json | validates programmes and batches; missing becomes empty manifest |
student-streams.json | validates public stream definitions; missing becomes empty list |
search.json | validates searchable entries; missing becomes empty list |
staff/*.json | shared staff schemas; missing becomes empty list |
special/{programme}/{batch}.json | shared student schema; missing becomes empty list |
users/{identity}.json | must parse as staff or student; missing is an error |
Projects by-person/{username}.json | validates usable project summaries; missing becomes empty list |
Staff records use the owning schemas from @scs/schemas. Person profiles try
the unified staff schema before the student schema. The smaller manifest,
search, stream, and project-index contracts are validated at the loader
boundary and malformed entries either fail the collection or are omitted,
according to that contract.
The manifest is authoritative for which programmes contain which batches.
Programme pages divide batches into current and alumni groups using each
batch's defaultCategory. A batch route validates that the requested batch
belongs to the requested programme before loading its students, then filters
published students so an alumni directory contains alumni and a current
directory excludes them.
A registration number supplies the batch candidate, but not the final programme. Profile breadcrumbs test the matching student lists for every manifest programme containing that batch and choose the programme whose published list contains that exact registration number. This avoids guessing a student's directory from the first subject or department value on the profile.
Prefetching is bounded: likely batches are selected from current and alumni groups, adjacent batches are prefetched around the active one, and at most 12 visible profiles are prefetched. Preserve those bounds when changing directory navigation.
The header search index is loaded on focus or first input and reused in the component. Matching is case-insensitive substring search over precomputed keywords, and the panel displays at most eight results. It has explicit loading, failure, no-result, clear, and Escape-to-close states.
The sitemap publishes stable top-level staff, student, and subject pages. It
does not enumerate profiles or dynamic batches from remote data. Canonical
username requests redirect permanently to /people/{identity}.
People application internals
Runtime map of published people data, directory construction, profile interpretation, search, and project participation.
People profile interpretation
How staff and student fields become labelled profile sections, how directory context differs from academics, and how project roles are shown.
