Skip to content

D7 Theme / UI Commonalities Audit

Audience: Developers and Yuji (Sprint 2 planning) Date: 2026-08-25 Source: Legacy D7 mandala-drupal codebase (sites/all/themes/), checked out locally at ~/Sandbox/Mandala/Site/mandala-drupal Relates to: ADR 010 (internal remodeling permitted; user-facing behavior is the floor), Images interactive viewing surfaces gap, ADR 009 (Phase 2 forks to per-site tracks)

Purpose

Yuji's Images data migration is closing out Sprint 1. Before Phase 2 forks into independent per-site tracks (Than on Texts, Xiaoming on Sources, AV last), establish whether the D7 sites share enough theme structure to justify a baseline D11 theme now — so each site owner extends a shared foundation instead of independently reinventing chrome that was never actually site-specific in D7.

This audit inventories the real D7 theme layer (not memory, not assumption) and sorts findings into shared baseline vs. genuinely site-specific, per the ADR 010 vocabulary ("faithful port" vs. "internal remodeling permitted").

Key finding: it's one theme, not six

Every one of D7's site-facing themes is a Bootstrap-based sub-theme of a single parent, shanti_sarvaka ("A responsive Drupal 7 theme for all Shanti@UVa sites, built with Bootstrap"). This was not an accident of naming — it is declared in each .info file's base theme = shanti_sarvaka line, and confirmed by file size: the base theme's template.php is 1,677 lines; every sub-theme's is a fraction of that (193–860 lines).

Site (D7 site as built; Visuals excluded from D11 scope — see note below) D7 theme template.php size Sub-theme's only real additions
Images sarvaka_images 860 lines Gallery CSS, broken-image fallback setting, shanti-main-images.js (drives the OpenSeadragon deep-zoom viewer — see the interactive surfaces note)
AV sarvaka_mediabase 653 lines AV-transcript CSS, media-edit-form CSS, body_tag/icon_class = audio-video. Kaltura itself is not theme code — it's the contrib kaltura module + KalturaClient library (sites/all/modules/contrib/kaltura, sites/all/libraries/KalturaClient)
Texts shanti_sarvaka_texts 193 lines (smallest) Texts CSS/JS, jquery.ui.tabs; depends on the shanti_texts and shanti_collections_admin modules, not theme code
Sources sources_theme 450 lines Sources CSS/JS; several commented-out stylesheets (search-bibliography, fancytree, biblio-full-page) — bibcite/citation UI was present but partially disabled even in D7
Visuals sarvaka_shiva 488 lines Depends on the shivanode module; body_tag/icon_class = visuals
Home / KMaps browse sarvaka_kmaps (not separately measured) KMaps subject/places explorer CSS/JS (kmaps-explorer.css, shanti-mandala-homepage-kmaps.css) — this is the taxonomy-browse chrome, not a distinct site skin

Note on scope: the current top-level docs describe "five sites (AV, Images, Sources, Texts, Mandala Home)," and the D7 codebase's actual sites.php multisite map and theme set show a sixth, distinct Visuals site (sarvaka_shiva, host visuals.mandala.library.virginia.edu) — matching what session logs from 2026-06-12, 2026-06-15, and 2026-07-30 already independently found (Visuals appears in the reindeer_x site list and the Spike 6 host map). Decided 2026-08-25 (team): Visuals is being retired, not migrated to D11 — no content migration, no theme work. So the top-level docs' "five sites" count is correct for D11 scope; Visuals is included in this audit only as a data point about what D7 actually built, not as Phase 2 scope. A ninth theme, sarvaka_projects, also exists in the codebase but doesn't map to any live D11-bound host in sites.php — likely legacy/unused; worth a quick confirmation, not a blocker to this audit's conclusion.

What's actually shared (the baseline)

All six site sub-themes declare the exact same twelve regions, verbatim, in every .info file: header, banner, content, search_flyout, search_results, sidebar_first, sidebar_second, highlighted, help, page_top, page_bottom, footer, admin_footer. No site introduced or dropped a region. That's a strong signal this was one designed layout system, not six independent ones that happened to converge.

