<?php
namespace Relay\Import;

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

/**
 * Reads SQL statements one at a time from a .sql file (spec §4.5), from any byte offset, so a dump of any size
 * is read over many requests. Understands '…', "…" and `…` (backslash escapes, doubled quotes), -- and #
 * line comments, block comments, MySQL's executable comments /*!40101 … *\/ (their content is kept, the
 * markers dropped; MariaDB's /*M! … *\/ are dropped whole) and DELIMITER lines. Comments are left out of the
 * statements it returns.
 */
final class SqlReader
{
    public const MAX_STATEMENT = 67108864;
    private const CHUNK = 1048576;

    /** @var resource */
    private $h;
    private string $buf = '';
    /** File offset of $buf[0]. */
    private int $base;
    private int $i = 0;
    private bool $eof = false;
    private string $delimiter;

    public function __construct(string $path, int $offset = 0, string $delimiter = ';')
    {
        $h = Fs::open($path, 'rb');
        if ($h === false) {
            throw new RelayError('import.unreadable', esc_html__('The uploaded SQL file could not be read.', 'siterelay'));
        }
        $this->h = $h;
        fseek($h, $offset);
        $this->base = $offset;
        $this->delimiter = $delimiter !== '' ? $delimiter : ';';
    }

    public function __destruct()
    {
        if (is_resource($this->h)) {
            Fs::close($this->h);
        }
    }

    /** The delimiter in force after the last statement (a DELIMITER line changes it). */
    public function delimiter(): string
    {
        return $this->delimiter;
    }

    /** @return array{sql: string, end: int}|null the next statement and the offset just after it; null at the end of the file */
    public function next(): ?array
    {
        $out = '';
        $conditional = false;
        $line_start = true;
        while (true) {
            if (!$this->ensure(64)) {
                $sql = trim($out);
                return $sql === '' ? null : ['sql' => $sql, 'end' => $this->base + $this->i];
            }
            if ($line_start && preg_match('/\G[ \t]*DELIMITER[ \t]+(\S+)[ \t]*(?:\r?\n|$)/i', $this->buf, $m, 0, $this->i) && trim($out) === '') {
                $this->delimiter = $m[1];
                $this->i += strlen($m[0]);
                continue;
            }
            $line_start = false;
            if (substr_compare($this->buf, $this->delimiter, $this->i, strlen($this->delimiter)) === 0) {
                $this->i += strlen($this->delimiter);
                $sql = trim($out);
                if ($sql !== '') {
                    return ['sql' => $sql, 'end' => $this->base + $this->i];
                }
                $out = '';
                $line_start = true;
                continue;
            }
            $c = $this->buf[$this->i];
            if ($c === "'" || $c === '"' || $c === '`') {
                $out .= $this->quoted($c);
            } elseif ($c === '#' || ($c === '-' && preg_match('/\G--(?:[ \t\r\n]|$)/', $this->buf, $m, 0, $this->i))) {
                $this->skip_to("\n");
                $out .= "\n";
                $line_start = true;
            } elseif ($c === '/' && ($this->buf[$this->i + 1] ?? '') === '*') {
                if (preg_match('/\G\/\*!\d*/', $this->buf, $m, 0, $this->i)) {
                    $this->i += strlen($m[0]);
                    $conditional = true;
                    $out .= ' ';
                } else {
                    $this->i += 2;
                    $this->skip_to('*/');
                    $out .= ' ';
                }
            } elseif ($conditional && $c === '*' && ($this->buf[$this->i + 1] ?? '') === '/') {
                $this->i += 2;
                $conditional = false;
                $out .= ' ';
            } elseif ($c === "\n") {
                $out .= "\n";
                $this->i++;
                $line_start = true;
            } else {
                $stop = "'\"`#-/\n" . $this->delimiter[0] . ($conditional ? '*' : '');
                $len = max(1, strcspn($this->buf, $stop, $this->i + 1) + 1);
                $out .= substr($this->buf, $this->i, $len);
                $this->i += $len;
            }
            if (strlen($out) > self::MAX_STATEMENT) {
                throw new RelayError('import.statement_too_large', esc_html__('The SQL file contains a statement larger than 64 MB, which SiteRelay will not load.', 'siterelay'));
            }
        }
    }

    /** A quoted string or identifier, kept as written. */
    private function quoted(string $q): string
    {
        $this->i++;
        $out = '';
        while (true) {
            if (!$this->ensure(2)) {
                throw new RelayError('import.unterminated', esc_html__('The SQL file ends inside a quoted value. It may be cut off.', 'siterelay'));
            }
            $len = strcspn($this->buf, $q === '`' ? '`' : $q . '\\', $this->i);
            if ($this->i + $len >= strlen($this->buf)) {
                $out .= substr($this->buf, $this->i);
                $this->i = strlen($this->buf);
                if (strlen($out) > self::MAX_STATEMENT) {
                    throw new RelayError('import.statement_too_large', esc_html__('The SQL file contains a statement larger than 64 MB, which SiteRelay will not load.', 'siterelay'));
                }
                continue;
            }
            $out .= substr($this->buf, $this->i, $len);
            $this->i += $len;
            if ($this->buf[$this->i] === '\\') {
                $this->ensure(2);
                $out .= substr($this->buf, $this->i, 2);
                $this->i += 2;
                continue;
            }
            // The quote: doubled means a literal quote character.
            $this->ensure(2);
            if (($this->buf[$this->i + 1] ?? '') === $q) {
                $out .= $q . $q;
                $this->i += 2;
                continue;
            }
            $this->i++;
            return $q . $out . $q;
        }
    }

    private function skip_to(string $end): void
    {
        while (true) {
            $at = strpos($this->buf, $end, $this->i);
            if ($at !== false) {
                $this->i = $at + strlen($end);
                return;
            }
            $this->i = max($this->i, strlen($this->buf) - strlen($end) + 1);
            if (!$this->ensure(strlen($end) + 1)) {
                $this->i = strlen($this->buf);
                return;
            }
        }
    }

    /** At least $n bytes from the current position, unless the file ends; false at the end of the file. */
    private function ensure(int $n): bool
    {
        while (strlen($this->buf) - $this->i < $n && !$this->eof) {
            $chunk = Fs::read($this->h, self::CHUNK);
            if ($chunk === false || $chunk === '') {
                $this->eof = true;
                break;
            }
            if ($this->i > 0) {
                $this->buf = substr($this->buf, $this->i);
                $this->base += $this->i;
                $this->i = 0;
            }
            $this->buf .= $chunk;
        }
        return $this->i < strlen($this->buf);
    }
}