<?php
// Wave 209 · /api/duplicates-registry.php
// Duplicate screens registry. Non-destructive audit: returns groups with
// canonical file + legacy candidates + recommendations. Zero delete.
@require_once __DIR__ . '/wevia-sanitizer-guard.php';
header('Content-Type: application/json; charset=utf-8');
header('Access-Control-Allow-Origin: *');

$DOCROOT = '/var/www/html';
$pages = array_filter(scandir($DOCROOT), function($f) use ($DOCROOT) {
    return substr($f, -5) === '.html' && is_file("$DOCROOT/$f");
});

// Normalize base name (strip version/legacy/saas suffixes)
$base_groups = [];
foreach ($pages as $p) {
    $base = substr($p, 0, -5);
    $base = preg_replace('/-v\d+$|-new$|-legacy$|-old$|\d{8,14}$|-saas$|-v\d+-\w+$/', '', $base);
    $base = preg_replace('/-pre-\w+$/', '', $base);
    $base_groups[$base][] = $p;
}

// Duplicates only
$dups = [];
$total_dups = 0;
foreach ($base_groups as $base => $files) {
    if (count($files) < 2) continue;

    // Determine canonical = file with cleanest name (shortest), fall back to newest
    usort($files, function($a, $b) use ($DOCROOT) {
        // Prefer files without -v / -legacy / -old / -saas suffixes
        $has_suffix_a = preg_match('/-v\d+|-legacy|-old|-saas|-new/', $a);
        $has_suffix_b = preg_match('/-v\d+|-legacy|-old|-saas|-new/', $b);
        if ($has_suffix_a !== $has_suffix_b) return $has_suffix_a - $has_suffix_b;
        // Then prefer shorter name
        if (strlen($a) !== strlen($b)) return strlen($a) - strlen($b);
        // Then newest mtime
        return filemtime("$DOCROOT/$b") - filemtime("$DOCROOT/$a");
    });

    $canonical = $files[0];
    $legacy = array_slice($files, 1);

    $entries = [];
    foreach ($files as $f) {
        $path = "$DOCROOT/$f";
        $size = filesize($path);
        $entries[] = [
            'file' => $f,
            'size_bytes' => $size,
            'last_modified' => date('c', filemtime($path)),
            'is_canonical' => $f === $canonical,
            'url' => "/$f"
        ];
    }

    // Size similarity between canonical and legacy
    $can_size = filesize("$DOCROOT/$canonical");
    $close_copies = [];
    foreach ($legacy as $l) {
        $l_size = filesize("$DOCROOT/$l");
        if ($can_size > 0 && abs($can_size - $l_size) / max($can_size, $l_size) < 0.1) {
            $close_copies[] = $l;
        }
    }

    $recommendation = count($close_copies) > 0
        ? 'Archive legacy copies (add HTTP 301 to canonical) - content quasi-identical'
        : 'Keep for now - legacy has distinct content, verify if still referenced';

    $dups[] = [
        'base' => $base,
        'count' => count($files),
        'canonical' => $canonical,
        'canonical_url' => "/$canonical",
        'legacy_files' => $legacy,
        'close_copies' => $close_copies,
        'recommendation' => $recommendation,
        'files' => $entries
    ];
    $total_dups += count($legacy);
}

// Sort by count desc
usort($dups, function($a, $b) { return $b['count'] - $a['count']; });

$out = [
    'ok' => true,
    'ts' => date('c'),
    'total_pages' => count($pages),
    'total_duplicate_groups' => count($dups),
    'total_legacy_files' => $total_dups,
    'total_close_copies' => array_sum(array_map(function($g){ return count($g['close_copies']); }, $dups)),
    'groups' => $dups,
    'version' => 'wave-209',
    'doctrine' => 'non-destructive · document canonical + legacy · zero delete'
];

echo json_encode($out, JSON_PRETTY_PRINT | JSON_UNESCAPED_UNICODE);