Re: [O] R code block produces only partial output

[John Hendy]

On Mon, Aug 4, 2014 at 7:46 PM, Andreas Kiermeier
<address@hidden> wrote:
> Hi Eric,
> I'm running Org 8.2.7c (via ELPA) and ESS 14.05.
> I've tried to run with a minimal Org setup using only the following in
> my init file.
> The shortcut I use to start Emacs (on Windows 8.1) is:
> C:\emacs\bin\runemacs.exe -Q -l "C:\emacs\bin\minimal-org.el"
>
> However, the results are the same as before. Any thoughts on what else
> I could try?
>

[snip]

I can reproduce with a minimal config and on Linux (to toss in a
confirmation). I *also* accidentally just stumbled on a way I can
toggle the behavior between correct and incorrect! It arose because my
file was not able to use the #+PROPERTY settings you have -- my
results were not obeying what I had set. I looked around and believe
the syntax is correct per this page:
- http://orgmode.org/manual/Header-arguments-in-Org-mode-properties.html

However, I guarantee that the results weren't updating to ":results
output drawer" as I had them set. I also noticed that the
"header-args:R :session *R*" also didn't seem to be working, as if I
changed the name, Emacs wouldn't ask me if I wanted to create a new R
session upon block execution. Here's what I've found:

Correct results:

#+begin_src org

#+STARTUP: showall indent hidestars
#+PROPERTY: header-args:R :session r

#+BEGIN_SRC R :results output drawer
  require(rms)
  set.seed(123)
  x <- rnorm(100)
  describe(x)
#+END_SRC

#+RESULTS:
:RESULTS:
x
       n  missing   unique     Mean      .05      .10      .25      .50
     100        0      100  0.09041 -1.26508 -1.06822 -0.49385  0.06176
     .75      .90      .95
 0.69182  1.26450  1.56653

lowest : -2.309 -1.967 -1.687 -1.549 -1.265
highest:  1.715  1.787  2.050  2.169  2.187
:END:


#+end_src


Incorrect results:

#+begin_src org

#+STARTUP: showall indent hidestars


#+BEGIN_SRC R :session r :results output drawer
  require(rms)
  set.seed(123)
  x <- rnorm(100)
  describe(x)
#+END_SRC

#+RESULTS:
:RESULTS:
75      .90      .95
 0.69182  1.26450  1.56653

lowest : -2.309 -1.967 -1.687 -1.549 -1.265
highest:  1.715  1.787  2.050  2.169  2.187
:END:

#+end_src


The reason, I think, is this (also correct):

#+begin_src org

#+STARTUP: showall indent hidestars


#+BEGIN_SRC R :results output drawer
  require(rms)
  set.seed(123)
  x <- rnorm(100)
  describe(x)
#+END_SRC

#+RESULTS:
:RESULTS:
x
       n  missing   unique     Mean      .05      .10      .25      .50
     100        0      100  0.09041 -1.26508 -1.06822 -0.49385  0.06176
     .75      .90      .95
 0.69182  1.26450  1.56653

lowest : -2.309 -1.967 -1.687 -1.549 -1.265
highest:  1.715  1.787  2.050  2.169  2.187
:END:

#+end_src

So, session settings in the #+PROPERTY config *or* no session set at
all yields correct results. Again, on a fresh start the #+PROPERTY
method does not ask to start a new R session. Last confirmation is
that if one leaves the session #+PROPERTY line (with none in the code
block) and comments out the "require(rms)" line after already having
run the block, you get the error "Could not find the function
'describe'". So, there's no session activity finding the
already-loaded library.

The remaining questions for me is why would results differ depending
on whether or not it's going through a session?

Eric: this also explains why your minimal .org file couldn't
reproduce: there's no setting of the session (again, not sure why that
matters... but it seems to).


John