<?php
namespace Relay\Import;

use Relay\Files\Fs;
use Relay\Jobs\Budget;
use Relay\Support\RelayError;

/**
 * A first pass over an uploaded dump: which tables it creates, which databases it switches to and, for every
 * *options table, its siteurl and home. From that (or from Relay's own export header) site() works out the
 * WordPress site in the file — its table prefix, URLs and path — over as many requests as the pass takes.
 */
final class DumpScanner
{
    public static function start(): array
    {
        return ['offset' => 0, 'delimiter' => ';', 'tables' => [], 'options' => [], 'databases' => [], 'statements' => 0, 'done' => false];
    }

    /** SiteRelay's export header (first lines of the file): site and home URL, path, table prefix. */
    public static function header(string $path): array
    {
        $h = Fs::open($path, 'rb');
        $head = $h !== false ? (string) Fs::read($h, 8192) : '';
        if ($h !== false) {
            Fs::close($h);
        }
        if (strpos($head, '-- SiteRelay SQL export') !== 0 && strpos($head, '-- Relay SQL export') !== 0) { // the second: before the plugin was called SiteRelay
            return [];
        }
        $out = [];
        $labels = [
            'Site URL' => 'site_url', 'Home URL' => 'home_url', 'Path' => 'path', 'Table prefix' => 'prefix',
            'Uploads URL' => 'uploads_url', 'Uploads path' => 'uploads_dir', 'Content URL' => 'content_url',
            'Main site uploads URL' => 'alt_uploads_url', 'Main site content URL' => 'alt_content_url',
        ];
        foreach ($labels as $label => $key) {
            if (preg_match('/^-- ' . $label . ': (.+)$/m', $head, $m)) {
                $out[$key] = trim($m[1]);
            }
        }
        return $out;
    }

    public static function run(string $path, array $cursor, Budget $budget): array
    {
        $reader = new SqlReader($path, (int) $cursor['offset'], (string) $cursor['delimiter']);
        while (!$budget->exhausted()) {
            $s = $reader->next();
            if ($s === null) {
                $cursor['done'] = true;
                break;
            }
            $sql = $s['sql'];
            if (preg_match('/^CREATE\s+TABLE\s+(?:IF\s+NOT\s+EXISTS\s+)?`((?:[^`]|``)+)`/i', $sql, $m)) {
                $t = str_replace('``', '`', $m[1]);
                if (!in_array($t, $cursor['tables'], true)) {
                    $cursor['tables'][] = $t;
                }
            } elseif (preg_match('/^USE\s+(?:`((?:[^`]|``)+)`|(\S+))\s*$/i', $sql, $m)) {
                $db = $m[1] !== '' ? str_replace('``', '`', $m[1]) : $m[2];
                if (!in_array($db, $cursor['databases'] ?? [], true)) {
                    $cursor['databases'][] = $db;
                }
            } elseif (preg_match('/^(?:INSERT|REPLACE)\s+(?:IGNORE\s+)?INTO\s+`((?:[^`]|``)+options)`/i', $sql, $m)) {
                $t = str_replace('``', '`', $m[1]);
                foreach (['siteurl', 'home'] as $name) {
                    // Only an address: with an unusual column order the value after the name is something else.
                    if (!isset($cursor['options'][$t][$name]) && preg_match("/'" . $name . "'\\s*,\\s*'(https?:(?:[^'\\\\]|\\\\.|'')*)'/i", $sql, $v)) {
                        $cursor['options'][$t][$name] = self::unescape($v[1]);
                    }
                }
            }
            $cursor['statements']++;
            $cursor['offset'] = $s['end'];
            $cursor['delimiter'] = $reader->delimiter();
            $budget->consume(strlen($sql));
        }
        return $cursor;
    }

    /**
     * The WordPress site in the dump. Throws when there is none, when it is a multisite network, or when the dump
     * holds several databases (mysqldump --all-databases): their tables would land on top of each other.
     *
     * @return array{prefix: string, site_url: string, home_url: string, path: string, tables: string[]}
     */
    public static function site(array $cursor, array $header): array
    {
        $databases = $cursor['databases'] ?? [];
        if (count($databases) > 1) {
            throw new RelayError(
                'import.several_databases',
                /* translators: %s: database names */
                esc_html(sprintf(__('This SQL file holds several databases (%s).', 'siterelay'), implode(', ', array_slice($databases, 0, 5)))),
                esc_html__('Export only the WordPress site\'s database and import that file.', 'siterelay')
            );
        }
        $tables = $cursor['tables'];
        $candidates = [];
        foreach ($tables as $t) {
            if (substr($t, -7) === 'options') {
                $p = substr($t, 0, -7);
                if (in_array($p . 'posts', $tables, true)) {
                    $candidates[] = $p;
                }
            }
        }
        $prefix = null;
        if (isset($header['prefix']) && in_array($header['prefix'], $candidates, true)) {
            $prefix = $header['prefix'];
        } elseif ($candidates !== []) {
            // The site whose URL is known first, then the shortest prefix (wp_ before wp_2_).
            usort($candidates, static function ($a, $b) use ($cursor) {
                $ka = isset($cursor['options'][$a . 'options']) ? 0 : 1;
                $kb = isset($cursor['options'][$b . 'options']) ? 0 : 1;
                return $ka <=> $kb ?: strlen($a) <=> strlen($b) ?: strcmp($a, $b);
            });
            $prefix = $candidates[0];
        }
        if ($prefix === null) {
            throw new RelayError('import.no_wordpress', esc_html__('This SQL file has no WordPress tables (an options and a posts table with the same prefix).', 'siterelay'), esc_html__('Choose a database export of a WordPress site.', 'siterelay'));
        }
        if (in_array($prefix . 'blogs', $tables, true) || in_array($prefix . 'site', $tables, true)) {
            throw new RelayError('import.multisite', esc_html__('This SQL file holds a whole multisite network, which SiteRelay can\'t import.', 'siterelay'), esc_html__('Export the one subsite you need from that network, and import that file.', 'siterelay'));
        }
        $options = $cursor['options'][$prefix . 'options'] ?? [];
        $home = (string) ($header['home_url'] ?? ($options['home'] ?? ($options['siteurl'] ?? '')));
        $extra = array_intersect_key($header, array_flip(['uploads_url', 'uploads_dir', 'content_url', 'alt_uploads_url', 'alt_content_url']));
        return [
            'prefix' => $prefix,
            'site_url' => (string) ($header['site_url'] ?? ($options['siteurl'] ?? $home)),
            'home_url' => $home,
            'path' => (string) ($header['path'] ?? ''),
            'tables' => array_values(array_filter($tables, static fn ($t) => strpos($t, $prefix) === 0)),
        ] + $extra;
    }

    /** MySQL string escapes in a quoted value. */
    private static function unescape(string $v): string
    {
        return strtr(str_replace("''", "'", $v), ['\\\\' => '\\', "\\'" => "'", '\\"' => '"', '\\n' => "\n", '\\r' => "\r", '\\t' => "\t", '\\0' => "\0", '\\/' => '/']);
    }
}