Revert "DONE take default values for header args from properties"

[rgr-org-mode.git] / org-babel.org

#+TITLE: org-babel --- facilitating communication between programming languages and people
#+SEQ_TODO: PROPOSED TODO STARTED | DONE DEFERRED REJECTED
#+OPTIONS: H:3 num:nil toc:t
#+STARTUP: oddeven hideblocks

Through Org-Babel Org-Mode can communicate with programming languages.
Code contained in source-code blocks can be evaluated and data can
pass seamlessly between different programming languages, Org-Mode
constructs (tables, file links, example text) and interactive comint
buffers.

In this document:
- The [[* Introduction][Introduction]] :: provides a brief overview of the design and use
     of Org-Babel including tutorials and examples.
- In [[* Getting started][Getting Started]] :: find instructions for installing org-babel
     into your emacs configuration.
- The [[* Tasks][Tasks]] :: section contains current and past tasks roughly ordered
     by TODO state, then importance or date-completed.  This would be
     a good place to suggest ideas for development.
- The [[* Bugs][Bugs]] :: section contains bug reports.
- The [[* Tests][Tests]] :: section consists of a large table which can be
     evaluated to run Org-Babel's functional test suite.  This
     provides a good overview of the current functionality with
     pointers to example source blocks.
- The [[* Sandbox][Sandbox]] :: demonstrates much of the early/basic functionality
     through commented source-code blocks.

Also see the [[file:library-of-babel.org][Library of Babel]], an extensible collection of ready-made
and easily-shortcut-callable source-code blocks for handling common
tasks.

* Introduction

Org-Babel enables *communication* between programming languages and
between people.

Org-Babel provides:
- communication between programs :: Data passes seamlessly between
     different programming languages, Org-Mode constructs (tables,
     file links, example text) and interactive comint buffers.
- communication between people :: Data and calculations are embedded
     in the same document as notes explanations and reports.

** communication between programs

Org-Mode supports embedded blocks of source code (in any language)
inside of Org documents.  Org-Babel allows these blocks of code to be
executed from within Org-Mode with natural handling of their inputs
and outputs.

*** simple execution
with both scalar, file, and table output

*** reading information from tables

*** reading information from other source blocks (disk usage in your home directory)

This will work for Linux and Mac users, not so sure about shell
commands for windows users.

To run place the cursor on the =#+begin_src= line of the source block
labeled directory-pie and press =\C-c\C-c=.

#+srcname: directories
#+begin_src bash :results replace
cd ~ && du -sc * |grep -v total
#+end_src

#+resname: directories
|       64 | "Desktop"   |
| 11882808 | "Documents" |
|  8210024 | "Downloads" |
|   879800 | "Library"   |
|    57344 | "Movies"    |
|  7590248 | "Music"     |
|  5307664 | "Pictures"  |
|        0 | "Public"    |
|      152 | "Sites"     |
|        8 | "System"    |
|       56 | "bin"       |
|  3274848 | "mail"      |
|  5282032 | "src"       |
|     1264 | "tools"     |

#+srcname: directory-pie
#+begin_src R :var dirs = directories :session R-pie-example
pie(dirs[,1], labels = dirs[,2])
#+end_src






*** operations in/on tables

#+tblname: grades-table
| student | grade | letter |
|---------+-------+--------|
|       1 |    99 | A      |
|       2 |    59 | F      |
|       3 |    75 | C      |
|       4 |    15 | F      |
|       5 |     7 | F      |
|       6 |    13 | F      |
#+TBLFM: $2='(sbe random-score-generator)::$3='(sbe assign-grade (score $2))

#+srcname: assign-grade
#+begin_src ruby :var score=99
case score
   when 0..59: "F"
   when 60..69: "D"
   when 70..79: "C"
   when 80..89: "B"
   when 90..100: "A"
   else "Invalid Score"
end
#+end_src

#+srcname: random-score-generator
#+begin_src ruby 
rand(100)
#+end_src

#+srcname: show-distribution
#+begin_src R :var grades=grades-table :session *R*
hist(grades$grade)
#+end_src





** communication between people
Quick overview of Org-Mode's exportation abilities, with links to the
online Org-Mode documentation, a focus on source-code blocks, and the
exportation options provided by Org-Babel.

*** Interactive tutorial
This would demonstrate applicability to Reproducible Research, and
Literate Programming.

*** Tests embedded in documentation
org-babels own functional tests are contained in a large org-mode
table, allowing the test suite to be run be evaluation of the table
and the results to be collected in the same table.

*** Emacs initialization files stored in Org-Mode buffers
Using `org-babel-tangle' it is possible to embed your Emacs
initialization into org-mode files.  This allows for folding,
note-taking, todo's etc... embedded with the source-code of your Emacs
initialization, and through org-mode's publishing features aids in
sharing your customizations with others.

It may be worthwhile to create a fork of Phil Hagelberg's
[[http://github.com/technomancy/emacs-starter-kit/tree/master][emacs-starter-kit]] which uses literate org-mode files for all of the
actual elisp customization.  These org-mode files could then be
exported to html and used to populate the repositories wiki on [[http://github.com/][github]].


** features

*** code evaluation (comint buffer sessions and external processes)
There are two main ways to evaluate source blocks with org-babel.

- external :: By default (if the =:session= header argument is not
              present) all source code blocks are evaluated in
              external processes.  In these cases an external process
              is used to evaluate the source-code blocks.
- session :: Session based evaluation uses persistent sessions in
             comint buffers.  Sessions can be used across multiple
             source blocks setting and accessing variables in the
             global environment.

             Evaluating source blocks in sessions also allows for
             interaction with the code.  To jump to the session of a
             source block use the `org-babel-pop-to-session' command
             or press =M-[down]= while inside of a source code block.
             When called with a prefix argument
             `org-babel-pop-to-session' will evaluate all header
             arguments before jumping to the source-code block.

*** results (values and outputs)
Either the *value* or the *output* of source code blocks can be
collected after evaluation.

- value :: The default way to collect results from a source-code block
           is to return the value of the last statement in the block.
           This can be thought of as the return value of the block.
           In this case any printed output of the block is ignored.
           This can be though of a similar to a "functional" value of
           evaluation.
- output :: Another way of generating results from a source-code block
            is to collect the output generated by the execution of the
            block.  In this case all printed output is collected
            throughout the execution of the block.  This can be
            thought of as similar to a "script" style of evaluation.


* Getting started
Add the following lines to your .emacs, replacing the path as
appropriate. A good place to check that things are up and running
would then be [[#sandbox][the sandbox]].
#+begin_src emacs-lisp
  (add-to-list 'load-path "/path/to/org-babel/lisp")
  (require 'org-babel-init)
#+end_src


* Tasks [37/58]
** PROPOSED raise elisp error when source-blocks return errors
Not sure how/if this would work, but it may be desirable.

** PROPOSED allow `anonymous' function block with function call args?
   My question here is simply whether we're going to allow
#+begin_src python(arg=ref)
# whatever
#+end_src

but with preference given to
#+srcname blockname(arg=ref)
** PROPOSED allow :result as synonym for :results?
** PROPOSED allow 'output mode to return stdout as value?
   Maybe we should allow this. In fact, if block x is called
   with :results output, and it references blocks y and z, then
   shouldn't the output of x contain a concatenation of the outputs of
   y and z, together with x's own output? That would raise the
   question of what happens if y is defined with :results output and z
   with :results value. I guess z's (possibly vector/tabular) output
   would be inside a literal example block containing the whole lot.
** PROPOSED optional timestamp for output
   Add option to place an (inactive) timestamp at the #+resname, to
   record when that output was generated.

*** source code block timestamps (optional addition)
    [Eric] If we did this would we then want to place a timestamp on the
    source-code block, so that we would know if the results are
    current or out of date?  This would have the effect of caching the
    results of calculations and then only re-running if the
    source-code has changed.  For the caching to work we would need to
    check not only the timestamp on a source-code block, but also the
    timestamps of any tables or source-code blocks referenced by the
    original source-code block.

    [Dan] I do remember getting frustrated by Sweave always having to
    re-do everything, so this could be desirable, as long as it's easy
    to over-ride of course. I'm not sure it should be the default
    behaviour unless we are very confident that it works well.

**** maintaining source-code block timestamps
     It may make sense to add a hook to `org-edit-special' which could
     update the source-code blocks timestamp.  If the user edits the
     contents of a source-code block directly I can think of no
     efficient way of maintaining the timestamp.

** TODO make tangle files read-only?
   With a file-local variable setting, yea that makes sense.  Maybe
   the header should reference the related org-mode file.

** TODO take default values for header args from properties
   Use file-wide and subtree wide properties to set default values for
   header args.
** TODO support for working with =*Org Edit Src Example*= buffers [2/4]
*** TODO optionally evaluate header references when we switch to =*Org Edit Src*= buffer
That seems to imply that the header references need to be evaluated
and transformed into the target language object when we hit C-c ' to
enter the *Org Edit Src* buffer [DED]

Good point, I heartily agree that this should be supported [Eric]

(or at least before the first time we attempt to evaluate code in that
buffer -- I suppose there might be an argument for lazy evaluation, in
case someone hits C-c ' but is "just looking" and not actually
evaluating anything.) Of course if evaluating the reference is
computationally intensive then the user might have to wait before they
get the *Org Edit Src* buffer. [DED]

I fear that it may be hard to anticipate when the references will be
needed, some major-modes do on-the-fly evaluation while the buffer is
being edited.  I think that we should either do this before the buffer
is opened or not at all, specifically I think we should resolve
references if the user calls C-c ' with a prefix argument.  Does that
sound reasonable? [Eric]

Yes [Dan]

[Dan] So now that we have org-src-mode and org-src-mode-hook, I guess
org-babel should do this by using the hook to make sure that, when C-c
C-' is issued on a source block, any references are resolved and
assignments are made in the appropriate session.
*** TODO set buffer-local-process variables appropriately [DED]
    I think something like this would be great. You've probably
already thought of this, but just to note it down: it would be really
nice if org-babel's notion of a buffer's 'session/process' played
nicely with ESS's notion of the buffer's session/process. ESS keeps
the current process name for a buffer in a buffer-local variable
ess-local-process-name. So one thing we will probably want to do is
make sure that the *Org Edit Src Example* buffer sets that variable
appropriately. [DED]

I had not thought of that, but I agree whole heartedly. [Eric]

Once this is done every variable should be able to dump regions into
their inferior-process buffer using major-mode functions.
*** DEFERRED send code to inferior process
Another thought on this topic: I think we will want users to send
chunks of code to the interpreter from within the *Org Edit Src*
buffer, and I think that's what you have in mind already. In ESS that
is done using the ess-eval-* functions. [DED]

I think we can leave this up to the major-mode in the source code
buffer, as almost every source-code major mode will have functions for
doing things like sending regions to the inferior process.  If
anything we might need to set the value of the buffer local inferior
process variable. [Eric]

*** DONE some possible requests/proposed changes for Carsten [4/4]
    While I remember, some possible requests/proposed changes for Carsten
    come to mind in that regard:

**** DONE Remap C-x C-s to save the source to the org buffer?
     I've done this personally and I find it essential. I'm using 
#+begin_src emacs-lisp
(defun org-edit-src-save ()
  "Update the parent org buffer with the edited source code, save
the parent org-buffer, and return to the source code edit
buffer."
  (interactive)
  (let ((p (point)))
    (org-edit-src-exit)
    (save-buffer)
    (org-edit-src-code)
    (goto-char p)))

(define-key org-exit-edit-mode-map "\C-x\C-s" 'org-edit-src-save)
#+end_src     
    which seems to work.

I think this is great, but I think it should be implemented in the
org-mode core

**** DEFERRED Rename buffer and minor mode?
     Something shorter than *Org Edit Src Example* for the buffer
     name. org-babel is bringing org's source code interaction to a
     level of maturity where the 'example' is no longer
     appropriate. And if further keybindings are going to be added to
     the minor mode then maybe org-edit-src-mode is a better name than
     org-exit-edit-mode.

     Maybe we should name the buffer with a combination of the source
     code and the session.  I think that makes sense.

