1-to-1 execution model is too rigid

[chipkent]

Plugin: 1.3.0 pre

The current version of the plugin is modeled off of the Deephaven IDE. As such, VS Code has one connection to one DH instance. This is basically like having an IDE and saying that it can only ever have one process running at a time. This is fine for simple cases, but multi-process / multi-query is where things often end up.

Consider JetBrains. When a job is run, it can be configured to run a specific script in a specific container or environment. This is much more flexible.

Imagine a case where a developer has two queries.

• Query 1: Has access to restricted data. Does some massaging of the data to produce less restricted output data.
• Query 2: Uses the less restricted output data to do stuff.

With the 1-to-1 model, we have to tell them to do all of query 1 and then reconnect and do all of query 2. Maybe this is fine. Maybe it becomes annoying in reality.

There are other limitations such as no way to display plots from two queries at once.

With the DH IDE, it is possible to spin up multiple servers on different ports and connect a different browser tab to each. Is there a similarly clean way to spawn multiple VS Code sessions pointed at different DH instances without it feeling clunky?

No matter what, there are tradeoffs with respect to 1-vscode-to-1-deephaven vs 1-vscode-to-N-deephaven. Have the relative tradeoffs been considered?

[vbabich]

We need to consider the UX on this. Ties in with #7

[bmingles]

@chipkent I think this will be addressed as part of DHC: Worker lifecycle management #7