<?php
/*************************************************************
# 076 License

Copyright (c) テクニカル諏訪子

Permission is hereby granted to any person obtaining a copy of the software
Little Beast (the "Software") to use, modify, merge, copy, publish, distribute,
sublicense, and/or sell copies of the Software, subject to the following conditions:

    1. **Origin Attribution**:
       - You must not misrepresent the origin of the Software; you must not claim
         you created the original Software.
       - If the Software is used in a product, you must either:
         a. Provide clear attribution in the product's documentation, user interface,
            or other visible areas, **OR**
         b. Pay the original developers a fee they specify in writing.
    2. **Usage Restriction**:
       - The Software, or any derivative works, dependencies, or libraries
         incorporating it, must not be used for censorship or to suppress freedom of
         speech, expression, or creativity. Prohibited uses include, but are not
         limited to:
         - Censorship of so-called "hate speech", visuals, non-mainstream opinions,
           ideas, or objective reality.
         - Tools or systems designed to restrict access to information or
           artistic works.
    3. **Notice Preservation**:
       - This license and the above copyright notice must remain intact in all copies
         of the source code.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED,
INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A
PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT
HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF
CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE
OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 */
namespace Std\Lib\Git;

/**
 * 受け取ったパックを読んで保存する（git index-pack、git unpack-objects）
 *
 * ストリームから読みながら一時ファイルに書き、差分を解決してIDを計算し、.idx（v2）と.revを作る。
 * オブジェクトが少なければルーズオブジェクトにする。
 */
final class PackIndexer {
  /** @var resource */
  private $in;
  /** @var resource */
  private $tmp;
  private string $buf = '';
  private $hash;
  private int $pos = 0;
  private string $tmpPath = '';
  /** @var array オフセット => [種類, 大きさ, データの位置, 元のオフセット, 元のID(bin), crc] */
  private array $entries = [];
  /** @var array オフセット => ID(bin) */
  private array $ids = [];
  /** @var array ID(hex) => オフセット */
  private array $byId = [];
  /** @var array オフセット => 種類 */
  private array $types = [];
  private array $cache = [];
  private int $cacheBytes = 0;
  private ?Pack $pack = null;
  /** 参照しているID（接続の確認用） */
  private array $links = [];
  private bool $thin = false;

  public function __construct(private Repository $repo) {}

  /**
   * パックを受け取る
   *
   * @param resource $in  入力（パックの始まりから）
   * @param int $unpackLimit  これより少なければルーズオブジェクトにする
   * @param callable|null $progress  fn(string $what, int $done, int $total)
   * @return array  objects（ID => 種類）、links（参照しているID => true）、pack（パックのパス、ルーズならnull）
   */
  public function receive($in, int $unpackLimit = 100, ?callable $progress = null): array {
    $this->in = $in;
    $this->hash = hash_init($this->repo->hashAlgo);
    $dir = $this->repo->objects->dir().'/pack';
    $this->repo->mkdir($dir);
    $this->tmpPath = $dir.'/tmp_pack_'.bin2hex(random_bytes(6));
    $tmp = @fopen($this->tmpPath, 'w+b');
    if ($tmp === false) throw new GitException("一時ファイルを作れません：{$this->tmpPath}");
    $this->tmp = $tmp;
    try {
      $head = $this->read(12);
      if (substr($head, 0, 4) !== 'PACK' || !in_array(unpack('N', $head, 4)[1], [2, 3], true)) throw new GitException('パックではありません');
      $count = unpack('N', $head, 8)[1];
      for ($i = 0; $i < $count; $i++) {
        $this->readEntry();
        if ($progress && (($i + 1) % 1000 === 0 || $i + 1 === $count)) $progress('受信', $i + 1, $count);
      }
      $sum = hash_final($this->hash, true);
      $trailer = $this->read($this->repo->hashLen, false);
      if ($trailer !== $sum) throw new GitException('パックのチェックサムが違います');
      fwrite($this->tmp, $trailer);
      fflush($this->tmp);
      $this->resolveAll($progress);
      $objects = [];
      foreach ($this->ids as $off => $bin) $objects[bin2hex($bin)] = $this->types[$off];
      // 中に無い物を参照しているか（相手が持っている物はOK、無い物は接続の確認で）
      foreach (array_keys($this->links) as $l) if (isset($objects[$l])) unset($this->links[$l]);
      if ($count < $unpackLimit || $this->thin) {
        foreach (array_keys($this->entries) as $off) {
          [$type, $data] = $this->resolve($off);
          $this->repo->objects->write($type, $data);
        }
        $path = null;
      } else {
        $path = $this->install($sum);
      }
      return ['objects' => $objects, 'links' => $this->links, 'pack' => $path];
    } finally {
      $this->pack = null;
      if (is_resource($this->tmp)) fclose($this->tmp);
      @unlink($this->tmpPath);
    }
  }