     [ES] Are you also suggesting a new org-edit-src minor mode?
     [DED] org-exit-edit-mode is a minor mode that already exists:

     Minor mode installing a single key binding, "C-c '" to exit special edit.

     org-edit-src-save now has a binding in that mode, so I guess all
     I'm saying at this stage is that it's a bit of a misnomer. But
     perhaps we will also have more functionality to add to that minor
     mode, making it even more of a misnomer. Perhaps something like
     org-src-mode would be better.
**** DONE Changed minor mode name and added hooks
     
**** DONE a hook called when the src edit buffer is created
     This should be implemented in the org-mode core

** TODO resolve references to other org buffers/files
   This would allow source blocks to call upon tables, source-blocks,
   and results in other org buffers/files.
   
   See...
   - [[file:lisp/org-babel-ref.el::TODO%20allow%20searching%20for%20names%20in%20other%20buffers][org-babel-ref.el:searching-in-other-buffers]]
   - [[file:lisp/org-babel.el::defun%20org-babel%20find%20named%20result%20name][org-babel.el#org-babel-find-named-result]]
** TODO resolve references to other non-org files
   - tabular data in .csv, .tsv etc format
   - files of interpreted code: anything stopping us giving such files
     similar status to a source code block?
   - Would be nice to allow org and non-org files to be remote
** TODO figure out how to handle errors during evaluation
   I expect it will be hard to do this properly, but ultimately it
   would be nice to be able to specify somewhere to receive STDERR,
   and to be warned if it is non-empty.

   Probably simpler in non-session evaluation than session? At least
   the mechanism will be different I guess.

   R has a try function, with error handling, along the lines of
   python. I bet ruby does too. Maybe more of an issue for functional
   style; in my proposed scripting style the error just gets dumped to
   the org buffer and the user is thus alerted.
** TODO figure out how to handle graphic output
This is listed under [[* graphical output][graphical output]] in out objectives.

This should take advantage of the =:results file= option, and
languages which almost always produce graphical output should set
=:results file= to true by default (this is currently done for the
gnuplot and ditaa languages).  That would handle placing these results
in the buffer.  Then if there is a combination of =silent= and =file=
=:results= headers we could drop the results to a temp buffer and pop
open that buffer...

Display of file results is addressed in the [[* =\C-c \C-o= to open results of source block][open-results-task]].

*** TODO R graphics to screen means session evaluation
    If R graphical output is going to screen then evaluation must be
    in a session, otherwise the graphics will disappear as soon as the
    R process dies.

*** Adding to a discussion started in email
    I'm not deeply wedded to these ideas, just noting them down. I'm
    probably just thinking of R and haven't really thought about how
    this fits with the other graphics-generating languages.
Dan:
> I used the approach below to get graphical file output
> today, which is one idea at least. Maybe it could be linked up with
> your :results file variable. (Or do we need a :results image for R?)
>
Eric:
I don't think we need a special image results variable, but I may be
missing what the code below accomplishes.  Would the task I added about
adding org-open-at-point functionality to source code blocks take care
of this need?

Dan: I'm not sure. I think the ability for a script to generate both
text and graphical output might be a natural expectation, at least for
R users.

>
> Dan
>
> #+srcname: cohort-scatter-plots-2d(org_babel_graphical_output_file="cohort-scatter-plots-2d.png")
> #+begin_src R 
>   if(exists("org_babel_output_file"))
>       png(filename=org_babel_graphical_output_file, width=1000, height=1000)
>   ## plotting code in here
>   if(exists("org_babel_graphical_output_file")) dev.off()
> #+end_src

Dan: Yes, the results :file option is nice for dealing with graphical
output, and that could well be enough. Something based on the scheme
above would have a couple of points in its favour:
1. It's easy to switch between output going to on-screen graphics and
   output going to file: Output will go to screen unless a string variable
   with a standard name (e.g. ""org_babel_graphical_output_file"")
   exists in which case it will go to the file indicated by the value
   of that variable.
2. The block can return a result / script output, as well as produce
   graphical output.

In interactive use we might want to allow the user to choose between
screen and file output. In non-interactive use such as export, it
would be file output (subject to the :exports directives).

** TODO =\C-c \C-o= to open results of source block
by adding a =defadvice= to =org-open-at-point= we can use the common
=\C-c \C-o= keybinding to open the results of a source-code block.
This would be especially useful for source-code blocks which generate
graphical results and insert a file link as the results in the
org-mode buffer.  (see [[* figure out how to handle graphic output][TODO figure out how to handle graphic output]]).
This could also act reasonably with other results types...

- file :: use org-open-at-point to open the file
- scalar :: open results unquoted in a new buffer
- tabular :: export the table to a new buffer and open that buffer
** TODO Finalise behaviour regarding vector/scalar output
*** DONE Stop spaces causing vector output
This simple example of multilingual chaining produces vector output if
there are spaces in the message and scalar otherwise.

[Not any more]

#+srcname: msg-from-R(msg=msg-from-python)
#+begin_src R
paste(msg, "und R", sep=" ")
#+end_src

#+resname:
: org-babel speaks elisp y python und R

#+srcname: msg-from-python(msg=msg-from-elisp)
#+begin_src python
msg + " y python"
#+end_src

#+srcname: msg-from-elisp(msg="org-babel speaks")
#+begin_src emacs-lisp
(concat msg " elisp")
#+end_src
** STARTED share org-babel [1/4]
how should we share org-babel?

*** DONE post to org-mode
*** TODO post to ess mailing list
*** TODO create a org-babel page on worg
*** TODO create a short screencast demonstrating org-babel in action

*** examples
we need to think up some good examples

**** interactive tutorials
This could be a place to use [[* org-babel assertions][org-babel assertions]].

for example the first step of a tutorial could assert that the version
of the software-package (or whatever) is equal to some value, then
source-code blocks could be used with confidence (and executed
directly from) the rest of the tutorial.

**** answering a text-book question w/code example
org-babel is an ideal environment enabling both the development and
demonstrationg of the code snippets required as answers to many
text-book questions.

**** something using tables
maybe something along the lines of calculations from collected grades

**** file sizes
Maybe something like the following which outputs sizes of directories
under the home directory, and then instead of the trivial =emacs-lisp=
block we could use an R block to create a nice pie chart of the
results.

#+srcname: sizes
#+begin_src bash :results replace
du -sc ~/*
#+end_src

#+begin_src emacs-lisp :var sizes=sizes :results replace
(mapcar #'car sizes)
#+end_src

** TODO command line execution
Allow source code blocks to be called form the command line.  This
will be easy using the =sbe= function in [[file:lisp/org-babel-table.el][org-babel-table.el]].

This will rely upon [[* resolve references to other buffers][resolve references to other buffers]].

** TODO inline source code blocks [3/5]
   Like the =\R{ code }= blocks

   not sure what the format should be, maybe just something simple
   like =src_lang[]{}= where lang is the name of the source code
   language to be evaluated, =[]= is optional and contains any header
   arguments and ={}= contains the code.

   (see [[* (sandbox) inline source blocks][the-sandbox]])

*** DONE evaluation with \C-c\C-c
Putting aside the header argument issue for now we can just run these
with the following default header arguments
- =:results= :: silent
- =:exports= :: results

*** DONE inline exportation
Need to add an interblock hook (or some such) through org-exp-blocks
*** DONE header arguments
We should make it possible to use header arguments.

*** TODO fontification
we should color these blocks differently

*** TODO refine html exportation
should use a span class, and should show original source in tool-tip

** TODO LoB: re-implement plotting and analysis functions from org-R
   I'll do this soon, now that we things are a bit more settled and we
   have column names in R.
** PROPOSED conversion between org-babel and noweb (e.g. .Rnw) format
   I haven't thought about this properly. Just noting it down. What
   Sweave uses is called "R noweb" (.Rnw).
   
   I found a good description of noweb in the following article (see
   the [[http://www.cs.tufts.edu/~nr/pubs/lpsimp.pdf][pdf]]).
   
   I think there are two parts to noweb, the construction of
   documentation and the extraction of source-code (with notangle).

   *documentation*: org-mode handles all of our documentation needs in
   a manner that I believe is superior to noweb.
   
   *source extraction* At this point I don't see anyone writing large
   applications with 100% of the source code contained in org-babel
   files, rather I see org-babel files containing things like
   - notes with active code chunks
   - interactive tutorials
   - requirements documents with code running test suites
   - and of course experimental reports with the code to run the
     experiment, and perform analysis

   Basically I think the scope of the programs written in org-babel
   (at least initially) will be small enough that it wont require the
   addition of a tangle type program to extract all of the source code
   into a running application.

   On the other hand, since we already have named blocks of source
   code which reference other blocks on which they rely, this
   shouldn't be too hard to implement either on our own, or possibly
   relying on something like noweb/notangle.

** PROPOSED support for passing paths to files between source blocks
Maybe this should be it's own result type (in addition to scalars and
vectors).  The reason being that some source-code blocks (for example
ditaa or anything that results in the creation of a file) may want to
pass a file path back to org-mode which could then be inserted into
the org-mode buffer as a link to the file...

This would allow for display of images upon export providing
functionality similar to =org-exp-blocks= only in a more general
manner.
** DEFERRED Support rownames and other org babel table features?

   The full org table features are detailed in the manual [[http://orgmode.org/manual/Advanced-features.html#Advanced-features][here]].

*** rownames
   Perhaps add a :rownames header arg. This would be an integer
    (usually 1) which would have the effect of post-processing all the
    variables created in the R session in the following way: if the
    integer is j, set the row names to the contents of column j and
    delete column j. Perhaps it is artificial to allow this integer to
    take any value other than 1. The default would be nil which would
    mean no such behaviour.

    Actually I don't know about that. If multiple variables are passed
    in, it's not appropriate to alter them all in the same way. The
    rownames specification would normally refer to just one of the
    variables. For now maybe just say this has to be done in R. E.g.

#+TBLNAME: sample-sizes
  | collection      | size | exclude | include | exclude2 | include2 |
  |-----------------+------+---------+---------+----------+----------|
  | 58C             | 2936 |       8 |    2928 |      256 |     2680 |
  | MS              | 5852 |     771 |    5081 |      771 |     5081 |
  | NBS             | 2929 |      64 |    2865 |      402 |     2527 |
  | POBI            | 2717 |       1 |    2716 |        1 |     2716 |
  | 58C+MS+NBS+POBI |      |         |   13590 |          |    13004 |
#+TBLFM: @2$4=@2$2 - @2$3::@2$6=@2$2 - @2$5::@3$4=@3$2-@3$3::@3$6=@3$2 - @3$5::@4$4=@4$2 - @4$3::@4$6=@4$2 - @4$5::@5$4=@5$2-@5$3::@5$6=@5$2 - @5$5::@6$4=vsum(@2$4..@5$4)::@6$6=vsum(@2$6..@5$6)

#+srcname: make-size-table(size=sample-sizes)
#+begin_src R 
  rownames(size) <- size[,1]
  size <- size[,-1]
#+end_src


*** Old notes
    [I don't think it's as problematic as this makes out]
    This is non-trivial, but may be worth doing, in particular to
    develop a nice framework for sending data to/from R.
**** Notes
    In R, indexing vector elements, and rows and columns, using
    strings rather than integers is an important part of the
    language.
 - elements of a vector may have names
 - matrices and data.frames may have "column names" and "row names"
   which can be used for indexing
 - In a data frame, row names *must* be unique
Examples
#+begin_example
> # a named vector
> vec <- c(a=1, b=2)
> vec["b"]
b 
2 
> mat <- matrix(1:4, nrow=2, ncol=2, dimnames=list(c("r1","r2"), c("c1","c2")))
> mat
   c1 c2
r1  1  3
r2  2  4
> # The names are separate from the data: they do not interfere with operations on the data
> mat * 3
   c1 c2
r1  3  9
r2  6 12
> mat["r1","c2"]
[1] 3
> df <- data.frame(var1=1:26, var2=26:1, row.names=letters)
> df$var2
 [1] 26 25 24 23 22 21 20 19 18 17 16 15 14 13 12 11 10  9  8  7  6  5  4  3  2  1
> df["g",]
  var1 var2
g    7   20
#+end_example

 So it's tempting to try to provide support for this in org-babel. For example
 - allow R to refer to columns of a :var reference by their names
 - When appropriate, results from R appear in the org buffer with "named
   columns (and rows)"

   However none (?) of the other languages we are currently supporting
   really have a native matrix type, let alone "column names" or "row
   names". Names are used in e.g. python and perl to refer to entries
   in dicts / hashes.

