Scala intellisense in Scala paragraphs + general latency issues

[zzhu8192]

Really fantastic project! There's not a whole lot of support for Zeppelin out there these days, so it's really great to see someone working on this

A couple of things I'm seeing:

• For Scala based paragraphs, even when I select "Scala" as the Language at the lower right, intellisense doesn't seem to work. Not sure if that's the intent? (I do have Metals installed, and on Scala projects, intellisense does work). This works for Python based paragraphs, which is really nice.

• In general, the response gets back to VSCode on the average about 5 seconds after running the paragraph. I can see that on the Zeppelin server side, the result was rendered in well under 1 second. Why is there such a long delay? I'm running zeppelin locally so there's little to no network latency at all.

[allen-li1231]

Hi, @zzhu8192

• For Scala based paragraphs, even when I select "Scala" as the Language at the lower right, intellisense doesn't seem to work. Not sure if that's the intent?

From my end, the Scala intellisense looks just fine, I tried some example code:

Would you like to be more specific to the problem you've encountered?

• In general, the response gets back to VSCode on the average about 5 seconds after running the paragraph.

I did notice the same 5 seconds delay from my end. In short, the extension is built upon Zeppelin REST API. The delay is not caused by extension, It is the server side that blocks the request when calling paragraph execution asynchronously.
However, the latency seems to go away when I tested the synchronous execution API to execute paragraphs. The extension applies the asynchronous one by default, you may switch between the two APIs in settings:

[zzhu8192]

Thanks for the response @allen-li1231 !
I tried the sequential access setting, and that works great!

In terms of the scala intellisense, here's a screenshot of what I'm seeing.
Note that, Zeppelin by default doesn't support %scala paragraph, so I'm using %spark, which is supported by default.
But the intellisense I'm getting is not the same that I'd get in Metals on a proper SBT/Maven project.

[allen-li1231]

Confirmed the discrepancy. I will look into that, by far I cannot tell whether the extension is responsible for this, it is possible that the language server simply does not work with codes in cell, thanks for the screenshot!

[allen-li1231]

It seems that metals support for notebook has bee halted, related PR: scalameta/metals#4253
So the metals currently do not officially support notebook, and the team is putting this in feature request: scalameta/metals#707
I am currently looking for workaround.