emacs-orgmode
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [O] Selectively export RESULTS


From: Eric Schulte
Subject: Re: [O] Selectively export RESULTS
Date: Fri, 02 Mar 2012 12:33:58 -0700
User-agent: Gnus/5.13 (Gnus v5.13) Emacs/24.0.93 (gnu/linux)

Matthew Landis <address@hidden> writes:

> Eric Schulte <eric.schulte <at> gmx.com> writes:
>
>> 
>> Matthew Landis <landis <at> isciences.com> writes:
>> 
>> Have you tried this?  The cache header argument has special handling of
>> code blocks with sessions to handle such cases.
>
> Fair enough!  I hadn't tried it.  But I did just try this example:
>
> -------------------------------------
> #+TITLE: Super simple example using cache header arguments
> #+PROPERTIES: eval no
>

The above line should be #+PROPERTY: singular.  That explains why
file-wide settings aren't working for you.

>
> Here is a really simple example.
>
> #+begin_src R :session :results silent :exports code :cache yes
> x <- rnorm(100)
> cat('ran this code block')
> #+end_src
>
> here is code block two.
>
> #+begin_src R :session :results graphics :exports both :file hist.png :cache 
> yes
> hist(x, main = 'A new title' )
> #+end_src
>
> #+results[1987d49b46ffd8b7263dc04e7c7b5d25f90342aa]:
> [[file:hist.png]]
> ---------------------------------
>

Thanks for the example, after working through it I now know what is
causing this issue.  The first code block will always be run for two
reasons.
1. it is run in a session, which means that Babel can not guess at to
   what state it could be changing internal to the session, so it
   defaults to the safest option which is allowing the code block to run
2. since this code block returns no results, there is no place for Babel
   to store the hash key holding the information on the code block and
   arguments used to produce the results.  Remember that Org-mode is
   nothing more than the text in the file, so without this stored hash,
   there is no way for Babel to know that the code block was previously
   run, or that it may have results.

For these reasons I would suggest either fixing the properties issue
above, which will allow you to set eval to "no" before export when you
are content with your current results, or I would suggest that you
switch from session based evaluation to explicitly passing the values
through Org-mode, which will allow babel to cache results.

I will update the ":cache" section of the manual so that it mentions the
issues around mixing ":session" and ":cache" header arguments.

Best,

>
> RESULTS:
> On 1st export to html, both code blocks are run, and #+results block is *not* 
> created.  
>
> On 2nd export, both code blocks are run again, counter to my desires.
>
> IF I run the code blocks interactively first using C-c C-c, the #+results 
> block 
> is created for the code block 2.  After that, subsequent exports run code 
> block 
> 1 but not code block 2.  Again, this is counter to my desires because I want 
> code block 1 to be ignored as well.  I understand why code block 1 is rerun, 
> because I have :results silent, but I'd rather not have to change that.  
> Generally, code block 1 is producing large R data.frames which don't need to 
> be 
> viewed anywhere.
>
>> Are you requesting a new option to :cache, such that even if the code
>> block has changes the old results are used anyway?
>
> Sort of.  I think I'm asking for a file wide option to decide whether to run 
> any 
> code blocks, or just do the export based on whatever is currently existing in 
> the #+results blocks.
>
> As in this example, I tried #+PROPERTIES: eval no but it is ignored.  It 
> would 
> be great if I could set an eval argument at the top of the file to be either 
> 'no' (don't run any code blocks) 'yes' (run all code blocks) or 'block' 
> respect 
> block level eval settings 
>
> Maybe such functionality exists in some other way - I claim only beginners 
> knowledge of org-babel.
>
>
>
>
>

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



reply via email to

[Prev in Thread] Current Thread [Next in Thread]