   It currently seems to me that support for this in org-babel would
   require setting rules about when org tables are considered to have
   named columns/fields, and ensuring that (a) languages with a notion
   of named columns/fields use them appropriately and (b) languages
   with no such notion do not treat then as data.

 - Org allows something that *looks* like column names to be separated
   by a hline
 - Org also allows a row to *function* as column names when special
   markers are placed in the first column. An hline is unnecessary
   (indeed hlines are purely cosmetic in org [correct?]
 - Org does not have a notion of "row names" [correct?]
    
   The full org table functionality exeplified [[http://orgmode.org/manual/Advanced-features.html#Advanced-features][here]] has features that
   we would not support in e.g. R (like names for the row below).
   
**** Initial statement: allow tables with hline to be passed as args into R
   This doesn't seem to work at the moment (example below). It would
   also be nice to have a natural way for the column names of the org
   table to become the column names of the R data frame, and to have
   the option to specify that the first column is to be used as row
   names in R (these must be unique). But this might require a bit of
   thinking about.


#+TBLNAME: egtable
| col1 | col2    | col3 |
|------+---------+------|
|    1 | 2       |    3 |
|    4 | schulte |    6 |

#+TBLNAME: egtable2
| 1 |         2 | 3 |
| 4 | schulte   | 6 |

#+begin_src R :var tabel=egtable :colnames t
tabel
#+end_src

#+resname:
| "col1" | "col2"    | "col3" |
|--------+-----------+--------|
|      1 | 2         |      3 |
|      4 | "schulte" |      6 |


Another example is in the [[*operations%20in%20on%20tables][grades example]].

** DEFERRED use textConnection to pass tsv to R?
   When passing args from the org buffer to R, the following route is
   used: arg in buffer -> elisp -> tsv on file -> data frame in R. I
   think it would be possible to avoid having to write to file by
   constructing an R expression in org-babel-R-assign-elisp, something
   like this

#+begin_src emacs-lisp
(org-babel-R-input-command
 (format  "%s <- read.table(textConnection(\"%s\"), sep=\"\\t\", as.is=TRUE)"
          name (orgtbl-to-tsv value '(:sep "\t" :fmt org-babel-R-quote-tsv-field))))
#+end_src

   I haven't tried to implement this yet as it's basically just
   fiddling with something that works. The only reason for it I can
   think of would be efficiency and I haven't tested that.

   This Didn't work after an initial test.  I still think this is a
   good idea (I also think we should try to do something similar when
   writing out results frmo R to elisp) however as it wouldn't result
   in any functional changes I'm bumping it down to deferred for
   now. [Eric]

for quick tests

#+tblname: quick-test
| 1 | 2 | 3 |

#+srcname: quick-test-src-blk
#+begin_src R :var vec=quick-test
mean(mean(vec))
#+end_src

#+resname:
: 2


: 2

** DEFERRED Rework Interaction with Running Processes [2/5]
*** DONE robust to errors interrupting execution

#+srcname: long-runner-ruby
#+begin_src ruby :results silent
  sleep(10)
  :patton_is_an_grumpy
#+end_src

*** DEFERRED use =C-g= keyboard-quit to push processing into the background
This may be possible using the `run-with-timer' command.

I have no idea how this could work...

#+srcname: long-runner-ruby
#+begin_src ruby :results silent
  sleep(10)
  :patton_is_an_grumpy
#+end_src

*** TODO ability to select which of multiple sessions is being used
    Increasingly it is looking like we're going to want to run all
    source code blocks in comint buffer (sessions).  Which will have
    the benefits of
    1) allowing background execution
    2) maintaining state between source-blocks
       - allowing inline blocks w/o header arguments 

**** R sessions
     (like ess-switch-process in .R buffers)
     
     Maybe this could be packaged into a header argument, something
     like =:R_session= which could accept either the name of the
     session to use, or the string =prompt=, in which case we could use
     the =ess-switch-process= command to select a new process.
     
*** TODO evaluation of shell code as background process? 
    After C-c C-c on an R code block, the process may appear to
    block, but C-g can be used to reclaim control of the .org buffer,
    without interrupting the R evalution. However I believe this is not
    true of bash/sh evaluation. [Haven't tried other languages] Perhaps
    a solution is just to background the individual shell commands.

    The other languages (aside from emacs lisp) are run through the
    shell, so if we find a shell solution it should work for them as
    well.
    
    Adding an ampersand seems to be a supported way to run commands in
    the background (see [[http://www.emacswiki.org/emacs/ExecuteExternalCommand#toc4][external-commands]]).  Although a more extensible
    solution may involve the use of the [[elisp:(progn (describe-function 'call-process-region) nil)][call-process-region]] function.
    
    Going to try this out in a new file [[file:lisp/org-babel-proc.el][org-babel-proc.el]].  This should
    contain functions for asynchronously running generic shell commands
    in the background, and then returning their input.

**** partial update of org-mode buffer
    The sleekest solution to this may be using a comint buffer, and
    then defining a filter function which would incrementally interpret
    the results as they are returned, including insertion into the
    org-mode buffer.  This may actually cause more problems than it is
    worth, what with the complexities of identifying the types of
    incrementally returned results, and the need for maintenance of a
    process marker in the org buffer.

**** 'working' spinner
     It may be nice and not too difficult to place a spinner on/near the
     evaluating source code block

*** TODO conversion of output from interactive shell, R (and python) sessions to org-babel buffers
    [DED] This would be a nice feature I think. Although an org-babel
    purist would say that it's working the wrong way round... After
    some interactive work in a *R* buffer, you save the buffer, maybe
    edit out some lines, and then convert it to org-babel format for
    posterity. Same for a shell session either in a *shell* buffer, or
    pasted from another terminal emulator. And python of course.

** DEFERRED improve the source-block snippet
any real improvement seems somewhat beyond the ability of yasnippet
for now.

[[file:~/src/emacs-starter-kit/src/snippets/text-mode/rst-mode/chap::name%20Chapter%20title][file:~/src/emacs-starter-kit/src/snippets/text-mode/rst-mode/chap::name Chapter title]]
#+begin_example
,#name : Chapter title
,# --
${1:Chapter}
${1:$(make-string (string-width text) ?\=)}

$0
#+end_example

[[file:snippets/org-mode/sb][sb -- snippet]]

waiting for guidance from those more familiar with yasnippets

** REJECTED re-implement R evaluation using ess-command or ess-execute
   I don't have any complaints with the current R evaluation code or
   behaviour, but I think it would be good to use the ESS functions
   from a political point of view. Plus of course it has the normal
   benefits of an API (insulates us from any underlying changes etc). [DED]

   I'll look into this.  I believe that I looked at and rejected these
   functions initially but now I can't remember why.  I agree with
   your overall point about using API's where available.  I will take
   a look back at these and either switch to using the ess commands,
   or at least articulate under this TODO the reasons for using our
   custom R-interaction commands. [Eric]

   ess-execute

   Lets just replace =org-babel-R-input-command= with =ess-execute=.

   I tried this, and although it works in some situations, I find that
   =ess-command= will often just hang indefinitely without returning
   results.  Also =ess-execute= will occasionally hang, and pops up
   the buffer containing the results of the command's execution, which
   is undesirable.  For now these functions can not be used.  Maybe
   someone more familiar with the ESS code can recommend proper usage
   of =ess-command= or some other lower-level function which could be
   used in place of [[file:lisp/org-babel-R.el::defun%20org-babel%20R%20input%20command%20command][org-babel-R-input-command]].

*** ess functions
   
#+begin_quote ess-command
(ess-command COM &optional BUF SLEEP NO-PROMPT-CHECK)

Send the ESS process command COM and delete the output
from the ESS process buffer.  If an optional second argument BUF exists
save the output in that buffer. BUF is erased before use.
COM should have a terminating newline.
Guarantees that the value of .Last.value will be preserved.
When optional third arg SLEEP is non-nil, `(sleep-for (* a SLEEP))'
will be used in a few places where `a' is proportional to `ess-cmd-delay'.
#+end_quote

#+begin_quote ess-execute
(ess-execute COMMAND &optional INVERT BUFF MESSAGE)

Send a command to the ESS process.
A newline is automatically added to COMMAND.  Prefix arg (or second arg
INVERT) means invert the meaning of
`ess-execute-in-process-buffer'.  If INVERT is 'buffer, output is
forced to go to the process buffer.  If the output is going to a
buffer, name it *BUFF*.  This buffer is erased before use.  Optional
fourth arg MESSAGE is text to print at the top of the buffer (defaults
to the command if BUFF is not given.)
#+end_quote

*** out current setup

    1) The body of the R source code block is wrapped in a function
    2) The function is called inside of a =write.table= function call
       writing the results to a table
    3) The table is read using =org-table-import=
** DONE add =:tangle= family of header arguments
values are
- no :: don't include source-code block when tangling
- yes :: do include source-code block when tangling

this is tested in [[file:test-tangle.org::*Emacs%20Lisp%20initialization%20stuff][test-tangle.org]]

** DONE Default args
   This would be good thing to address soon. I'm imagining that
   e.g. here, the 'caller' block would return the answer 30. I believe
   there's a few issues here: i.e. the naked 'a' without a reference
   is not understood; the default arg b=6 is not understood.

#+srcname: adder(a, b=6)
#+begin_src python 
a+b
#+end_src

#+srcname: caller(var=adder(a=24))
#+begin_src python :results silent
var
#+end_src

** DONE extensible library of callable source blocks
*** Current design
    This is covered by the [[file:library-of-babel.org][Library of Babel]], which will contain
    ready-made source blocks designed to carry out useful common tasks.
*** Initial statement [Eric]
    Much of the power of org-R seems to be in it's helper functions for
    the quick graphing of tables.  Should we try to re-implement these
    functions on top of org-babel?

    I'm thinking this may be useful both to add features to org-babel-R and
    also to potentially suggest extensions of the framework.  For example
    one that comes to mind is the ability to treat a source-code block
    like a function which accepts arguments and returns results. Actually
    this can be it's own TODO (see [[* source blocks as functions][source blocks as functions]]).
*** Objectives [Dan]
    - We want to provide convenient off-the-shelf actions
      (e.g. plotting data) that make use of our new code evaluation
      environment but do not require any actual coding.
*** Initial Design proposal [Dan]
    - *Input data* will be specified using the same mechanism as :var
      references, thus the input data may come from a table, or
      another source block, and it is initially available as an elisp
      data structure.
    - We introduce a new #+ line, e.g.  #+BABELDO. C-c C-c on that
      line will apply an *action* to the referenced data.
    - *Actions correspond to source blocks*: our library of available
      actions will be a library of org-babel source blocks. Thus the
      code for executing an action, and the code for dealing with the
      output of the action will be the same code as for executing
      source blocks in general
    - Optionally, the user can have the relevant source block inserted
      into the org buffer after the (say) #+BABELDO line. This will
      allow the user to fine tune the action by modifying the code
      (especially useful for plots).
    - So maybe a #+BABELDO line will have header args
      - :data (a reference to a table or source code block)
      - :action (or should that be :srcname?) which will be something
        like :action pie-chart, referring to a source block which will
        be executed with the :data referent passed in using a :var arg.
      - :showcode or something controlling whether to show the code
      
*** Modification to design
    I'm implementing this, at least initially, as a new interpreter
    named 'babel', which has an empty body. 'babel' blocks take
    a :srcname header arg, and look for the source-code block with
    that name. They then execute the referenced block, after first
    appending their own header args on to the target block's header
    args.

    If the target block is in the library of babel (a.o.t. e.g. the
    current buffer), then the code in the block will refer to the
    input data with a name dictated by convention (e.g. __data__
    (something which is syntactically legal in all languages...). Thus
    the babel block will use a :var __data__ = whatever header arg to
    reference the data to be plotted.

** DONE Column names in R input/output
   This has been implemented: Automatic on input to R; optional in
   output. Note that this equates column names with the header row in
   an org table; whereas org actually has a mechanism whereby a row
   with a '!' in the first field defines column names. I have not
   attempted to support these org table mechanisms yet. See [[*Support%20rownames%20and%20other%20org%20babel%20table%20features][this
   DEFERRED todo item]].
** DONE use example block for large amounts of stdout output?
   We're currently `examplizing' with : at the beginning of the line,
   but should larger amounts of output be in a
   \#+begin_example...\#+end_example block? What's the cutoff? > 1
   line?  This would be nice as it would allow folding of lengthy
   output. Sometimes one will want to see stdout just to check
   everything looks OK, and then fold it away.

