Re: org babel, ess, R

[John Hendy <jw.hendy@gmail.com>]

On Wed, Jul 5, 2017 at 4:22 PM, Nicolas Goaziou <mail@nicolasgoaziou.fr> wrote:
> Hello,
>
> Vikas Rawal <vikaslists@agrarianresearch.org> writes:
>
>> C-RET seems to be picking up the custom session name if it is already
>> there. But otherwise it generates the default *R* session there,
>
> Is there a bug to fix? If so, could start a new thread about it, ideally
> with an ECM demonstrating the issue?

I guess Vikas can speak for Vikas, but he was responding to my
observation of the same:

"... I'd like Org to pick up the fact that there's a :session
argument with a custom name and use that for C-RET in the edit buffer;
otherwise it generates the default *R* session there, but if you C-c
C-c the block later it puts it into the custom session name."

Here's an ECM:

#+begin_example

* heading

#+name: x
#+begin_src R :session r

x <- 1:10

#+end_src

#+name: y
#+begin_src R :session r

y <- x^2

#+end_src

#+end_example

So, say I start working on a document for the day and enter the
editing buffer with C-c ' on block x. I then C-RET from the edit
buffer, which evaluates x and starts a new R session, *R*. Now I'm
done and go on to block y, but don't need to edit. Maybe I just want
to update a plot or other result in the document, so I C-c C-c:

Error: object 'x' not found

This really, really common for my work. I define a setup block
somewhere in which I load libraries, read in thousands of lines of
sensor code, and process things (moving averages, summary stats, etc.)
and create my sort of "master object." From there I have sections of a
report that subset the data, plot various things, but typically have
them all set to :eval no by default. If I pick up work on a file, edit
my setup code block, but only interactively because I forget I need to
C-c C-c something first to kick off the custom session, the work was
sort of for naught as everything is in the wrong session.

In this example, I realize the error and go back to block x and C-c
C-c after failing on block y. At that point, when I go to a block
interactively to do C-RET again, I get this in the minibuffer.

Process to load into: {r | *R* | *new* }

Then I switch to *R* to quit() and end the spurious session.

Really, it was a comment about how Org behaves, not necessarily a bug.
That said, speaking for myself, I would never want a default session
in interactive mode when I've defined one in the code heading. I'd
love if Org could pick up on that setting, such that the first C-RET
of the day dropped into the session named by the block I'm editing. At
present, interactive mode appears to ignore this.


Hope that helps clarify,
John





> Regards,
>
> --
> Nicolas Goaziou