The répertoire page was loading the full v_theses_public view
(15 JOINs + 8 GROUP_CONCAT temp B-trees) via getAllPublishedTheses()
just to build the student name → thesis-id map on the index page.
Only two columns (id, authors) were ever consumed by the template.
Add Database::getPublishedAuthors(): array
- Queries thesis_authors JOIN authors directly on the theses base table
- Filters on theses.is_published = 1 using the existing index
- Returns only id + GROUP_CONCAT(authors) — no view expansion
- Results verified identical to the old getAllPublishedTheses() output
Update search.php to call getPublishedAuthors() instead.
Mark getAllPublishedTheses() @deprecated in Database.php.
All tests pass.
admin/index.php showed "TFE total / Publiés / En attente" by running
array_filter() over the already-filtered $theses array returned by
getThesesList(). When any search or year filter was active the three
numbers reflected only the matching subset, making the stats misleading
(e.g. searching for a single student would show "1 total, 0 publiés").
Add Database::getThesesStats(): array — a single SQL aggregation query:
SELECT COUNT(*), SUM(is_published), SUM(NOT is_published) FROM theses
This runs against the raw theses table with no filters, so the counters
always display the true whole-database figures regardless of what filter
the admin has active. admin/index.php now calls getThesesStats() and
reads $stats['total'], $stats['published'], $stats['pending'] instead
of the array_filter expressions.
- Add idx_theses_pub_year composite index on theses(is_published, year DESC) to
schema.sql; replaces the need for the query planner to pick between the two
separate idx_theses_published / idx_theses_year indexes and sort with a temp
B-tree. Every public query filters on is_published=1 and orders/filters by year,
so this covering index eliminates the sort pass for those queries.
- Create storage/migrations/006_add_composite_index.sql and apply to both
posterg.db and test.db.
- Fix storage/migrations/005_add_banner.sql: the view recreation in that file
still referenced the pre-migration-001 table/column names (thesis_keywords,
keywords.keyword). Updated to use thesis_tags / tags tg to match the canonical
schema.sql. The live DB was unaffected (migration 001 ran before 005), but the
file was misleading and would fail if ever re-run from scratch.
SQLite performance (Database::__construct):
- PRAGMA journal_mode = WAL: eliminates full-DB read locks on write, safe
for concurrent PHP-FPM workers
- PRAGMA synchronous = NORMAL: durable on commit without full fsync per write
- PRAGMA cache_size = -8000: ~8 MB page cache per connection
Accessibility foundation (WCAG 2.1 AA):
- common.css: add .sr-only utility, .skip-link (hidden until focused),
global :focus-visible (2px purple outline, 2px offset),
prefers-reduced-motion guard; remove bare outline:none from
.site-search__input
- admin.css: same :focus-visible, skip-link, and motion guard scoped to
admin purple; remove outline:none from .admin-input/.admin-select/
.admin-textarea and .admin-filters select (both had :focus border rules
already, so focus is still visually communicated)
- search.css: remove outline:none from .search-filter-select (already has
:focus border-color rule)
- All 5 public pages (index, search, tfe, apropos, licence): add
<a href="#main-content" class="skip-link"> as first child of <body>;
add id="main-content" to <main>
- templates/admin/head.php: same skip link; aria-label="Navigation admin"
on <nav>; id="main-content" on all 10 admin <main> elements
All 4 test suites pass (unit, integration, security, rate-limit).
SecurityTest::Test1 was calling $db->searchTheses($string) with a plain
string, but searchTheses() was refactored to require array $params when
the tag M2M work landed. This caused an immediate PHP fatal TypeError
before any SQL ever ran, killing the entire Security test suite with
exit code 255 and masking all three tests.
Fix: pass each malicious payload via ['query' => $string] which is the
correct API and properly exercises the parameterised query path through
validateSearchParams() + buildSearchConditions(). Added a clarifying
comment explaining why the array form is required.
All 4 test suites now pass:
- Database (Unit): 7/7
- Rate Limit (Unit): 5/5
- Search (Integration): 6/6
- Security: 3/3
- tests/Unit/DatabaseTest.php: tests 5-7 for findOrCreateTag round-trip, getUsedTags column, alias
- tests/Integration/SearchTest.php: tests 4-6 for tag subquery, full-text query, count consistency
- Database: getAllPublishedTheses() bypasses 100-row search cap for student index
- search.php: uses getAllPublishedTheses() for étudiantes column; all tests pass