The base theme owns the actual page skeleton and the vast majority of shared behavior — sub-themes add CSS/JS on top, they don't override structure:

  • Page templates (shanti_sarvaka/templates/): page.tpl.php, html.tpl.php, node.tpl.php, page--403.tpl.php, page--404.tpl.php, page-user.tpl.php, block.tpl.php, region--header.tpl.php, plus search-result templates (search-result.tpl.php, search-results.tpl.php) and a KMaps typeahead template (shanti-kmaps-typeahead.tpl.php) — the last one shared by every site with a KMaps field, i.e. all of them per Spike 1.
  • Shared JS: shanti-main.js, shanti-search.js, shanti-iframe.js, plus a large shared vendor set (Bootstrap, multilevelpushmenu, wookmark, jssor slider, mCustomScrollbar, progressive-image, hammer, icheck, bootstrap-select) — none of this is per-site.
  • Shared theme functions (shanti_sarvaka_* in template.php): breadcrumb, carousel, faceted search (facetapi_count), search-result preprocessing (preprocess_apachesolr_search_snippets), block/field/node/page preprocessing, a shared "get mandala home" helper, and a shared explore-menu renderer (menu_tree__shanti_explore_menu). This is the actual navigation, search UX, and cross-site "return to Mandala home" chrome — genuinely common, not duplicated per site.
  • Shared visual language: Bootstrap grid + vertical tabs, shanticon icon font, a common color-strip banner region, faceted search styling, collections-view CSS, all declared once in the base theme, inherited everywhere.

What's genuinely site-specific (do not converge these)

Each site's real divergence is narrow and maps directly to what the content type needs to be viewable, not to look-and-feel:

  • Images — the interactive deep-zoom viewer (already scoped in the interactive surfaces deferred note).
  • AV — Kaltura player integration, but this lives at the module layer (contrib kaltura + KalturaClient), not the theme — so it doesn't factor into a theme-baseline decision at all. It's tracked separately as Spike 7, currently ○ Pending.
  • Texts — reader-specific chrome (tabs) and dependency on footnote-adjacent modules; overlaps with the already-proven Spike 4b work, which is a content/markup concern, not a theme one.
  • Sources — citation/bibliography display (search-bibliography, biblio-full-page, fancytree styles) — relates directly to Spike 5's bibcite work. Notably, D7 itself had already disabled most of this (stylesheets commented out in the live .info file), so "faithful port" here may mean porting less than expected.
  • Visuals — depends on the shivanode module. Decided 2026-08-25 (team): Visuals is being retired, not migrated. No D11 content migration and no theme work needed for it. Retained here only as a record of what D7 actually shipped.

Recommendation

A shared D11 base theme is justified by evidence, not just intuition — D7 already built and ran one for eleven years across all six sites. The commonalities aren't incidental overlap to be discovered case-by-case as each site migrates; they're a single designed system (identical regions, identical page skeleton, identical JS/CSS foundation) that Phase 2's per-site owners would otherwise each partially rebuild in isolation.

Concretely:

  1. Port shanti_sarvaka's region layout and shared JS/CSS foundation as a D11 base theme before Phase 2 forks. This is squarely "faithful migration" per ADR 008/010 — it's porting what's already there, not introducing a new design.
  2. Each site's real add-on work is a module/content concern, not a theme concern — IIIF viewer (Images), Kaltura (AV, already its own spike), footnote rendering (Texts, already proven), bibcite display (Sources, already its own spike). None of these compete with or complicate a shared base theme; they layer on top of it exactly as they did in D7.
  3. Confirm whether sarvaka_projects is truly dead — a five-minute check (grep sites.php history / ask if any redirect still points at it) before anyone spends time auditing it further.

This audit does not itself decide when the base theme gets built (Sprint 2 vs. folded into Images' close-out) — that's the open scheduling question already raised with Yuji. What it settles is that the shared-baseline approach is sound: the D7 sites were never six independently-designed UIs, they were one theme with six thin, content-driven skins.