Thanks, Eric.  That's very clear.  (This is a great list.)

-- Mike


From: Eric Schulte <schulte.eric@gmail.com>
To: Michael Hannon <jm_hannon@yahoo.com>
Cc: Org-Mode List <emacs-orgmode@gnu.org>
Sent: Tuesday, September 27, 2011 5:42 PM
Subject: Re: [O] Question about using :cache in Babel (for R)

Michael Hannon <jm_hannon@yahoo.com> writes:

> Greetings.  I have a question that relates to the use of the :cache option in
> Babel sessions using R.
>
> Here's the relevant configuration information:
>
>     Emacs  : GNU Emacs 23.2.1 (x86_64-redhat-linux-gnu, GTK+ Version 2.24.4)
>      of 2011-05-23 on x86-12.phx2.fedoraproject.org
>     Package: Org-mode version 7.7 (release_7.7.328.g1a97)
>
> I've appended the text of a file that exhibits the issue, which is the
> following.  If I explicitly evaluate a given code block, using, let's say, C-c
> C-c, I get the results stuck into the Org file, as in the first source-code
> block below.  If I then repeatedly export the file (to PDF, HTML, etc.), I get
> the cached results in the exported document.  For instance, the mean of the
> random variable does not change as I repeatedly export the file.
>
> If for some reason I have NOT explicitly evaluated a given code block, as with
> the second code block below, then during export I see a message "Evaluating
> code block" (or similar), indicating that Babel is NOT using a cached version,
> and no cacheable output is produced as a result, meaning that the entire
> calculation will be repeated during the next export.  For instance, the mean
> of the random variable will be different every time I export the file.
>

Hi Mike,

This is the expected behavior.  Org-mode files are just plain text, so
if the results are not saved in the Org-mode file then there is no place
to cache them.  Since Org does not change the contents of Org-mode files
the only way to save the result of a code block is to interactively
evaluate the code block -- there is no way to generate and save cached
results during export.

>
> I guess I can see the logic of this, but I wonder if there is some way to
> instruct Babel to automatically cache the output of any code blocks it
> evaluates, i.e., via a header argument or elisp variable or ... .
>

Nope, in the past we have played with things like an in-memory variable,
but have abandoned such things as they would move away from the "just
plain text" spirit of Org-mode, would require external storage on disk,
and could ultimately lead to more confusing behavior than the current
situation.

>
> Thanks,
>
> -- Mike
>
> P.S. Is the SHA1(?) identifier for the results a new feature?
>

No, this is how Babel knows if the contents of the code block have
changed and hence if the cached results are still valid.

Best -- Eric

>
> ----------
>
> #+TITLE: 
> #+AUTHOR:
> #+DATE:
>
> #+STARTUP: hideblocks
> #+BABEL: :session *R* :cache yes :results output verbatim :exports both
> :tangle yes
>
> * This is a test
>
> #+begin_src R
>
>  x = rnorm(10)
>  mean(x)
>
> #+end_src
>
> #+results[9b9372488662083d0d211b9aad9c2985faba20cd]:
> : [1] -0.07904561
>
> * Here's a second test
>
> #+begin_src R
>
>  x = rnorm (50)
>  summary(x)
>
> #+end_src

--
Eric Schulte
http://cs.unm.edu/~eschulte/