通过REST客户端 - 服务器同步 [英] Client-server synchronization over REST

问题描述

这个问题是什么，我认为是在安卓/ iOS开发一个非常普遍的问题，但我还没有发现任何标准的解决方案呢。

This question is about what I think is a very common problem in Android / iOS development but I haven't found any "standard" solution yet.

让我们说我们有一个相当正常的REST API。服务器数据库包含（其中包括）表国家和城镇以1：N的关系

Let's say we have a fairly normal REST API. The server database contains (among others) the tables countries and towns with a 1:N relationship.

客户端（手机应用程序）要保持这两个表的本地快照。所以，当它的下线，就可以做到这一点就可以正常进行了REST查询，例如：得奥城镇人口的名单> = 100

The client (mobile app) wants to maintain a local snapshot of these two tables. So that when it's offline, it can do queries that would be normally done over REST, e.g.: "get a list of Austrian towns with population >= 100"?

如何处理呢？

第一个问题：一致性。客户端应该具有两个表的快照。如果城镇表的客户端下载更新和下线，一些镇可参照的国家，这不是在国家的本地副本表。

First problem: consistency. The client should have a snapshot of the two tables. If the client downloads updates of the towns table and goes offline, some town may reference a country that's not in the local copy of the countries table.

问题二：客户端只下载新的/删除/修改的行。沟休息和使用像一些自定义的RPC调用 get_updates_since（...）？

Second problem: the client should only download the new / deleted / changed rows. Ditch the REST and use some custom RPC call like get_updates_since(...)?

第三个问题：如何本地数据库更改（可能脱机使用）的客户端副本与服务器同步？自定义RPC调用？

Third problem: how should local changes to the client's copy of the database (possibly made offline) be synchronized with the server? Custom RPC calls?

推荐答案

我不认为有一个灵丹妙药，但你要寻找的模式是高速缓存。在过去的项目我已经从服务器到本地存储器（SQLite的或平面文件）复制的项目，并维持在约更新基于简单的规则/上传/吹扫的条目的元数据。这不是一个简单的问题，并最终被大量的code。

I don't think there's a silver bullet but the pattern you're looking for is caching. In past projects I have copied items from the server to local storage (SQLite or flat file), and maintained meta-data about the entries to update/upload/purge based on simple rules. It's not a simple problem and ends up being a lot of code.

一致性：在你的榜样，无论是确保您下载国表第一，或进行这两种操作原子 - 例如，使新表副本，并只替换缓存的版本，如果两个副本都成功完成（双打上存储）。

Consistency: in your example, either ensure you download the countries table first, or make the two operations atomic - e.g., make copies of the "new" tables and only replace your cached versions if both copies complete successfully (doubles up on storage).

只有下载新的/删除/修改：是的，这需要客户端/服务器集成 - 时间戳的所有记录（与格林尼治标准​​时间），要求从元数据服务器，然后步行通过当地的元数据决定做什么。 （在行提示的UUID是有益的）。

Only download new/deleted/changed: Yep, this requires client/server integration - timestamp all the records (with GMT), ask for metadata from the server, and walk through your local metadata deciding what to do. (Hint UUIDs on the rows are helpful).

同步本地更改：是的，更多的客户机/服务器集成。见上面的段落。

Synchronize local changes: Yes, more client/server integration. See above paragraph.

处理所有的异常和边缘的情况下是具有挑战性的。看看在问题的iCloud正在同步的coredata 。也许不是一个公平的比较，因为苹果正试图解决这个为一个完全分布式的数据库，但有趣的阅读不过。

Handling all the exceptions and edge cases is challenging. Take a look at the problems with iCloud sync'ing CoreData. Perhaps not a fair comparison, as Apple is trying to solve this for a totally distributed database, but interesting reading nonetheless.

这篇关于通过REST客户端 - 服务器同步的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！