  /** 入力からnバイト（ハッシュと一時ファイルにも） */
  private function read(int $n, bool $hash = true): string {
    while (strlen($this->buf) < $n) {
      $s = fread($this->in, max(65536, $n - strlen($this->buf)));
      if ($s === false || $s === '') {
        if (feof($this->in)) throw new GitException('パックが途中で終わりました');
        continue;
      }
      $this->buf .= $s;
    }
    $out = substr($this->buf, 0, $n);
    $this->buf = (string)substr($this->buf, $n);
    if ($hash) {
      hash_update($this->hash, $out);
      fwrite($this->tmp, $out);
      $this->pos += $n;
    }
    return $out;
  }

  private function readEntry(): void {
    $offset = $this->pos;
    $crc = hash_init('crc32b');
    $b = ord($h = $this->read(1));
    $raw = $h;
    $type = ($b >> 4) & 7;
    $size = $b & 15;
    $shift = 4;
    while ($b & 0x80) {
      $b = ord($h = $this->read(1));
      $raw .= $h;
      $size |= ($b & 0x7f) << $shift;
      $shift += 7;
    }
    $baseOff = null;
    $baseId = null;
    if ($type === Pack::OFS_DELTA) {
      $b = ord($h = $this->read(1));
      $raw .= $h;
      $n = $b & 0x7f;
      while ($b & 0x80) {
        $b = ord($h = $this->read(1));
        $raw .= $h;
        $n = (($n + 1) << 7) | ($b & 0x7f);
      }
      $baseOff = $offset - $n;
      if (!isset($this->entries[$baseOff])) throw new GitException('壊れたパック（OFS_DELTAの元）');
    } else if ($type === Pack::REF_DELTA) {
      $baseId = $this->read($this->repo->hashLen);
      $raw .= $baseId;
    } else if (!isset(Pack::TYPES[$type])) {
      throw new GitException("壊れたパック（種類{$type}）");
    }
    hash_update($crc, $raw);
    $dataAt = $this->pos;
    // zlibのデータ（長さは展開しないと分からない）
    $ctx = inflate_init(ZLIB_ENCODING_DEFLATE);
    $out = '';
    $readBefore = 0;
    while (true) {
      if ($this->buf === '') {
        $s = fread($this->in, 65536);
        if ($s === false || $s === '') {
          if (feof($this->in)) throw new GitException('パックが途中で終わりました');
          continue;
        }
        $this->buf = $s;
      }
      $res = @inflate_add($ctx, $this->buf, ZLIB_SYNC_FLUSH);
      if ($res === false) throw new GitException('壊れたパック（zlib）');
      $out .= $res;
      if (strlen($out) > $size) throw new GitException('壊れたパック（大きさ）');
      $total = inflate_get_read_len($ctx);
      $used = $total - $readBefore;
      $readBefore = $total;
      $chunk = substr($this->buf, 0, $used);
      $this->buf = (string)substr($this->buf, $used);
      hash_update($this->hash, $chunk);
      hash_update($crc, $chunk);
      fwrite($this->tmp, $chunk);
      $this->pos += $used;
      if (inflate_get_status($ctx) === ZLIB_STREAM_END) break;
    }
    if (strlen($out) !== $size) throw new GitException('壊れたパック（大きさ）');
    $this->entries[$offset] = [$type, $size, $dataAt, $baseOff, $baseId, unpack('N', hash_final($crc, true))[1]];
    if (isset(Pack::TYPES[$type])) $this->found($offset, Pack::TYPES[$type], $out);
  }

  /** IDが分かったオブジェクト */
  private function found(int $off, string $type, string $data): void {
    $id = $this->repo->hashObject($type, $data);
    $this->ids[$off] = hex2bin($id);
    $this->byId[$id] = $off;
    $this->types[$off] = $type;
    $this->check($id, $type, $data);
  }

  /** 中身の確認と、参照しているID */
  private function check(string $id, string $type, string $data): void {
    $repo = $this->repo;
    try {
      if ($type === 'commit') {
        $c = Commit::parse($data);
        if (!$repo->isId($c->tree) || $c->author === null || $c->committer === null) throw new GitException("壊れたコミット：{$id}");
        $this->links[$c->tree] = true;
        foreach ($c->parents as $p) {
          if (!$repo->isId($p)) throw new GitException("壊れたコミット：{$id}");
          $this->links[$p] = true;
        }
      } else if ($type === 'tree') {
        foreach (Tree::parse($data, $repo->hashLen)->entries as [$mode, $name, $eid]) {
          if (!in_array($mode, [Tree::FILE, Tree::EXEC, Tree::LINK, Tree::DIR, Tree::SUBMODULE, '100664'], true) || $name === '' || $name === '.' || $name === '..' || str_contains($name, '/')
              || strcasecmp($name, '.git') === 0) {
            throw new GitException("不正なツリー：{$id}（{$name}）");
          }
          if ($mode !== Tree::SUBMODULE) $this->links[$eid] = true;
        }
      } else if ($type === 'tag') {
        $t = Tag::parse($data);
        if (!$repo->isId($t->object)) throw new GitException("壊れたタグ：{$id}");
        $this->links[$t->object] = true;
      }
    } catch (GitException $e) {
      throw $e;
    }
  }

