Variables, evaluation, and escaping about debug-adapter-protocol

[int19h]

I think the point is rather that the spec doesn't say that the string that's reported via value is also what's going to be passed to setVariable when user edits it in the IDE. True, the IDE is not really interpreting that string. But it's passing it from one place to another, and since it crosses the protocol boundary twice in doing so, it needs to be a part of the contract.

It's always going to be best effort, of course, because not all values have a representation that's suitable for re-evaluation. But if DAP implementations are failing to do so for something as basic as string literals, clearly it needs some nudging from the spec side.

[weinand]

DAP only deals with strings rendered in the UI. The DAP does not define any "primitive types" for data (variables).

The strings are shown as is.
There is no interpretation, evaluation, quoting or escaping performed in the frontend.
This gives the DA full control over what is shown to the user, but it is an obligation too: the DA has to format the value as the user would expected them to look like in the specific language.

The "value" argument in the SetVariableArguments is again just a string that is not interpreted by the frontend in any way. It is the DAs responsibility to interpret it in a way that makes sense for the given language/debugger.

The frontend does not assume anywhere that a "foo" is a string, 123 a number, and null a special value. The values are received from the DA as ""foo"" and "123" and "null".

[vscodebot]

This issue has been closed because it represents a question. Questions are better addressed on StackOverflow.

Happy Coding!

[hoehrmann]

@weinand : That is not an acceptable response. I asked that the specification be clarified. I would expect either a pointer to the part of the specification or accompanying documents where this is already explained, meaning I did not read them properly, -- or a rationale for why this does not need explaining.

[weinand]

Sorry the bot closed this issue.
I properly tagged it as a "Clarification" request.

[gregg-miskelly]

One minor note: someday we might want to extend the debug adapter protocol to add a new format or context that a client can use to obtain the 'editable value' of an expression/variable. VS has this concept within our internal debugging APIs so that backends can provide one value in the normal display scenario, but then a different value when the user double clicks to edit.