mirror of
https://codeberg.org/PostERG/xamxam.git
synced 2026-05-06 11:09:18 +02:00
b981223ff4
Add system-fragment.php — a thin authenticated endpoint that returns only the tab-panel HTML (toolbar + meta + log/nginx-config output) for a given ?tab=&n= combination. No page shell, no status section, no DB queries. system.php changes: - Tab <a> elements gain data-tab= attributes used by JS to identify the target without parsing hrefs. - Tab panel content wrapped in <div id=sys-tab-panel data-tab= data-n=> which JS uses as both the swap target and its own state store. - JS rewritten: tab clicks and lines-select changes call loadPanel() which fetch()es system-fragment.php, swaps innerHTML, updates active tab ARIA attributes, and pushes state via history.pushState. - Browser back/forward handled via popstate listener. - bindPanelControls() re-wires the lines-select and copy-to-clipboard button after every innerHTML swap (event delegation not feasible here because log-output is replaced wholesale). - fetch() failure falls back to window.location.href (full page load). - Tabs without JS still work: <a href> links go to system.php?tab=… as before. system-fragment.php: - Requires AdminAuth::isAuthenticated(); returns 403 on failure. - Validates tab and n params against the same whitelist as system.php. - All helper functions namespaced with frag_ prefix to avoid redeclaration if PHP ever includes both files in the same process. - Renders identical HTML to the corresponding section in system.php. system.css: - #sys-tab-panel gets min-height:8rem and position:relative to prevent layout jump during fetch. - .sys-panel-loading: opacity 0.4 + pointer-events:none + subtle diagonal-stripe ::after overlay with shimmer animation.
2.0 KiB
2.0 KiB
System Page Caching - Database-Backed Status Cache
Problem
The admin system page (/admin/system.php) runs expensive operations on every load:
systemctlsubprocess calls (~4 checks × ~100ms each)curlHTTP self-check (~200-500ms)disk_total_space()/disk_free_space()(fast but unnecessary per-request)- Log file
tail+filesize+filemtime(I/O bound) - Nginx config file reading
Solution: system_cache table + background refresh
- Add
system_cachetable to schema:CREATE TABLE system_cache (key TEXT PRIMARY KEY, value TEXT NOT NULL, updated_at INTEGER NOT NULL) - Add migration
storage/migrations/007_system_cache.sql - Add
SystemCacheclass (src/SystemCache.php) with methods:get(string $key, int $maxAgeSec = 60): ?arrayset(string $key, array $data): voidisStale(string $key, int $maxAgeSec = 60): bool
- Refactor
system.phpstatus section:- Check
SystemCache::get('system_status', 120)— 2-minute TTL - If cache hit → render from cache, show “mis en cache il y a X sec” label
- If cache miss → run checks, store in cache, render
- Add
?refresh=1GET param to force-bypass cache
- Check
- Refactor
system.phplog sections — avoid re-reading on every tab switch; only read the active tab’s log - Cache disk info separately with 5-minute TTL:
SystemCache::get('disk_info', 300) - Cache PHP info separately with 1-hour TTL:
SystemCache::get('php_info', 3600)
Notes
- Log caching deliberately omitted:
tailoutput is inherently real-time and caching even 30s would show stale data during the moments it matters most (deploys, errors). The existing tab guard already ensures only the active log file is read. - nginx config could be cached but
file()on a small static config file is negligible; not worth the added complexity. - Log tab switching and line-count changes now use
fetch()viasystem-fragment.php; no full page reload on tab switch.