Workspace management

[JohnCoene]

Labeling this as question, it came to mind yesterday.

By default blockr uses one workspace object which I believe will cause issues if we have multiple concurrent users as they will be referencing the same workspace environment. I think that's bound to leads to data races and other issues.

Each user should have their own workspace I think.

Is there an easy way to define that when the application starts up?

[nbenn]

In principle, I believe this can be handled with what we have currently: all "workspace manipulating" functions (should) have a workspace arg. What is expected there is an environment created as

blockr/R/workspace.R

Lines 43 to 48 in 061ac78

	workspace_env <- structure(
	new.env(parent = emptyenv()),
	title = "",
	settings = list(),
	class = "workspace"
	)

Happy to export a ctor for this, something like

new_workspace <- function(env = new.env(parent = emptyenv()), title = "", settings = list()) {
  structure(
    env,
    title = title,
    settings = settings,
    class = "workspace"
  )
}

And with that we'd be good to go (for now at least), no?

As discussed elsewhere, I believe the whole "workspace" approach taken should be reconsidered...

[nbenn]

The current result_field implementation is not compatible with "multiple" workspaces, as it goes and fetches stack results from the default. This will change with #432 however as it does not work as intended anyways.