   I'm addressing this in branch 'examplizing-output'.
   Yea, that makes sense.  (either that or allow folding of large
   blocks escaped with =:=).

   Proposed cutoff of 10 lines, we can save this value in a user
   customizable variable.
*** DONE add ability to remove such results
** DONE exclusive =exports= params
   
#+srcname: implement-export-exclusivity
#+begin_src ruby 
:this_is_a_test
#+end_src

#+resname:
: :this_is_a_test
** DONE LoB: allow output in buffer
** DONE allow default header arguments by language
org-babel-default-header-args:lang-name

An example of when this is useful is for languages which always return
files as their results (e.g. [[*** ditaa][ditaa]], and [[*** gnuplot][gnuplot]]).
** DONE singe-function tangling and loading elisp from literate org-mode file [3/3]

This function should tangle the org-mode file for elisp, and then call
`load-file' on the resulting tangled file.

#+srcname: test-loading-embedded-emacs-lisp
#+begin_src emacs-lisp :results replace
  (setq test-tangle-advert nil)
  (setq test-tangle-loading nil)
  (setq results (list :before test-tangle-loading test-tangle-advert))
  (org-babel-load-file "test-tangle.org")
  (setq results (list (list :after test-tangle-loading test-tangle-advert) results))
  (delete-file "test-tangle.el")
  (reverse results)
#+end_src

#+resname: test-loading-embedded-emacs-lisp
| :before | nil                 | nil                                                              |
| :after  | "org-babel tangles" | "use org-babel-tangle for all your emacs initialization files!!" |

*** DONE add optional language limiter to org-babel-tangle
This should check to see if there is any need to re-export

*** DONE ensure that org-babel-tangle returns the path to the tangled file(s)

#+srcname: test-return-value-of-org-babel-tangle
#+begin_src emacs-lisp :results replace
  (mapcar #'file-name-nondirectory (org-babel-tangle-file "test-tangle.org" "emacs-lisp"))
#+end_src

#+resname:
| "test-tangle.el" |

*** DONE only tangle the file if it's actually necessary
** DONE add a function to jump to a source-block by name
   I've had an initial stab at that in org-babel-find-named-block
   (library-of-babel branch).

   At the same time I introduced org-babel-named-src-block-regexp, to
   match src-blocks with srcname.

   This is now working with the command
   `org-babel-goto-named-source-block', all we need is a good key
   binding.

** DONE add =:none= session argument (for purely functional execution) [4/4]
This would allow source blocks to be run in their own new process

- These blocks could then also be run in the background (since we can
  detach and just wait for the process to signal that it has terminated)
- We wouldn't be drowning in session buffers after running the tests
- we can re-use much of the session code to run in a more /functional/
  mode

While session provide a lot of cool features, like persistent
environments, [[* DONE function to bring up inferior-process buffer][pop-to-session]], and hints at exportation for
org-babel-tangle, they also have some down sides and I'm thinking that
session-based execution maybe shouldn't be the default behavior.

Down-sides to sessions
- *much* more complicated than functional evaluation
  - maintaining the state of the session has weird issues
  - waiting for evaluation to finish
  - prompt issues like [[* TODO weird escaped characters in shell prompt break shell evaluation][shell-prompt-escapes-bug]]
- can't run in background
- litter emacs with session buffers

*** DONE ruby

#+srcname: ruby-task-no-session
#+begin_src ruby :results replace output
puts :eric
puts :schulte
[1, 2, 3]
#+end_src

#+resname: ruby-task-no-session
| "eric"    |
| "schulte" |
*** DONE python

#+srcname: task-python-none-session
#+begin_src python :session none :results replace value
print 'something'
print 'output'
[1, 2, 3]
#+end_src

#+resname: task-python-none-session
| 1 | 2 | 3 |

*** DONE sh

#+srcname: task-session-none-sh
#+begin_src sh :results replace
echo "first"
echo "second"
#+end_src

#+resname: task-session-none-sh
| "first"  |
| "second" |

*** DONE R

#+srcname: task-no-session-R
#+begin_src R :results replace output
a <- 8
b <- 9
a + b
b - a
#+end_src

#+resname: task-no-session-R
| "[1]" | 17 |
| "[1]" |  1 |

** DONE fully purge org-babel-R of direct comint interaction
try to remove all code under the [[file:lisp/org-babel-R.el::functions%20for%20evaluation%20of%20R%20code][;; functions for evaluation of R code]] line

** DONE Create objects in top level (global) environment [5/5]
*sessions*

*** initial requirement statement [DED]
   At the moment, objects created by computations performed in the
   code block are evaluated in the scope of the
   code-block-function-body and therefore disappear when the code
   block is evaluated {unless you employ some extra trickery like
   assign('name', object, env=globalenv()) }. I think it will be
   desirable to also allow for a style wherein objects that are
   created in one code block persist in the R global environment and
   can be re-used in a separate block.

   This is what Sweave does, and while I'm not saying we have to be
   the same as Sweave, it wouldn't be hard for us to provide the same
   behaviour in this case; if we don't, we risk undeservedly being
   written off as an oddity by some.

   IOW one aspect of org-babel is that of a sort of functional
   meta-programming language. This is crazy, in a very good
   way. Nevertheless, wrt R I think there's going to be a lot of value
   in providing for a working style in which the objects are stored in
   the R session, rather than elisp/org buffer. This will be a very
   familiar working style to lots of people.

   There are no doubt a number of different ways of accomplishing
   this, the simplest being a hack like adding

#+begin_src R
for(objname in ls())
    assign(objname, get(objname), envir=globalenv())
#+end_src

to the source code block function body. (Maybe wrap it in an on.exit() call).

However this may deserve to be thought about more carefully, perhaps
with a view to having a uniform approach across languages. E.g. shell
code blocks have the same semantics at the moment (no persistence of
variables across code blocks), because the body is evaluated in a new
bash shell process rather than a running shell. And I guess the same
is true for python. However, in both these cases, you could imagine
implementing the alternative in which the body is evaluated in a
persistent interactive session. It's just that it's particularly
natural for R, seeing as both ESS and org-babel evaluate commands in a
single persistent R session.

*** sessions [Eric]

Thanks for bringing this up.  I think you are absolutely correct that we
should provide support for a persistent environment (maybe called a
*session*) in which to evaluate code blocks.  I think the current setup
demonstrates my personal bias for a functional style of programming
which is certainly not ideal in all contexts.

While the R function you mention does look like an elegant solution, I
think we should choose an implementation that would be the same across
all source code types.  Specifically I think we should allow the user to
specify an optional *session* as a header variable (when not present we
assume a default session for each language).  The session name could be
used to name a comint buffer (like the *R* buffer) in which all
evaluation would take place (within which variables would retain their
values --at least once I remove some of the functional method wrappings
currently in place-- ).

This would allow multiple environments to be used in the same buffer,
and once this setup was implemented we should be able to fairly easily
implement commands for jumping between source code blocks and the
related session buffers, as well as for dumping the last N commands from
a session into a new or existing source code block.

Please let me know if you foresee any problems with this proposed setup,
or if you think any parts might be confusing for people coming from
Sweave.  I'll hopefully find some time to work on this later in the
week.

*** can functional and interpreted/interactive models coexist?

Even though both of these use the same =*R*= buffer the value of =a=
is not preserved because it is assigned inside of a functional
wrapper.

#+srcname: task-R-sessions
#+begin_src R 
a <- 9
b <- 21
a + b
#+end_src

#+srcname: task-R-same-session
#+begin_src R 
a
#+end_src

This functional wrapper was implemented in order to efficiently return
the results of the execution of the entire source code block.  However
it inhibits the evaluation of source code blocks in the top level,
which would allow for persistence of variable assignment across
evaluations.  How can we allow *both* evaluation in the top level, and
efficient capture of the return value of an entire source code block
in a language independent manner?

Possible solutions...
1) we can't so we will have to implement two types of evaluation
   depending on which is appropriate (functional or imperative)
2) we remove the functional wrapper and parse the source code block
   into it's top level statements (most often but not always on line
   breaks) so that we can isolate the final segment which is our
   return value.
3) we add some sort of "#+return" line to the code block
4) we take advantage of each languages support for meta-programming
   through =eval= type functions, and use said to evaluate the entire
   blocks in such a way that their environment can be combined with the
   global environment, and their results are still captured.
5) I believe that most modern languages which support interactive
   sessions have support for a =last_result= type function, which
   returns the result of the last input without re-calculation.  If
   widely enough present this would be the ideal solution to a
   combination of functional and imperative styles.

None of these solutions seem very desirable, but for now I don't see
what else would be possible.

Of these options I was leaning towards (1) and (4) but now believe
that if it is possible option (5) will be ideal.

**** (1) both functional and imperative evaluation
Pros
- can take advantage of built in functions for sending regions to the
  inferior process
- retains the proven tested and working functional wrappers

Cons
- introduces the complication of keeping track of which type of
  evaluation is best suited to a particular context
- the current functional wrappers may require some changes in order to
  include the existing global context

**** (4) exploit language meta-programming constructs to explicitly evaluate code
Pros
- only one type of evaluation

Cons
- some languages may not have sufficient meta-programming constructs

**** (5) exploit some =last_value= functionality if present

Need to ensure that most languages have such a function, those without
will simply have to implement their own similar solution...

| language   | =last_value= function       |
|------------+-----------------------------|
| R          | .Last.value                 |
| ruby       | _                           |
| python     | _                           |
| shell      | see [[* last command for shells][last command for shells]] |
| emacs-lisp | see [[* emacs-lisp will be a special case][special-case]]            |

#+srcname: task-last-value
#+begin_src ruby
82 + 18
#+end_src

***** last command for shells
Do this using the =tee= shell command, and continually pipe the output
to a file.

Got this idea from the following [[http://linux.derkeiler.com/Mailing-Lists/Fedora/2004-01/0898.html][email-thread]].

suggested from mailing list

#+srcname: bash-save-last-output-to-file
#+begin_src sh 
while read line 
do 
  bash -c "$line" | tee /tmp/last.out1 
  mv /tmp/last.out1 /tmp/last.out 
done
#+end_src

another proposed solution from the above thread

#+srcname: bash-save-in-variable
#+begin_src sh 
#!/bin/bash 
# so - Save Output. Saves output of command in OUT shell variable. 
OUT=`$*` 
echo $OUT 
#+end_src

and another

#+begin_quote
.inputrc: 
"^[k": accept-line 
"^M": " | tee /tmp/h_lastcmd.out ^[k" 

.bash_profile: 
export __=/tmp/h_lastcmd.out 

If you try it, Alt-k will stand for the old Enter; use "command $__" to 
access the last output. 

Best, 

--

Herculano de Lima Einloft Neto
#+end_quote

***** emacs-lisp will be a special case
While it is possible for emacs-lisp to be run in a console type
environment (see the =elim= function) it is *not* possible to run
emacs-lisp in a different *session*.  Meaning any variable set top
level of the console environment will be set *everywhere* inside
emacs.  For this reason I think that it doesn't make any sense to
worry about session support for emacs-lisp.

*** Further thoughts on 'scripting' vs. functional approaches

    These are just thoughts, I don't know how sure I am about this.
    And again, perhaps I'm not saying anything very radical, just that
    it would be nice to have some options supporting things like
    receiving text output in the org buffer.

    I can see that you've already gone some way down the road towards
    the 'last value' approach, so sorry if my comments come rather
    late. I am concerned that we are not giving sufficient attention
    to stdout / the text that is returned by the interpreters. In
    contrast, many of our potential users will be accustomed to a
    'scripting' approach, where they are outputting text at various
    points in the code block, not just at the end. I am leaning
    towards thinking that we should have 2 modes of evaluation:
    'script' mode, and 'functional' mode.

    In script mode, evaluation of a code block would result in *all*
    text output from that code block appearing as output in the org
    buffer, presumably as an #+begin_example...#+end_example. There
    could be an :echo option controlling whether the input commands
    also appear in the output. [This is like Sweave].

    In functional mode, the *result* of the code block is available as
    an elisp object, and may appear in the org buffer as an org
    table/string, via the mechanisms you have developed already.

    One thing I'm wondering about is whether, in script mode, there
    simply should not be a return value. Perhaps this is not so
    different from what exists: script mode would be new, and what
    exists currently would be functional mode.

