I am analyzing the TRACE data in the RTOS and have encountered the following problems when using it:

[dingzhou-beck]

Your question:

1. How to set the display format and time zone of the timestamp? Refer to Window > Preferces > Tracing > Time Format in trace-compass. Is setting the time zone supported in the current version?

2. How can the time interval between each data packet and the previous one be observed? This is very useful for studying the delay of data packets.

3. When using search to filter events in the event header, the matching data will be highlighted. Does it support displaying only the filtered data?

[bhufmann]

Thanks for interest in the VsCode trace viewer. Certain features of Trace Compass classic (in Eclipse) have not been migrated to VsCode trace viewer and using Trace Compass trace server.

Your question:

1. How to set the display format and time zone of the timestamp? Refer to Window > Preferces > Tracing > Time Format in trace-compass. Is setting the time zone supported in the current version?
   This has not been implemented in the VsCode trace viewer. The timestamps are either shown as long values (nanoseconds) or formatted with the default formatter of the trace server backend.

2. How can the time interval between each data packet and the previous one be observed? This is very useful for studying the delay of data packets.

This sounds like some custom analysis in the trace server back-end. This can be implemented as server extension using Java API where you have full control over how to analyse the trace events. You might want look into example latency analysis in Trace Compass. The pattern latency analysis using XML definition might also be interesting.

The developer guides for Trace Compass classic and Trace Compass trace server are a bit spread out. Here are some references:

• Trace Developer guide and Trace Server Focused guides: https://github.com/eclipse-tracecompass/org.eclipse.tracecompass/wiki#user-guides
• Data Driven analysis in XML (base): https://archive.eclipse.org/tracecompass/doc/stable/org.eclipse.tracecompass.doc.user/Data-driven-analysis.html#Data_driven_analysis
• XML pattern analysis: https://archive.eclipse.org/tracecompass/doc/stable/org.eclipse.tracecompass.doc.user/Data-driven-analysis.html#Defining_an_XML_pattern_provider

3. When using search to filter events in the event header, the matching data will be highlighted. Does it support displaying only the filtered data?

Filtering in the events table has not been implemented either in the VsCode trace viewer.

[dingzhou-beck]

Ok. Thank you for your answer!

I found that both questions 1 and 3 are supported in Trace-Compass. Will they be supported in VsCode trace viewer later? Or rather, if I want to implement these two functions, are there any reference documents or prompts?

Thank you very much!

[bhufmann]

Ok. Thank you for your answer!

I found that both questions 1 and 3 are supported in Trace-Compass. Will they be supported in VsCode trace viewer later? Or rather, if I want to implement these two functions, are there any reference documents or prompts?

Thank you very much!

Hi, yes, we have both in our backlog, but due to team capacity and priorities I don't have an indication when we will be able to implement that. So, you offer to help is very welcome.

For 1) it's a mixed item between FE and BE. Some data that is being displayed are already formatted in the backend and the frontend just shows the string. For example, when you see some formatted timestamp (e.g. Events table) they are formatted in the server backend. Some other timestimps are just transported as long values (nanoseconds) and the frontend either shows that number, or converts it to seconds decimal number (see the time axis). To have consistent formatting there needs to be a full stack update.

However, I would focus first on the FE to show all timestamps that are in long numbers can be formatted and updated. The second, step would be how can we synchronize with the server. I'm not fully clear about this yet. If you have any other suggestion, please feel free to bring that up.

About 3), the events we wrote an ADR at one point about adding this feature: https://github.com/eclipse-cdt-cloud/theia-trace-extension/blob/master/doc/adr/0006-table-filtering.md. This was to implement pretty much as it was done in classic Trace Compass. It would update the existing table.

However, things evolved from the time I wrote that. One idea I had was to create new data providers on the fly when a filter expression was entered. That data provider we can show under the events table data provider and the user can open it. In the FE it would be just a different virtual table. We could have multiple ones and they can be deleted from the UI. There are already APIs to have this children parent relationship for data providers (views) and to be able to delete.

To create such children, derived events table data providers we could use the new configurable data provider API of the TSP. We can have json schema to define the parameters to provide and re-use the recently added json editor for that. See here for more information about configurable data providers:

• https://github.com/eclipse-cdt-cloud/theia-trace-extension/blob/master/doc/adr/0011-tsp-analysis-api.md#configure-customizable-outputs
• https://github.com/eclipse-tracecompass/org.eclipse.tracecompass/blob/master/doc/trace-server/trace-server-dev-guide.md#implementing-a-configurable-data-provider

Please note that the implementing a configurable data provider for filtering support in the events table would mean it's a custom feature of the Trace Compass server, instead of being a TSP specified feature that each server needs to implement.

In any case, please let us know if you have any suggestions.

Please let us also know, if you start tackling one of the 2 items so that we can coordinate and not duplicate effort.

[dingzhou-beck]

好的。谢谢你的回答！
我发现问题 1 和问题 3 在 Trace-Compass 中都支持。请问 VsCode trace viewer 后续会支持吗？或者说，如果我想实现这两个功能，有什么参考文档或提示吗？
非常感谢！

您好，是的，我们两个项目都积压了，但由于团队能力和优先级问题，我暂时无法确定何时能够实施。所以，非常欢迎您提供帮助。

对于 1)，它是 FE 和 BE 的混合体。一些正在显示的数据已经在后端格式化，而前端只显示字符串。例如，当您看到一些格式化的时间戳（例如事件表）时，它们在服务器后端进行了格式化。其他一些时间戳只是以长整型值（纳秒）的形式传输，前端要么显示该数字，要么将其转换为以秒为单位的十进制数（参见时间轴）。为了保持格式一致，需要进行完整的堆栈更新。

不过，我首先会关注FE，以显示所有长数字的时间戳都可以格式化和更新。第二步是如何与服务器同步。我目前还不太清楚这一点。如果您有任何其他建议，请随时提出。

关于 3) 事件，我们曾写过一个关于添加此功能的 ADR：https://github.com/eclipse-cdt-cloud/theia-trace-extension/blob/master/doc/adr/0006-table-filtering.md。这个功能的实现方式与经典 Trace Compass 中的实现方式非常相似。它会更新现有的表。

然而，从我写下那篇文章开始，事情就发生了变化。我有一个想法，就是在输入过滤表达式时动态创建新的数据提供程序。我们可以将该数据提供程序显示在事件表数据提供程序下，用户可以打开它。在 FE 中，它只是一个不同的虚拟表。我们可以有多个数据提供程序，并且可以从 UI 中删除它们。目前已经有 API 可以实现数据提供程序（视图）的这种子父关系，并且能够删除它们。

要创建此类子事件表数据提供程序，我们可以使用 TSP 新增的可配置数据提供程序 API。我们可以使用 JSON 模式来定义要提供的参数，并复用最近添加的 JSON 编辑器。有关可配置数据提供程序的更多信息，请参见此处：

• https://github.com/eclipse-cdt-cloud/theia-trace-extension/blob/master/doc/adr/0011-tsp-analysis-api.md#configure-customizable-outputs
• https://github.com/eclipse-tracecompass/org.eclipse.tracecompass/blob/master/doc/trace-server/trace-server-dev-guide.md#implementing-a-configurable-data-provider

请注意，在事件表中实现可配置数据提供程序以支持过滤意味着它是 Trace Compass 服务器的自定义功能，而不是每个服务器都需要实现的 TSP 指定功能。

无论如何，如果您有任何建议，请告诉我们。

如果您开始处理这两项中的一项，也请告知我们，以便我们协调，避免重复努力。

Thank you for your hint! I will update if there is any progress.