应该如何陈旧的指标在测试过程中如何处理？ [英] How should stale indexes be handled during testing?

问题描述

我在内存模式使用RavenDB的单元测试。我的查询是由静态指标支持。我不使用 WaitForNonStaleResults（） API（我也不想要）。

有关测试的典型工作流程

1. 初始化RavenDB在内存模式


2. 使用集成指数 IndexCreation.CreateIndexes（大会，IDocumentStore）


3. 插入测试数据（验证查询行为）


4. 运行查询


5. 验证查询输出

我注意到步骤1-3发生的如此之快，静态指标唐'T有时间在步骤4之前得到更新 - 因此指标是陈旧

我创建了一个快速的解决方法这一点。步骤3之后，我执行：

 而（！documentStore.DocumentDatabase.Statistics.StaleIndexes.Length = 0）
 Thread.sleep代码（10）; 
  

这感觉累赘。我想知道的是：

• 这是正常的索引在内存模式运行RavenDB当是过时


• 有没有一种更好的办法测试过程中避免旧索引吗？

解决方案

跨张贴这 RavenDB所在的用户组并有一个工作的解决方案。

这是正常的，在内存
模式下运行RavenDB当索引过时？

是的。索引是一个索引。

有没有更好的办法测试过程中避免旧索引吗？

是的。初始化文件存储时配置全局约定：

  VAR店=新EmbeddableDocumentStore（）; 
 store.RunInMemory = TRUE; 
 store.Conventions =新DocumentConvention 
 {
 DefaultQueryingConsistency = ConsistencyOptions.QueryYourWrites 
}; 
 
 store.Initialize（）; 
  

注意： ConsistencyOptions.QueryYourWrites 不与地图工作/缩小指标，即用减少=>指标; ... 部分。对于这些，你必须使用自定义（X => x.WaitForNonStale ...（））查询



<当点> 更新：有另一种方法，该可能会更好（没有亲自尝试过呢）。你可以实现IDocumentQueryListener强制所有的查询返回未失效的结果：

  VAR店=新EmbeddableDocumentStore {RunInMemory =真}; 
 store.Initialize（）; 
 
 store.RegisterListener（新ForceNonStaleQueryListener（））; 
 
公共类ForceNonStaleQueryListener：IDocumentQueryListener 
 {
公共无效BeforeQueryExecuted（IDocumentQueryCustomization定制）
 {
 queryCustomization.WaitForNonStaleResults（）; 
} 
} 
  

I am using RavenDB in In-Memory mode for unit testing. My queries are backed by static indexes. I am not using WaitForNonStaleResults() API (nor do I want to).

Typical workflow for a test is:

1. Initialise RavenDB in In-Memory mode
2. Integrate indexes using IndexCreation.CreateIndexes(Assembly, IDocumentStore)
3. Insert test data (for verifying query behaviour)
4. Run query
5. Verify query output

I have noticed steps 1-3 happen so quickly, that static indexes don't have time to get updated before step 4 - therefore the indexes are stale.

I have created a quick work-around for this. After step 3, I execute:

while (documentStore.DocumentDatabase.Statistics.StaleIndexes.Length != 0)
    Thread.Sleep(10);

This feels cumbersome. What I would like to know is:

• Is it normal for indexes to be stale when running RavenDB in In-Memory mode?
• Is there a better way to avoid stale indexes during testing?

解决方案

Cross-posted this to RavenDB usergroup and have a working solution.

Is it normal for indexes to be stale when running RavenDB in In-Memory mode?

Yes. An index is an index.

Is there a better way to avoid stale indexes during testing?

Yes. Configure global conventions when initialising document store:

var store = new EmbeddableDocumentStore();
store.RunInMemory = true;
store.Conventions = new DocumentConvention
{
    DefaultQueryingConsistency = ConsistencyOptions.QueryYourWrites
};

store.Initialize();

Note: ConsistencyOptions.QueryYourWrites doesn't work with Map/Reduce indexes, i.e. indexes with a Reduce => ... section. For these you have to use Customize(x => x.WaitForNonStale...()) when querying

Update: There is another approach, which may be better (haven't personally tried it yet). You could implement IDocumentQueryListener to force all queries to return non-stale results:

var store = new EmbeddableDocumentStore { RunInMemory = true };
store.Initialize();

store.RegisterListener(new ForceNonStaleQueryListener());

public class ForceNonStaleQueryListener : IDocumentQueryListener
{
    public void BeforeQueryExecuted(IDocumentQueryCustomization customization)
    {
        queryCustomization.WaitForNonStaleResults();
    }
}

这篇关于应该如何陈旧的指标在测试过程中如何处理？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！