    I think it's likely that, while code evaluation will be exciting
    to people, a large majority of our users in a large majority of
    their usage will not attempt to actually use the return value from
    a source code block in any meaningful way. In that case, it seems
    rather restrictive to only allow them to see output from the end
    of the code block.

    Instead I think the most accessible way to introduce org-babel to
    people, at least while they are learning it, is as an immensely
    powerful environment in which to embed their 'scripts', which now
    also allows them to 'run' their 'scripts'. Especially as such
    people are likely to be the least capable of the user-base, a
    possible design-rule would be to make the scripting style of usage
    easy (default?), perhaps requiring a special option to enable a
    functional style. Those who will use the functional style won't
    have a problem understanding what's going on, whereas the 'skript
    kiddies' might not even know the syntax for defining a function in
    their language of choice. And of course we can allow the user to
    set a variable in their .emacs controlling the preference, so that
    functional users are not inconveniennced by having to provide
    header args the whole time.

    Please don't get the impression that I am down-valuing the
    functional style of org-babel. I am constantly horrified at the
    messy 'scripts' that my colleagues produce in perl or R or
    whatever! Nevertheless that seems to be how a lot of people work.
    
    I think you were leaning towards the last-value approach because
    it offered the possibility of unified code supporting both the
    single evaluation environment and the functional style. If you
    agree with any of the above then perhaps it will impact upon this
    and mean that the code in the two branches has to differ a bit. In
    that case, functional mode could perhaps after all evaluate each
    code block in its own environment, thus (re)approaching 'true'
    functional programming (side-effects are hard to achieve).

#+begin_src sh
ls > files
echo "There are `wc -l files` files in this directory"

#+end_src

*** even more thoughts on evaluation, results, models and options

Thanks Dan, These comments are invaluable.

What do you think about this as a new list of priorities/requirements
for the execution of source-code blocks.

- Sessions
   1)  we want the evaluation of the source code block to take place in a
       session which can persist state (variables, current directory,
       etc...).
   2)  source code blocks can specify their session with a header argument
   3)  each session should correspond to an Emacs comint buffer so that the
       user can drop into the session and experiment with live code
       evaluation.
- Results
  1) each source-code block generates some form of results which (as
     we have already implemented) is transfered into emacs-lisp
     after which it can be inserted into the org-mode buffer, or
     used by other source-code blocks
  2) when the results are translated into emacs-lisp, forced to be
     interpreted as a scalar (dumping their raw values into the
     org-mode buffer), as a vector (which is often desirable with R
     code blocks), or interpreted on the fly (the default option).
     Note that this is very nearly currently implemented through the
     [[* DONE results-type header (vector/file)][results-type-header]].
  3) there should be *two* means of collecting results from the
     execution of a source code block.  *Either* the value of the
     last statement of the source code block, or the collection of
     all that has been passed to STDOUT during the evaluation.

**** header argument or return line (*header argument*)

   Rather than using a header argument to specify how the return value
   should be passed back, I'm leaning towards the use of a =#+RETURN=
   line inside the block.  If such a line *is not present* then we
   default to using STDOUT to collect results, but if such a line *is
   present* then we use it's value as the results of the block.  I
   think this will allow for the most elegant specification between
   functional and script execution.  This also cleans up some issues
   of implementation and finding which statement is the last
   statement.

   Having given this more thought, I think a header argument is
   preferable.  The =#+return:= line adds new complicating syntax for
   something that does little more than we would accomplish through
   the addition of a header argument.  The only benefit being that we
   know where the final statement starts, which is not an issue in
   those languages which contain 'last value' operators.

   new header =:results= arguments
   - script :: explicitly states that we want to use STDOUT to
               initialize our results
   - return_last :: stdout is ignored instead the *value* of the final
                    statement in the block is returned
   - echo :: means echo the contents of the source-code block along
             with the results (this implies the *script* =:results=
             argument as well)

*** DONE rework evaluation lang-by-lang [4/4]

This should include...
- functional results working with the comint buffer
- results headers
  - script :: return the output of STDOUT
    - write a macro which runs the first redirection, executes the
      body, then runs the second redirection
  - last :: return the value of the last statement
    - 

- sessions in comint buffers

**** DONE Ruby [4/4]
- [X] functional results working with comint
- [X] script results
- [X] ensure scalar/vector results args are taken into consideration
- [X] ensure callable by other source block

#+srcname: ruby-use-last-output
#+begin_src ruby :results replace
a = 2
b = 4
c = a + b
[a, b, c, 78]
#+end_src

#+resname: ruby-use-last-output
| 2 | 4 | 6 | 78 |

#+srcname: task-call-use-last-output
#+begin_src ruby :var last=ruby-use-last-output :results replace
last.flatten.size + 1
#+end_src

#+resname: task-call-use-last-output
: 5

***** ruby sessions

#+srcname: first-ruby-session-task
#+begin_src ruby :session schulte :results silent
schulte = 27
#+end_src

#+srcname: second-ruby-session-task
#+begin_src ruby :session schulte :results silent
schulte + 3
#+end_src

#+srcname: without-the-right-session
#+begin_src ruby :results silent
schulte
#+end_src

**** DONE R [4/4]

- [X] functional results working with comint
- [X] script results
- [X] ensure scalar/vector results args are taken into consideration
- [X] ensure callable by other source block

To redirect output to a file, you can use the =sink()= command.

#+srcname: task_R_B
#+begin_src R :results value vector silent
a <- 9
b <- 10
b - a
a + b
#+end_src

#+srcname: task-R-use-other-output
#+begin_src R :var twoentyseven=task_R_B() :results replace value
83
twoentyseven + 9
#+end_src

#+resname: task-R-use-other-output
: 28

**** DONE Python [4/4]
- [X] functional results working with comint
- [X] script results
- [X] ensure scalar/vector results args are taken into consideration
- [X] ensure callable by other source block

#+srcname: task-new-eval-for-python
#+begin_src python :results silent output scalar
8
9
10
#+end_src

#+srcname: task-use-new-eval
#+begin_src python :var tasking=task-new-eval-for-python() :results replace
tasking + 2
#+end_src

#+resname: task-use-new-eval
: 12

**** DONE Shells [4/4]
- [X] functional results working with comint
- [X] script results
- [X] ensure scalar/vector results args are taken into consideration
- [X] ensure callable by other source block

#+srcname: task-shell-new-evaluation
#+begin_src sh :results silent value scalar
echo 'eric'
date
#+end_src

#+srcname: task-call-other-shell
#+begin_src sh :var other=task-shell-new-evaluation() :results replace  scalar
echo $other ' is the old date'
#+end_src

#+resname: task-call-other-shell
: $ Fri Jun 12 13:08:37 PDT 2009  is the old date

*** DONE implement a *session* header argument [4/4]
=:session= header argument to override the default *session* buffer

**** DONE ruby

#+srcname: task-ruby-named-session
#+begin_src ruby :session schulte :results replace
schulte = :in_schulte
#+end_src

#+resname: task-ruby-named-session
: :in_schulte

#+srcname: another-in-schulte
#+begin_src ruby :session schulte 
schulte
#+end_src

#+resname: another-in-schulte
: :in_schulte
: :in_schulte
: :in_schulte

**** DONE python

#+srcname: python-session-task
#+begin_src python :session what :results silent
what = 98
#+end_src

#+srcname: python-get-from-session
#+begin_src python :session what :results replace
what
#+end_src

#+resname: python-get-from-session
: 98

**** DONE shell

#+srcname: task-shell-sessions
#+begin_src sh :session what
WHAT='patton'
#+end_src

#+srcname: task-shell-sessions-what
#+begin_src sh :session what :results replace
echo $WHAT
#+end_src

#+resname: task-shell-sessions-what
: patton

**** DONE R

#+srcname: task-R-session
#+begin_src R :session what :results replace
a <- 9
b <- 8
a + b
#+end_src

#+resname: task-R-session
: 17

#+srcname: another-task-R-session
#+begin_src R :session what :results replace
a + b
#+end_src

*** DONE function to bring up inferior-process buffer [4/4]

This should be callable from inside of a source-code block in an
org-mode buffer.  It should evaluate the header arguments, then bring
up the inf-proc buffer using =pop-to-buffer=.

For lack of a better place, lets add this to the `org-metadown-hook'
hook.

To give this a try, place the cursor on a source block with variables,
(optionally git a prefix argument) then hold meta and press down.

**** DONE ruby

#+srcname: task-ruby-pop-to-session
#+begin_src ruby :var num=9 :var another="something else"
num.times{|n| puts another}
#+end_src

**** DONE python

#+srcname: task-python-pop-to-session
#+begin_src python :var num=9 :var another="something else"
another * num
#+end_src
**** DONE R

#+srcname: task-R-pop-to-session
#+begin_src R :var a=9 :var b=8
a * b
#+end_src

**** DONE shell

#+srcname: task-shell-pop-sessions
#+begin_src sh :var NAME="eric"
echo $NAME
#+end_src

*** DEFERRED function to dump last N lines from inf-proc buffer into the current source block

Callable with a prefix argument to specify how many lines should be
dumped into the source-code buffer.

*** REJECTED comint notes

Implementing comint integration in [[file:lisp/org-babel-comint.el][org-babel-comint.el]].

Need to have...
- handling of outputs
  - split raw output from process by prompts
  - a ring of the outputs, buffer-local, `org-babel-comint-output-ring'
  - a switch for dumping all outputs to a buffer
- inputting commands

Lets drop all this language specific stuff, and just use
org-babel-comint to split up our outputs, and return either the last
value of an execution or the combination of values from the
executions.

**** comint filter functions
: ;;  comint-input-filter-functions     hook    process-in-a-buffer
: ;;  comint-output-filter-functions    hook    function modes.
: ;;  comint-preoutput-filter-functions   hook
: ;;  comint-input-filter                       function ...

#+srcname: obc-filter-ruby
#+begin_src ruby :results last
1
2
3
4
5
#+end_src

** DONE Remove protective commas from # comments before evaluating
   org inserts protective commas in front of ## comments in language
   modes that use them. We need to remove them prior to sending code
   to the interpreter.

#+srcname: testing-removal-of-protective-comas
#+begin_src ruby
,# this one might break it??
:comma_protection
#+end_src

** DONE pass multiple reference arguments into R
   Can we do this? I wasn't sure how to supply multiple 'var' header
   args. Just delete this if I'm being dense.

   This should be working, see the following example...

#+srcname: two-arg-example
#+begin_src R :var n=2 :var m=8
n + m
#+end_src

#+resname: two-arg-example
: 10

** DONE ensure that table ranges work
when a table range is passed to org-babel as an argument, it should be
interpreted as a vector.

| 1 | 2 | simple       |
| 2 | 3 | Fixnum:1     |
| 3 | 4 | Array:123456 |
| 4 | 5 |              |
| 5 | 6 |              |
| 6 | 7 |              |
#+TBLFM: @1$3='(sbe simple-sbe-example (n 4))::@2$3='(sbe task-table-range (n @1$1..@6$1))::@3$3='(sbe task-table-range (n (@1$1..@6$1)))

#+srcname: simple-sbe-example
#+begin_src emacs-lisp 
"simple"
#+end_src

#+srcname: task-table-range
#+begin_src ruby :var n=simple-sbe-example
"#{n.class}:#{n}"
#+end_src

#+srcname: simple-results
#+begin_src emacs-lisp :var n=task-table-range(n=(1 2 3))
n
#+end_src

#+resname: simple-results
: Array:123

#+srcname: task-arr-referent
#+begin_src ruby :var ar=(1 2 3)
ar.size
#+end_src

#+resname: task-arr-referent
: 3

** DONE global variable indicating default to vector output
how about an alist... =org-babel-default-header-args= this may already
exist... just execute the following and all source blocks will default
to vector output

#+begin_src emacs-lisp 
(setq org-babel-default-header-args '((:results . "vector")))
#+end_src

** DONE name named results if source block is named
currently this isn't happening although it should be

#+srcname: test-naming-named-source-blocks
#+begin_src emacs-lisp 
:namer
#+end_src

#+resname: test-naming-named-source-blocks
: :namer
** DONE (simple caching) check for named results before source blocks
see the TODO comment in [[file:lisp/org-babel-ref.el::TODO%20This%20should%20explicitly%20look%20for%20resname%20lines%20before][org-babel-ref.el#org-babel-ref-resolve-reference]]
** DONE set =:results silent= when eval with prefix argument

#+begin_src emacs-lisp
'silentp
#+end_src
** DONE results-type header (vector/file) [3/3]
   In response to a point in Dan's email.  We should allow the user to
   force scalar or vector results.  This could be done with a header
   argument, and the default behavior could be controlled through a
   configuration variable.
   
