objective-c - ARC 和重用指针会耗尽内存

Question

我认为在 ARC 实现之前我一直在考虑保留和释放，我倾向于尝试重用指针以避免一直重新声明它们。

但是对于 ARC，它会造成泄漏。我的猜测是它只是释放分配给“tmp”指针的最后一个对象，我假设当我给一个指针一个新对象时，旧对象会自动释放。将函数包装在 autoreleasepool 中似乎可以完成这项工作，但这真的是理想的解决方案还是我的编码风格有误？

这是一个有罪的函数，在我运行它几百次之后似乎把我所有的 ram 都吃光了:

- (void) saveJDictEntry:(JDictEntry *)entry {

  const char * query;
  char * errMsg;
  NSString *queryString;    
  sqlite3_stmt * statement;

  [self connect];   

  // Insert the word..
  queryString = [NSString stringWithFormat:@"INSERT INTO words (seq) VALUES (%i)", entry.number];
  query = [queryString UTF8String];
  if (sqlite3_exec(database, query, NULL, NULL, &errMsg) != SQLITE_OK) {
      NSLog(@"Statement failed %@ %@", errMsg, queryString);
  }


  // Get the last id we inserted..
  NSUInteger lastId = 0;
  queryString = @"SELECT last_insert_rowid()";
  query = [queryString UTF8String];
  sqlite3_prepare_v2(database, query, -1, &statement, NULL);    
  if (sqlite3_step(statement) == SQLITE_ROW) {
      lastId =  sqlite3_column_int(statement, 0);
  } else {
      NSLog(@"Failed to fetch last Id after saving");        
  }
  sqlite3_finalize(statement);    

  entry.databaseId = lastId;
  entry.saved = true;



  // Kana...
  queryString = [NSString stringWithFormat:@"INSERT INTO words_kana (word_id, kana) VALUES (%i, '%@')", entry.databaseId, entry.kana ];
  query = [queryString UTF8String];    
  if (sqlite3_exec(database, query, NULL, NULL, &errMsg) != SQLITE_OK) {
      NSLog(@"Statement failed %@ %@", errMsg, queryString);
  }        

  // Kanji ...
  for(NSString * _kanji in entry.kanji) {
      queryString = [NSString stringWithFormat:@"INSERT INTO words_kanji (word_id, kanji) VALUES (%i, '%@')", entry.databaseId, _kanji];
      query = [queryString UTF8String];    
      if (sqlite3_exec(database, query, NULL, NULL, &errMsg) != SQLITE_OK) {
          NSLog(@"Statement failed %@ %@", errMsg, queryString);
      }              
  }

  // Pos..
  for(NSString * _pos in entry.pos) {
      queryString = [NSString stringWithFormat:@"INSERT INTO words_pos (word_id, pos) VALUES (%i, '%@')", entry.databaseId, _pos];
      query = [queryString UTF8String];    
      if (sqlite3_exec(database, query, NULL, NULL, &errMsg) != SQLITE_OK) {
          NSLog(@"Statement failed %@ %@", errMsg, queryString);
      }              
  }

  // Gloss
  for(NSString * _gloss in entry.gloss) {
      NSString * glossField = [_gloss stringByReplacingOccurrencesOfString:@"'" withString:@"''"];            
      queryString = [NSString stringWithFormat:@"INSERT INTO words_gloss (word_id, gloss) VALUES (%i, '%@')", entry.databaseId, glossField];
      query = [queryString UTF8String];    
      if (sqlite3_exec(database, query, NULL, NULL, &errMsg) != SQLITE_OK) {
          NSLog(@"Statement failed %@ %@", errMsg, queryString);
      }              
  }   


}

Answer 1

旧的内存管理概念仍然适用，全局释放池中的自动释放实例在完成当前运行循环之前不会被释放，有或没有 ARC。

如果您正在编写非平凡的迭代，例如数千次迭代，用自动释放池包装它是可行的方法。你可能会使用

@autoreleasepool {
    // your code here
}

为了简化使用。使用自动释放池并没有错或不理想，因为出于某种原因它一直在那里。 ARC，顾名思义，只做自动引用计数。管理内存使用的任务仍然属于我们。