<?php
namespace Relay\Files;

/**
 * Walks the backup roots and appends one JSON line per file ({"g": group, "r": relative path, "s": size})
 * to a list file. Resumable: the cursor holds the pending directory stack, so a walk spans many requests.
 */
final class Scanner
{
    private const MAX_SKIPPED = 200;

    /** @var array<string, string> group => absolute root */
    private array $roots;
    /** @var array<string, Excludes> */
    private array $excludes;
    private string $list;

    public function __construct(array $roots, array $excludes, string $list_path)
    {
        $this->roots = $roots;
        $this->excludes = $excludes;
        $this->list = $list_path;
    }

    public static function start(): array
    {
        return ['group' => 0, 'stack' => null, 'list_bytes' => 0, 'files' => [], 'skipped' => [], 'skipped_count' => 0, 'done' => false];
    }

    /** Scan until $max_entries entries were looked at (a directory is always finished once started). */
    public function run(array $cursor, int $max_entries): array
    {
        $groups = array_keys($this->roots);
        $seen = 0;
        $lines = '';
        while (!$cursor['done'] && $seen < $max_entries) {
            if (!isset($groups[$cursor['group']])) {
                $cursor['done'] = true;
                break;
            }
            $group = $groups[$cursor['group']];
            $root = rtrim($this->roots[$group], '/');
            if ($cursor['stack'] === null) {
                $cursor['stack'] = is_dir($root) && !is_link($root) ? [''] : [];
                $cursor['files'][$group] = $cursor['files'][$group] ?? ['count' => 0, 'bytes' => 0];
            }
            if ($cursor['stack'] === []) {
                $cursor['group']++;
                $cursor['stack'] = null;
                continue;
            }
            $dir = array_pop($cursor['stack']);
            $path = $root . ($dir === '' ? '' : '/' . $dir);
            $names = is_dir($path) && is_readable($path) ? scandir($path) : false;
            if ($names === false) {
                $cursor = self::skip($cursor, $group . '/' . $dir, 'unreadable');
                continue;
            }
            $subdirs = [];
            foreach ($names as $name) {
                if ($name === '.' || $name === '..') {
                    continue;
                }
                $seen++;
                $rel = $dir === '' ? $name : $dir . '/' . $name;
                if ($this->excludes[$group]->excluded($rel)) {
                    continue;
                }
                $abs = $root . '/' . $rel;
                if (is_link($abs)) {
                    $cursor = self::skip($cursor, $group . '/' . $rel, 'symlink');
                } elseif (is_dir($abs)) {
                    $subdirs[] = $rel;
                } elseif (!is_file($abs) || !is_readable($abs)) {
                    $cursor = self::skip($cursor, $group . '/' . $rel, 'unreadable');
                } else {
                    $size = (int) filesize($abs);
                    $lines .= json_encode(['g' => $group, 'r' => $rel, 's' => $size], JSON_UNESCAPED_SLASHES | JSON_UNESCAPED_UNICODE | JSON_INVALID_UTF8_SUBSTITUTE) . "\n";
                    $cursor['files'][$group]['count']++;
                    $cursor['files'][$group]['bytes'] += $size;
                }
            }
            // Reverse so that popping the stack visits subdirectories in name order.
            foreach (array_reverse($subdirs) as $sub) {
                $cursor['stack'][] = $sub;
            }
        }
        if ($lines !== '') {
            Fs::put($this->list, $lines, true); // a list missing files would leave them out, or move them aside in a mirror
        }
        clearstatcache(true, $this->list);
        $cursor['list_bytes'] = is_file($this->list) ? (int) filesize($this->list) : 0;
        return $cursor;
    }

    private static function skip(array $cursor, string $path, string $reason): array
    {
        $cursor['skipped_count']++;
        if (count($cursor['skipped']) < self::MAX_SKIPPED) {
            $cursor['skipped'][] = ['path' => $path, 'reason' => $reason];
        }
        return $cursor;
    }
}