#+srcname: task-trivial-vector
#+begin_src ruby :results replace vector
:scalar
#+end_src

#+resname:
| ":scalar" |

   since it doesn't make sense to turn a vector into a scalar, lets
   just add a two values...
   
   - vector :: forces the results to be a vector (potentially 1 dimensional)
   - file :: this throws an error if the result isn't a string, and
             tries to treat it as a path to a file.

   I'm just going to cram all of these into the =:results= header
   argument.  Then if we allow multiple header arguments it should
   work out, for example one possible header argument string could be
   =:results replace vector file=, which would *replace* any existing
   results forcing the results into an org-mode table, and
   interpreting any strings as file paths.

*** DONE multiple =:results= headers

#+srcname: multiple-result-headers
#+begin_src ruby :results replace silent
:schulte
#+end_src

#+resname:

*** DONE file result types
When inserting into an org-mode buffer create a link with the path
being the value, and optionally the display being the
=file-name-nondirectory= if it exists.

#+srcname: task-file-result
#+begin_src python :results replace file
"something"
#+end_src

#+resname:
[[something][something]]


This will be useful because blocks like =ditaa= and =dot= can return
the string path of their files, and can add =file= to their results
header.

*** DONE vector result types

#+srcname: task-force-results
#+begin_src emacs-lisp :results vector
8
#+end_src

#+resname:
| 8 |

** DONE results name
    In order to do this we will need to start naming our results.
    Since the source blocks are named with =#+srcname:= lines we can
    name results with =#+resname:= lines (if the source block has no
    name then no name is given to the =#+resname:= line on creation,
    otherwise the name of the source block is used).

    This will have the additional benefit of allowing results and
    source blocks to be located in different places in a buffer (and
    eventually in different buffers entirely).

#+srcname: developing-resnames
#+begin_src emacs-lisp  :results silent
'schulte
#+end_src

    Once source blocks are able to find their own =#+resname:= lines
    we then need to...

#+srcname: sbe-w-new-results
#+begin_src emacs-lisp :results replace
(sbe "developing-resnames")
#+end_src

#+resname:
: schulte

*** TODO change the results insertion functions to use these lines

*** TODO teach references to resolve =#+resname= lines.

** DONE org-babel tests org-babel [1/1]
since we are accumulating this nice collection of source-code blocks
in the sandbox section we should make use of them as unit tests.
What's more, we should be able to actually use org-babel to run these
tests.

We would just need to cycle over every source code block under the
sandbox, run it, and assert that the return value is equal to what we
expect.

I have the feeling that this should be possible using only org-babel
functions with minimal or no additional elisp.  It would be very cool
for org-babel to be able to test itself.

This is now done, see [[* Tests]].

*** DEFERRED org-babel assertions (may not be necessary)
These could be used to make assertions about the results of a
source-code block.  If the assertion fails then the point could be
moved to the block, and error messages and highlighting etc... could
ensue

** DONE make C-c C-c work anywhere within source code block?
   This seems like it would be nice to me, but perhaps it would be
   inefficient or ugly in implementation? I suppose you could search
   forward, and if you find #+end_src before you find #+begin_src,
   then you're inside one. [DED]

   Agreed, I think inside of the =#+srcname: line= would be useful as
   well.

#+srcname: testing-out-cc
#+begin_src emacs-lisp
'schulte
#+end_src

** DONE integration with org tables
We should make it easy to call org-babel source blocks from org-mode
table formulas.  This is practical now that it is possible to pass
arguments to org-babel source blocks.

See the related [[* (sandbox) integration w/org tables][sandbox]] header for tests/examples.

*** digging in org-table.el
In the past [[file:~/src/org/lisp/org-table.el::org%20table%20el%20The%20table%20editor%20for%20Org%20mode][org-table.el]] has proven difficult to work with.

Should be a hook in [[file:~/src/org/lisp/org-table.el::defun%20org%20table%20eval%20formula%20optional%20arg%20equation][org-table-eval-formula]].

Looks like I need to change this [[file:~/src/org/lisp/org-table.el::if%20lispp][if statement]] (line 2239) into a cond
expression.

** DONE source blocks as functions

Allow source code blocks to be called like functions, with arguments
specified.  We are already able to call a source-code block and assign
it's return result to a variable.  This would just add the ability to
specify the values of the arguments to the source code block assuming
any exist.  For an example see 

When a variable appears in a header argument, how do we differentiate
between it's value being a reference or a literal value?  I guess this
could work just like a programming language.  If it's escaped or in
quotes, then we count it as a literal, otherwise we try to look it up
and evaluate it.

** DONE folding of code blocks? [2/2]
   [DED] In similar way to using outline-minor-mode for folding function
   bodies, can we fold code blocks?  #+begin whatever statements are
   pretty ugly, and in any case when you're thinking about the overall
   game plan you don't necessarily want to see the code for each Step.

*** DONE folding of source code block
    Sounds good, and wasn't too hard to implement.  Code blocks should
    now be fold-able in the same manner as headlines (by pressing TAB
    on the first line).

*** REJECTED folding of results
    So, lets do a three-stage tab cycle... First fold the src block,
    then fold the results, then unfold.
    
    There's no way to tell if the results are a table or not w/o
    actually executing the block which would be too expensive of an
    operation.

** DONE selective export of text, code, figures
   [DED] The org-babel buffer contains everything (code, headings and
   notes/prose describing what you're up to, textual/numeric/graphical
   code output, etc). However on export to html / LaTeX one might want
   to include only a subset of that content. For example you might
   want to create a presentation of what you've done which omits the
   code.

   [EMS] So I think this should be implemented as a property which can
   be set globally or on the outline header level (I need to review
   the mechanics of org-mode properties).  And then as a source block
   header argument which will apply only to a specific source code
   block.  A header argument of =:export= with values of
   
   - =code= :: just show the code in the source code block
   - =none= :: don't show the code or the results of the evaluation
   - =results= :: just show the results of the code evaluation (don't
                  show the actual code)
   - =both= :: show both the source code, and the results

this will be done in [[* (sandbox) selective export][(sandbox) selective export]].

** DONE a header argument specifying silent evaluation (no output)
This would be useful across all types of source block.  Currently
there is a =:replace t= option to control output, this could be
generalized to an =:output= option which could take the following
options (maybe more)

- =t= :: this would be the default, and would simply insert the
         results after the source block
- =replace= :: to replace any results which may already be there
- =silent= :: this would inhibit any insertion of the results

This is now implemented see the example in the [[* silent evaluation][sandbox]]

** DONE assign variables from tables in R
This is now working (see [[* (sandbox table) R][(sandbox-table)-R]]).  Although it's not that
impressive until we are able to print table results from R.

** DONE insert 2-D R results as tables
everything is working but R and shell

*** DONE shells

*** DONE R

This has already been tackled by Dan in [[file:existing_tools/org-R.el::defconst%20org%20R%20write%20org%20table%20def][org-R:check-dimensions]].  The
functions there should be useful in combination with [[http://cran.r-project.org/doc/manuals/R-data.html#Export-to-text-files][R-export-to-csv]]
as a means of converting multidimensional R objects to emacs lisp.

It may be as simple as first checking if the data is multidimensional,
and then, if so using =write= to write the data out to a temporary
file from which emacs can read the data in using =org-table-import=.

Looking into this further, is seems that there is no such thing as a
scalar in R [[http://tolstoy.newcastle.edu.au/R/help/03a/3733.html][R-scalar-vs-vector]]  In that light I am not sure how to
deal with trivial vectors (scalars) in R.  I'm tempted to just treat
them as vectors, but then that would lead to a proliferation of
trivial 1-cell tables...

** DONE allow variable initialization from source blocks
Currently it is possible to initialize a variable from an org-mode
table with a block argument like =table=sandbox= (note that the
variable doesn't have to named =table=) as in the following example

#+TBLNAME: sandbox
| 1 |       2 | 3 |
| 4 | schulte | 6 |

#+begin_src emacs-lisp :var table=sandbox :results replace
(message (format "table = %S" table))
#+end_src

: "table = ((1 2 3) (4 \"schulte\" 6))"

It would be good to allow initialization of variables from the results
of other source blocks in the same manner.  This would probably
require the addition of =#+SRCNAME: example= lines for the naming of
source blocks, also the =table=sandbox= syntax may have to be expanded
to specify whether the target is a source code block or a table
(alternately we could just match the first one with the given name
whether it's a table or a source code block).

At least initially I'll try to implement this so that there is no need
to specify whether the reference is to a table or a source-code block.
That seems to be simpler both in terms of use and implementation.

This is now working for emacs-lisp, ruby and python (and mixtures of
the three) source blocks.  See the examples in the [[* (sandbox) referencing other source blocks][sandbox]].

This is currently working only with emacs lisp as in the following
example in the [[* emacs lisp source reference][emacs lisp source reference]].


** TODO Add languages [2/6]
I'm sure there are many more that aren't listed here.  Please add
them, and bubble any that you particularly care about up to the top.

Any new language should be implemented in a org-babel-lang.el file.
Follow the pattern set by [[file:lisp/org-babel-script.el][org-babel-script.el]], [[file:lisp/org-babel-shell.el][org-babel-shell.el]] and
[[file:lisp/org-babel-R.el][org-babel-R.el]].

*** TODO perl
This could probably be added to [[file:lisp/org-babel-script.el][org-babel-script.el]]

*** TODO java

*** DONE ditaa
(see [[* file result types][file result types]])

#+srcname: implementing-ditaa
#+begin_src ditaa :results replace :file blue.png :cmdline -r
+---------+
| cBLU    |
|         |
|    +----+
|    |cPNK|
|    |    |
+----+----+
#+end_src

#+resname: implementing-ditaa
[[file:blue.png][blue.png]]

*** DONE gnuplot [7/7]
(see [[* file result types][file result types]])

#+PLOT: title:"Citas" ind:1 deps:(3) type:2d with:histograms set:"yrange [0:]"
#+TBLNAME: gnuplot-data
| independent var | first dependent var | second dependent var |
|-----------------+---------------------+----------------------|
|             0.1 |               0.425 |                0.375 |
|             0.2 |              0.3125 |               0.3375 |
|             0.3 |          0.24999993 |           0.28333338 |
|             0.4 |               0.275 |              0.28125 |
|             0.5 |                0.26 |                 0.27 |
|             0.6 |          0.25833338 |           0.24999993 |
|             0.7 |          0.24642845 |           0.23928553 |
|             0.8 |             0.23125 |               0.2375 |
|             0.9 |          0.23333323 |            0.2333332 |
|               1 |              0.2225 |                 0.22 |
|             1.1 |          0.20909075 |           0.22272708 |
|             1.2 |          0.19999998 |           0.21458333 |
|             1.3 |          0.19615368 |           0.21730748 |

#+srcname: implementing-gnuplot
#+begin_src gnuplot :var data=gnuplot-data :results silent
set title "Implementing Gnuplot"
plot data using 1:2 with lines
#+end_src

**** DONE add variables
     gnuplot 4.2 and up support user defined variables.  This is how
     we will handle variables with org-babel (meaning we will need to
     require gnuplot 4.2 and up for variable support, which can be
     install using [[http://www.macports.org/install.php][macports]] on Mac OSX).

     - scalar variables should be replaced in the body of the gnuplot code
     - vector variables should be exported to tab-separated files, and
       the variable names should be replaced with the path to the files

**** DONE direct plotting w/o session
**** DEFERRED gnuplot support for column/row names
This should be implemented along the lines of the [[* STARTED Column (and row) names of tables in R input/output][R-colname-support]].

We can do something similar to the :labels param in org-plot, we just
have to be careful to ensure that each label is aligned with the
related data file.

This may be walking too close to an entirely prebuilt plotting tool
rather than straight gnuplot code evaluation.  For now I think this
can wait.

