%sessioninfo magic should add versions to depended packages

[gaow]

Eg, if I have a workflow step in a notebook:

[flash]
depends: R_library("stephenslab/flash")

Then it should not be too much to ask %sessioninfo to display an R session summary with version of flash package, eg:

> library(flashr)
> sessionInfo()
attached base packages:
[1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
[1] flashr_0.6-1         RevoUtils_10.0.6     RevoUtilsMath_10.0.1

...

[BoPeng]

If the library is actually loaded, sessioninfo would include it, right?

[gaow]

Well, I'm talking about SoS workflow in SoS notebook. eg,

(BTW there seems to be a bug in sos notebook or lab. I'm on the master of both these packages)

[BoPeng]

They do not work with master of sos now.

[BoPeng]

I see. These (workflow and notebook) are two systems that do not work with each other.

[gaow]

I know. That's why I was asking if it is too much of a request to have the information added.

[BoPeng]

The problem is that sessioninfo needs live R (and others) sessions to work, and the workflow version does not. Let me think harder.

[gaow]

A dumb implementation would be to add a section to %sessioninfo output called "Workflow environment", call Rscript -e "library(ggplot2); sessionInfo()", capture the stdout and append it to this section. (need to hide the stderr or use suppressMessages(library(ggplot2))).

[gaow]

Bump this one. Would be a nice point to make in justifying that SoS workflow with Jupyter IDE helps with reproducibility and book-keeping.

[BoPeng]

It is even more difficult now because %run and %sosrun are executed by sos CLI outside of sos notebook. I tend to think we should implement things like Py_Requirements, Conda, R_Requirements that

1. Centralize workflow requirement
2. Consistent with binder because they corresponds to binder requirement files and can be used to create binder environments
3. Allow systematic way to create session info, rather than hacking the workflow.