Core Data结合CloudKit去重后Book仅保留首个关联Author问题

DevOps 小满

2026-8-28

问题根因

核心bug出在对Core Data双向关系的实时可变集合做遍历时直接修改集合，导致枚举过程跳过元素，最终丢失作者关联。

Core Data的双向关系会自动同步两边的变更：当你从Author.books集合里删掉待删除的重复Book时，Core Data会自动把这个Author从反向的Book.authors集合里移除。你现在的代码直接遍历book.authors这个会实时变动的可变集合，实际执行流程如下：

某本待删除的重复Book原本关联了3个作者，集合为[authorA, authorB, authorC]
for-in遍历第一个元素authorA，执行关联转移：把authorA的books里的重复Book删掉，加上保留的winner Book。这一步触发双向同步，authorA被自动从当前重复Book的authors集合里移除，集合变成[authorB, authorC]
for-in的内部索引往后移一位，原本应该取索引1位置的authorB，但此时索引1对应的元素已经变成authorC，authorB被直接跳过，完全没有执行关联到winner Book的逻辑
最终只有第一个被遍历到的作者能正确关联到保留的Book，其余作者的关联全部丢失。

这个遍历实时集合的问题在作者去重、出版社去重函数里也存在，只是因为Book和出版社是多对一关系，单本书仅关联一个出版社，不会出现跳元素的情况，所以问题没有暴露。

修复方案

所有遍历Core Data关系集合的场景，先把集合转为不可变的本地快照再遍历，不要直接遍历实时更新的关系集合，避免枚举过程中集合变动导致元素被跳过。
修正后的代码如下：

private func remove(duplicatedBooks: [Book], winner: Book, performingContext: NSManagedObjectContext) {
    for book in duplicatedBooks {
        // 拷贝当前关联为不可变快照，避免遍历过程中集合被双向同步修改
        guard let authors = book.authors?.allObjects as? [Author] else {
            // 无作者关联时直接处理出版社
            if let publisher = book.publisher,
               let publisherBooks = publisher.books as? NSMutableSet {
                publisherBooks.remove(book)
                if !publisherBooks.contains(winner) {
                    publisherBooks.add(winner)
                }
            }
            performingContext.delete(book)
            continue
        }
        for author in authors {
            guard let authorBooks = author.books as? NSMutableSet else { continue }
            if authorBooks.contains(book) {
                authorBooks.remove(book)
                // 避免重复添加已存在的关联，减少无意义的变更
                if !authorBooks.contains(winner) {
                    authorBooks.add(winner)
                }
            }
        }
        // 处理出版社关联
        if let publisher = book.publisher,
           let publisherBooks = publisher.books as? NSMutableSet {
            if publisherBooks.contains(book) {
                publisherBooks.remove(book)
                if !publisherBooks.contains(winner) {
                    publisherBooks.add(winner)
                }
            }
        }
        performingContext.delete(book)
    }
}

private func remove(duplicatedAuthors: [Author], winner: Author, performingContext: NSManagedObjectContext) {
    for author in duplicatedAuthors {
        // 拷贝关联快照
        guard let books = author.books?.allObjects as? [Book] else {
            performingContext.delete(author)
            continue
        }
        for book in books {
            guard let bookAuthors = book.authors as? NSMutableSet else { continue }
            if bookAuthors.contains(author) {
                bookAuthors.remove(author)
                if !bookAuthors.contains(winner) {
                    bookAuthors.add(winner)
                }
            }
        }
        performingContext.delete(author)
    }
}

private func remove(duplicatedPublishers: [Pub], winner: Pub, performingContext: NSManagedObjectContext) {
    for pub in duplicatedPublishers {
        // 拷贝关联快照
        guard let books = pub.books?.allObjects as? [Book] else {
            performingContext.delete(pub)
            continue
        }
        for book in books {
            book.publisher = winner
        }
        performingContext.delete(pub)
    }
}

补充说明：

用allObjects将NSSet转为Swift数组，得到的是集合当前状态的不可变副本，遍历过程中不会被Core Data的双向同步逻辑修改，彻底避免枚举跳元素的问题
新增了关联存在性判断，避免向集合中添加已经存在的对象，减少不必要的NSManagedObjectContext变更通知，提升性能
Book与Publisher是多对一关系，直接给book.publisher赋值时，Core Data会自动维护反向的publisher.books集合，不需要手动操作集合，原有逻辑本身正确，只需要补全集合快照拷贝即可。

内容的提问来源于stack exchange，提问作者wildcard