**** DONE a =file= header argument
to specify a file holding the results

#+srcname: gnuplot-to-file-implementation
#+begin_src gnuplot :file plot.png :var data=gnuplot-data
plot data using 1:2, data using 1:3 with lines
#+end_src

#+resname:
[[file:plot.png][plot.png]]

**** DONE helpers from org-plot.el
There are a variety of helpers in org-plot which can be fit nicely
into custom gnuplot header arguments.

These should all be in place by now.

**** DEFERRED header argument specifying 3D data

#+tblname: org-grid
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 0 | 0 | 1 | 1 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 | 1 | 1 | 1 | 0 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 1 | 1 | 1 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 0 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 0 | 1 | 0 | 0 | 0 | 1 | 0 |
| 0 | 1 | 0 | 0 | 1 | 0 | 1 | 1 | 0 | 1 | 0 | 1 | 0 | 0 | 0 | 1 | 0 |
| 0 | 0 | 1 | 1 | 0 | 0 | 1 | 0 | 1 | 1 | 0 | 0 | 1 | 1 | 1 | 0 | 0 |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |
| 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 | 0 |

#+srcname: implementing-gnuplot-grid-plots
#+begin_src gnuplot :vars data=org-grid

#+end_src

**** DONE gnuplot sessions
Working on this, we won't support multiple sessions as `gnuplot-mode'
isn't setup for such things.

Also we can't display results with the default :none session, so for
gnuplot we really want the default behavior to be :default, and to
only run a :none session when explicitly specified.

#+srcname: implementing-gnuplot-sessions
#+begin_src gnuplot :var data=gnuplot-data :session none :file session.png
set title "Implementing Gnuplot Sessions"
plot data using 1:2 with lines
#+end_src

#+resname:
[[file:session.png][session.png]]

*** TODO dot
(see [[* file result types][file result types]])

*** TODO asymptote
(see [[* file result types][file result types]])


* Bugs [25/36]
** TODO allow srcname to omit function call parentheses
   Someone needs to revisit those regexps. Is there an argument for
   moving some of the regexps used to match function calls into
   defvars? (i.e. in o-b.el and o-b-ref.el)

** TODO creeping blank lines
   There's still inappropriate addition of blank lines in some circumstances. 

   Hmm, it's a bit confusing. It's to do with o-b-remove-result. LoB
   removes the entire (#+resname and result) and starts from scratch,
   whereas #+begin_src only removes the result. I haven't worked out
   what the correct fix is yet. Maybe the right thing to do is to make
   sure that those functions (o-b-remove-result et al.) are neutral
   with respect to newlines. Sounds easy, but...

   E.g.

#+begin_src sh
b=5
#+end_src



   Compare the results of
#+lob: python-add(a=5, b=17)

#+resname: python-add(a=5, b=17)
: 22
--------------------------------

#+begin_src python
23
#+end_src

#+resname:
: 23
---------------------
** TODO problem with newlines in output when :results value
#+begin_src python :results value
'\n'.join(map(str, range(4)))
#+end_src

#+resname:
: 0

Whereas I was hoping for

| 0 |
| 1 |
| 2 |
| 3 |

This is some sort of non-printing char / quoting issue I think. Note
that

#+begin_src python :results value
'\\n'.join(map(str, range(4)))
#+end_src

#+resname:
: 0\n1\n2\n3

Also, note that
#+begin_src python :results output
print('\n'.join(map(str, range(4))))
#+end_src

#+resname:
: 0
: 1
: 2
: 3

*** collapsing consecutive newlines in string output
    
    This is an example of the same bug

#+srcname: multi-line-string-output
#+begin_src ruby :results output
"the first line ends here


     and this is the second one

even a third"
#+end_src

This doesn't produce anything at all now. I believe that's because
I've changed things so that :results output really does *not* get the
value of the block, only the STDOUT. So if we add a print statement
this works OK.

#+srcname: multi-line-string-output
#+begin_src ruby :results output
print "the first line ends here


     and this is the second one

even a third"
#+end_src

#+resname:
: the first line ends here
: 
: 
:      and this is the second one
: 
: even a third

However, the behaviour with :results value is wrong

#+srcname: multi-line-string-value
#+begin_src ruby
"the first line ends here


     and this is the second one

even a third"
#+end_src

#+resname:
: 0

** TODO prompt characters appearing in output with R
#+begin_src R :session *R* :results output
  x <- 6
  y <- 8
  3
#+end_src      

#+resname:
: > [1] 3

** TODO o-b-execute-subtree overwrites heading when subtree is folded
*** Example
    Try M-x org-babel-execute-subtree with the subtree folded and
    point at the beginning of the heading line.
#+begin_src sh
size=5
#+end_src

** TODO Allow source blocks to be recognised when #+ are not first characters on the line
   I think Carsten has recently altered the core so that #+ can have
   preceding whitespace, at least for literal/code examples. org-babel
   should support this.

** TODO non-orgtbl formatted lists
for example

#+srcname: this-doesn't-match-orgtbl
#+begin_src emacs-lisp :results replace
'((:results . "replace"))
#+end_src

#+resname: this-doesn't-match-orgtbl
   
** PROPOSED external shell execution can't isolate return values
I have no idea how to do this as of yet.  The result is that when
shell functions are run w/o a session there is no difference between
the =output= and =value= result arguments.

Yea, I don't know how to do this either.  I searched extensively on
how to isolate the *last* output of a series of shell commands (see
[[* last command for
 shells][last command for shells]]).  The results of the search were basically
that it was not possible (or at least not accomplish-able with a
reasonable amount of effort).

That fact combined with the tenancy to all ways use standard out in
shell scripts led me to treat these two options (=output= and =value=)
as identical in shell evaluation.  Not ideal but maybe good enough for
the moment.

In the `results' branch I've changed this so that they're not quite
identical: output results in raw stdout contents, whereas value
converts it to elisp, perhaps to a table if it looks tabular. This is
the same for the other languages. [Dan]

** TODO are the org-babel-trim s necessary?
   at the end of e.g. org-babel-R-evaluate, org-babel-python-evaluate, but
   not org-babel-ruby-evaluate
** TODO use new merge function [[file:lisp/org-babel-ref.el::t%20nil%20org%20combine%20plists%20args%20nil][here]]?
   And at other occurrences of org-combine-plists?
** TODO LoB is not populated on startup
   org-babel-library-of-babel is nil for me on startup. I have to
   evaluate the [[file:lisp/org-babel-lob.el::][org-babel-lob-ingest]] line manually.
** DONE Fix nested evaluation
   The current parser / evaluator fails with greater levels of nested
   function block calls (example below).

*** Initial statement [ded]
    If we want to overcome this I think we'd have to redesign some of
    the evaluation mechanism. Seeing as we are also facing issues like
    dealing with default argument values, and seeing as we now know
    how we want the library of babel to behave in addition to the
    source blocks, now might be a good time to think about this. It
    would be nice to do the full thing at some point, but otoh we may
    not consider it a massive priority.
    
    AIui, there are two stages: (i) construct a parse tree, and (ii)
    evaluate it and return the value at the root. In the parse tree
    each node represents an unevaluated value (either a literal value
    or a reference). Node v may have descendent nodes, which represent
    values upon which node v's evaluation depends. Once that tree is
    constructed, then we evaluate the nodes from the tips towards the
    root (a post-order traversal).

    [This would also provide a solution for concatenating the STDOUTs
    of called blocks, which is a [[*allow%20output%20mode%20to%20return%20stdout%20as%20value][task below]]; we concatenate them in
    whatever order the traversal is done in.]

    In addition to the variable references (i.e. daughter nodes), each
    node would contain the information needed to evaluate that node
    (e.g. lang body). Then we would pass a function postorder over the
    tree which would call o-b-execute-src-block at each node, finally
    returning the value at the root.

    Fwiw I made a very tentative small start at stubbing this out in
    org-babel-call.el in the 'evaluation' branch. And I've made a start
    at sketching a parsing algorithm below.
**** Parse tree algorithm
    Seeing as we're just trying to parse a string like
    f(a=1,b=g(c=2,d=3)) it shouldn't be too hard. But of course there
    are 'proper' parsers written in elisp out there,
    e.g. [[http://cedet.sourceforge.net/semantic.shtml][Semantic]]. Perhaps we can find what we need -- our syntax is
    pretty much the same as python and R isn't it?

    Or, a complete hack, but maybe it would be we easy to transform it
    to XML and then parse that with some existing tool?
    
    But if we're doing it ourselves, something very vaguely like this?
    (I'm sure there're lots of problems with this)

#+srcname: org-babel-call-parse(call)
#+begin_src python 
  ## we are currently reading a reference name: the name of the root function
  whereami = "refname"
  node = root = Node()
  for c in call_string:
      if c == '(':
          varnum = 0
          whereami = "varname" # now we're reading a variable name
      if c == '=':
          new = Node()
          node.daughters = [node.daughters, new]
          new.parent = node
          node = new
          whereami = "refname"
      if c == ',':
          whereami = "varname"
          varnum += 1
      elif c == ')':
          node = node.parent
      elif c == ' ':
          pass
      else:
          if whereami = "varname":
              node.varnames[varnum] += c
          elif whereami = "refname":
              node.name += c
#+end_src
    
*** discussion / investigation
I believe that this issue should be addressed as a bug rather than as
a point for new development.   The code in [[file:lisp/org-babel-ref.el][org-babel-ref.el]] already
resolves variable references in a recursive manner which *should* work
in the same manner regardless of the depth of the number of nested
function calls.  This recursive evaluation has the effect of
implicitly constructing the parse tree that your are thinking of
constructing explicitly.

Through using some of the commented out debugging statements in
[[file:lisp/org-babel-ref.el][org-babel-ref.el]] I have looked at what may be going wrong in the
current evaluation setup, and it seems that nested variables are being
set using the =:var= header argument, and these variables are being
overridden by the *default* variables which are being entered through
the new functional syntax (see the demonstration header below).

I believe that once this bug is fixed we should be back to fully
resolution of nested arguments.  We should capture this functionality
in a test to ensure that we continue to test it as we move forward.  I
can take a look at implementing this once I get a chance.

Looks like the problem may be in [[file:lisp/org-babel.el::defun%20org%20babel%20merge%20params%20rest%20plists][org-babel-merge-params]], which seems
to be trampling the provided :vars values.

Nope, now it seems that we are actually looking up the results line,
rather than the actual source-code block, which would make sense given
that the results-line will return the same value regardless of the
arguments supplied.  See the output of this [[file:lisp/org-babel-ref.el::message%20type%20S%20type%20debugging][debug-statement]].

We need to be sure that we don't read from a =#+resname:= line when we
have a non-nil set of arguments.

**** demonstration
After uncommenting the debugging statements located [[file:lisp/org-babel-ref.el::message%20format%20first%20second%20S%20S%20new%20refere%20new%20referent%20debugging][here]] and more
importantly [[file:lisp/org-babel-ref.el::message%20nested%20args%20S%20args%20debugging][here]], we can see that the current reference code does
evaluate the references correctly, and it uses the =:var= header
argument to set =a=8=, however the default variables specified using
the functional syntax in =adder(a=3, b=2)= is overriding this
specification.

***** doesn't work with functional syntax

#+srcname: adder-func(a=3, b=2)
#+begin_src python 
a + b
#+end_src

#+resname: adder-func
: 5

#+srcname: after-adder-func(arg=adder-func(a=8))
#+begin_src python 
arg
#+end_src

#+resname: after-adder-func
: 5

***** still does work with =:var= syntax

so it looks like regardless of the syntax used we're not overriding
the default argument values.