  /** 差分を全部解決する */
  private function resolveAll(?callable $progress): void {
    $this->pack = new Pack($this->tmpPath, $this->repo->hashLen, false);
    $pending = array_keys(array_filter($this->entries, fn($e) => !isset(Pack::TYPES[$e[0]])));
    $total = count($pending);
    while (!empty($pending)) {
      $left = [];
      foreach ($pending as $off) {
        if (isset($this->ids[$off])) continue;
        if ($this->resolve($off, true) === null) $left[] = $off;
      }
      if (count($left) === count($pending)) {
        // 元がパックに無い（シンパック）：リポジトリの物を使う
        foreach ($left as $off) {
          $bid = $this->entries[$off][4];
          if ($bid === null || !$this->repo->objects->exists(bin2hex($bid))) throw new GitException('差分の元が有りません：'.($bid === null ? '?' : bin2hex($bid)));
          $this->thin = true;
        }
        foreach ($left as $off) $this->resolve($off);
        break;
      }
      $pending = $left;
      if ($progress) $progress('差分の解決', $total - count($pending), $total);
    }
  }

  /**
   * オフセットのオブジェクト
   *
   * @param bool $inPackOnly  元が（まだ）分からなければnull
   * @return array|null  [種類, データ]
   */
  private function resolve(int $off, bool $inPackOnly = false): ?array {
    if (isset($this->cache[$off])) return $this->cache[$off];
    [$type, $size, $dataAt, $baseOff, $baseId] = $this->entries[$off];
    if (isset(Pack::TYPES[$type])) {
      $obj = [Pack::TYPES[$type], $this->pack->inflate($dataAt, $size)];
    } else {
      if ($baseOff !== null) {
        $base = $this->resolve($baseOff, $inPackOnly);
      } else {
        $hex = bin2hex($baseId);
        if (isset($this->byId[$hex])) $base = $this->resolve($this->byId[$hex], $inPackOnly);
        else if ($inPackOnly) return null;
        else $base = $this->repo->objects->read($hex);
      }
      if ($base === null) return null;
      $obj = [$base[0], Delta::apply($base[1], $this->pack->inflate($dataAt, $size))];
      if (!isset($this->ids[$off])) $this->found($off, $obj[0], $obj[1]);
    }
    $n = strlen($obj[1]);
    if ($n < 8388608) {
      while ($this->cacheBytes + $n > 67108864 && !empty($this->cache)) {
        $k = array_key_first($this->cache);
        $this->cacheBytes -= strlen($this->cache[$k][1]);
        unset($this->cache[$k]);
      }
      $this->cache[$off] = $obj;
      $this->cacheBytes += $n;
    }
    return $obj;
  }

  /** パックと.idx・.revを置く */
  private function install(string $sum): string {
    $h = $this->repo->hashLen;
    $sorted = $this->ids;
    asort($sorted, SORT_STRING);
    $fan = array_fill(0, 256, 0);
    foreach ($sorted as $bin) $fan[ord($bin[0])]++;
    $idx = "\xfftOc".pack('N', 2);
    $acc = 0;
    for ($i = 0; $i < 256; $i++) {
      $acc += $fan[$i];
      $idx .= pack('N', $acc);
    }
    $names = '';
    $crcs = '';
    $offs = '';
    $large = '';
    $nLarge = 0;
    $position = [];
    $n = 0;
    foreach ($sorted as $off => $bin) {
      $names .= $bin;
      $crcs .= pack('N', $this->entries[$off][5]);
      if ($off < 0x80000000) {
        $offs .= pack('N', $off);
      } else {
        $offs .= pack('N', 0x80000000 | $nLarge++);
        $large .= pack('J', $off);
      }
      $position[$off] = $n++;
    }
    $idx .= $names.$crcs.$offs.$large.$sum;
    $idx .= hash($this->repo->hashAlgo, $idx, true);
    ksort($position);
    $rev = 'RIDX'.pack('NN', 1, $this->repo->hashAlgo === 'sha1' ? 1 : 2).pack('N*', ...array_values($position)).$sum;
    $rev .= hash($this->repo->hashAlgo, $rev, true);

    $base = $this->repo->objects->dir().'/pack/pack-'.bin2hex($sum);
    $this->pack = null;
    fclose($this->tmp);
    if (!is_file($base.'.pack')) {
      if (!@rename($this->tmpPath, $base.'.pack')) throw new GitException('パックを保存出来ません');
      $this->repo->chmodFile($base.'.pack', true);
    }
    $this->repo->writeFile($base.'.rev', $rev, true);
    $this->repo->writeFile($base.'.idx', $idx, true);
    $this->repo->objects->packs(true);
    return $base.'.pack';
  }
}