#+srcname: adder-header
#+begin_src python :var a=3 :var b=2
a + b
#+end_src

#+resname: adder-header
: 5

#+srcname: after-adder-header
#+begin_src python :var arg=adder-header(a=8, b=0)
arg
#+end_src

#+resname: after-adder-header
: 5

*** Set of test cases
**** Both defaults provided in definition
#+srcname: adder1(a=10,b=20)
#+begin_src python
a+b
#+end_src

#+resname: adder1
: 30

****** DONE Rely on defaults
#+lob: adder1()

#+resname: adder1()
: 30

## should be 30
## OK, but
******* DONE empty parens () not recognised as lob call
        E.g. remove spaces between parens above

        updated [[file:lisp/org-babel-lob.el::defvar%20org%20babel%20lob%20one%20liner%20regexp%20lob%20t%20n%20n%20t%20n][org-babel-lob-one-liner-regexp]]

****** DONE One supplied, one default
#+lob: adder1(a=0)

#+resname: adder1(a=0)
## should be 10
: 20

#+lob: adder1(b=0)

#+resname: adder1(b=0)
## should be 10
: 10

****** DONE Both supplied
#+lob: adder1(a=1,b=2)

#+resname: adder1(a=1,b=2)
## should be 3
: 3

**** One arg lacks default in definition
#+srcname: adder2(a=10,b)
#+begin_src python
a+b
#+end_src
****** DEFERRED Rely on defaults (one of which is missing)
#+lob: adder2( )

[no output]

## should be error: b has no default

Maybe we should let the programming language handle this case.  For
example python spits out an error in the =#+lob= line above.  Maybe
rather than catching these errors our-selves we should raise an error
when the source-block returns an error.  I'll propose a [[* PROPOSED raise elisp error when source-blocks return errors][task]] for this
idea, I'm not sure how/if it would work...

****** DEFERRED Default over-ridden
#+lob: adder2(a=1)

See the above [[* DEFERRED Rely on defaults (one of which is missing)][deferred]] and the new proposed [[* PROPOSED raise elisp error when source-blocks return errors][task]], I think it may be
more flexible to allow the source block language to handle the error.

[no output ]
## should be error: b has no default

****** DONE Missing default supplied
#+lob: adder2(b=1)

#+resname: adder2(b=1)
: 11

## should be 11
## OK

****** DONE One over-ridden, one supplied
#+lob: adder2(a=1,b=2)

#+resname: adder2(a=1,b=2)
: 3
## should be 3

*** Example that fails
    
#+srcname: adder(a=0, b=99)
#+begin_src python 
a+b
#+end_src

#+resname: adder

: 99

#+srcname: one()
#+begin_src python :results silent
2
#+end_src


#+srcname: level-one-nesting
#+begin_src python :var arg=adder(a=one(),b=one())
arg
#+end_src

#+resname:
: 4

#+srcname: level-one-nesting
#+begin_src python :var arg=adder(a=adder(a=one(),b=one()),b=adder(a=one(),b=one()))
arg
#+end_src

#+resname:
: 99

*** DONE deeply nested arguments still fails

#+srcname: deeply-nested-args-bug
#+begin_src python :var arg=adder(a=adder(a=one(),b=one()),b=adder(a=one(),b=one()))
arg
#+end_src

#+resname:
: 8

results in this error
: supplied params=nil
: new-refere="adder", new-referent="a=adder(a=one(),b=one()),b=adder(a=one(),b=one())"
: args=((:var . "a=adder(a=one()") (:var . "b=one())") (:var . "b=adder(a=one()") (:var . "b=one())"))
: type=source-block
: supplied params=((:var . "a=adder(a=one()") (:var . "b=one())") (:var . "b=adder(a=one()") (:var . "b=one())"))
: new-refere="adder", new-referent="a=one("
: args=((:var . "a=one("))
: type=source-block
: supplied params=((:var . "a=one("))
: reference 'one(' not found in this buffer

Need to change the regexp in [[file:lisp/org-babel-ref.el::assign%20any%20arguments%20to%20pass%20to%20source%20block][org-babel-ref-resolve-reference]] so that
it only matches when the parenthesis are balanced.  Maybe look at
[[http://www.gnu.org/software/emacs/elisp/html_node/List-Motion.html][this]].

** DONE avoid stripping whitespace from output when :results output
   This may be partly solved by using o-b-chomp rather than o-b-trim
   in the o-b-LANG-evaluate functions.
** DEFERRED weird escaped characters in shell prompt break shell evaluation
   E.g. this doesn't work. Should the shell sessions set a sane prompt
   when they start up? Or is it a question of altering
   comint-prompt-regexp? Or altering org-babel regexps?
   
#+begin_src sh   
   black=30 ; red=31 ; green=32 ; yellow=33 ; blue=34 ; magenta=35 ; cyan=36 ; white=37
   prompt_col=$red
   prompt_char='>'
   export PS1="\[\033[${prompt_col}m\]\w${prompt_char} \[\033[0m\]"
#+end_src

   I just pushed a good amount of changes, could you see if your shell
   problems still exist?

   The problem's still there. Specifically, aIui, at [[file:lisp/langs/org-babel-sh.el::raw%20org%20babel%20comint%20with%20output%20buffer%20org%20babel%20sh%20eoe%20output%20nil%20insert%20full%20body%20comint%20send%20input%20nil%20t][this line]] of
   org-babel-sh.el, raw gets the value

("" "\e[0m Sun Jun 14 19:26:24 EDT 2009\n" "\e[0m org_babel_sh_eoe\n" "\e[0m ")

   and therefore (member org-babel-sh-eoe-output ...) fails

   I think that `comint-prompt-regexp' needs to be altered to match
   the shell prompt.  This shouldn't be too difficult to do by hand,
   using the `regexp-builder' command and should probably be part of
   the user's regular emacs init.  I can't think of a way for us to
   set this automatically, and we are SOL without a regexp to match
   the prompt.
** DONE function calls in #+srcname: refs
   
   My srcname references don't seem to be working for function
   calls. This needs fixing.
   
#+srcname: called()
#+begin_src python 
59
#+end_src

srcname function call doesn't work for calling a source block
#+srcname: caller(var1=called())
#+begin_src python
var1
#+end_src

#+resname: caller
: 59






They do work for a simple reference
#+srcname: caller2(var1=56)
#+begin_src python 
var1
#+end_src

#+resname: caller2
: 59


and they do work for :var header arg
#+srcname: caller3
#+begin_src python :var var1=called() 
var1
#+end_src

#+resname:
: 58
** DONE LoB: with output to buffer, not working in buffers other than library-of-babel.org
*** Initial report
   I haven't fixed this yet. org-babel-ref-resolve-reference moves
   point around, inside a save-excursion. Somehow when it comes to
   inserting the results (after possible further recursive calls to
   org-babel-ref-resolve-reference), point hasn't gone back to the
   lob line.

#+tblname: test-data
| 1 |    1 |
| 2 |   .5 |
| 3 | .333 |

#+lob: R-plot(data=test-data)

#+lob: python-add(a=2, b=9)

#+resname: python-add(a=2, b=9)
: 11

*** Now
    I think this got fixed in the bugfixes before merging results into master.

** DONE cursor movement when evaluating source blocks
   E.g. the pie chart example. Despite the save-window-excursion in
   org-babel-execute:R. (I never learned how to do this properly: org-R
   jumps all over the place...)

   I don't see this now [ded]

** DONE LoB: calls fail if reference has single character name
   commit 21d058869df1ff23f4f8cc26f63045ac9c0190e2
**** This doesn't work
#+lob: R-plot(data=X)

#+tblname: X
| 1 |     1 |
| 2 |    .5 |
| 3 | .3333 |
| 4 |   .25 |
| 5 |    .2 |
| 6 | .1666 |

**** But this is OK
#+tblname: XX
| 1 |     1 |
| 2 |    .5 |
| 3 | .3333 |
| 4 |   .25 |
| 5 |    .2 |
| 6 | .1666 |

#+lob: R-plot(data=XX)

** DONE make :results replace the default?
   I'm tending to think that appending results to pre-existing results
   creates mess, and that the cleaner `replace' option should be the
   default. E.g. when a source block creates an image, we would want
   that to be updated, rather than have a new one be added.
   
   I agree.

** DONE ruby evaluation not working under ubuntu emacs 23
   With emacs 23.0.91.1 on ubuntu, for C-h f run-ruby I have the
   following, which seems to conflict with [[file:lisp/langs/org-babel-ruby.el::let%20session%20buffer%20save%20window%20excursion%20run%20ruby%20nil%20session%20current%20buffer][this line]] in org-babel-ruby.el.

#+begin_example
run-ruby is an interactive compiled Lisp function.

(run-ruby cmd)

Run an inferior Ruby process, input and output via buffer *ruby*.
If there is a process already running in `*ruby*', switch to that buffer.
With argument, allows you to edit the command line (default is value
of `ruby-program-name').  Runs the hooks `inferior-ruby-mode-hook'
(after the `comint-mode-hook' is run).
(Type C-h m in the process buffer for a list of commands.)
#+end_example

   So, I may have a non-standard inf-ruby.el.  Here's my version of
   run-ruby.

#+begin_example 
run-ruby is an interactive Lisp function in `inf-ruby.el'.

(run-ruby &optional COMMAND NAME)

Run an inferior Ruby process, input and output via buffer *ruby*.
If there is a process already running in `*ruby*', switch to that buffer.
With argument, allows you to edit the command line (default is value
of `ruby-program-name').  Runs the hooks `inferior-ruby-mode-hook'
(after the `comint-mode-hook' is run).
(Type C-h m in the process buffer for a list of commands.)
#+end_example

   It seems we could either bundle my version of inf-ruby.el (as it's
   the newest).  Or we could change the use of `run-ruby' so that it
   is robust across multiple distributions.  I think I'd prefer the
   former, unless the older version of inf-ruby is actually bundled
   with emacs, in which case maybe we should go out of our way to
   support it.  Thoughts?

   I think for now I'll just include the latest [[file:util/inf-ruby.el][inf-ruby.el]] in the
   newly created utility directory.  I doubt anyone would have a
   problem using the latest version of this file.
** DONE test failing forcing vector results with =test-forced-vector-results= ruby code block
Note that this only seems to happen the *second* time the test table
is evaluated

#+srcname: bug-trivial-vector
#+begin_src emacs-lisp :results vector silent
8
#+end_src

#+srcname: bug-forced-vector-results
#+begin_src ruby :var triv=test-trivial-vector :results silent
triv.class.name
#+end_src

mysteriously this seems to be fixed...
** DONE defunct R sessions
Sometimes an old R session will turn defunct, and newly inserted code
will not be evaluated (leading to a hang).

This seems to be fixed by using `inferior-ess-send-input' rather than `comint-send-input'.
** DONE ruby fails on first call to non-default session

#+srcname: bug-new-session
#+begin_src ruby :session is-new
:patton
#+end_src

** DONE when reading results from =#+resname= line

Errors when trying to read from resname lines.

#+resname: bug-in-resname
: 8

#+srcname: bug-in-resname-reader
#+begin_src emacs-lisp :var buggy=bug-in-resname() :results silent
buggy
#+end_src

** DONE R-code broke on "org-babel" rename

#+srcname: bug-R-babels
#+begin_src R 
8 * 2
#+end_src

** DONE error on trivial R results

So I know it's generally not a good idea to squash error without
handling them, but in this case the error almost always means that
there was no file contents to be read by =org-table-import=, so I
think it's ok.

#+srcname: bug-trivial-r1
#+begin_src R :results replace
pie(c(1, 2, 3), labels = c(1, 2, 3))
#+end_src

#+srcname: bug-trivial-r2
#+begin_src R :results replace
8
#+end_src

#+resname: bug-trivial-r2
: 8

#+srcname: bug-trivial-r3
#+begin_src R :results replace
c(1,2,3)
#+end_src

#+resname: bug-trivial-r3
| 1 |
| 2 |
| 3 |

** DONE ruby new variable creation (multi-line ruby blocks)
Actually it looks like we were dropping all but the last line.

#+srcname: multi-line-ruby-test
#+begin_src ruby :var table=bug-numerical-table :results replace
total = 0
table.each{|n| total += n}
total/table.size
#+end_src

#+resname:
: 2

** DONE R code execution seems to choke on certain inputs
Currently the R code seems to work on vertical (but not landscape)
tables

#+srcname: little-fake
#+begin_src emacs-lisp 
"schulte"
#+end_src


#+begin_src R :var num=little-fake
num
#+end_src

#+resname:
: schulte

#+srcname: set-debug-on-error
#+begin_src emacs-lisp :results silent
(setq debug-on-error t)
#+end_src

#+srcname: bug-numerical-table
#+begin_src emacs-lisp :results silent
'(1 2 3)
#+end_src




#+srcname: bug-R-number-evaluation
#+begin_src R :var table=bug-numerical-table
mean(mean(table))
#+end_src

#+resname:
: 2



#+tblname: bug-vert-table
| 1 |
| 2 |
| 3 |

#+srcname: bug-R-vertical-table
#+begin_src R :var table=bug-vert-table :results silent
mean(table)
#+end_src

** DONE org bug/request: prevent certain org behaviour within code blocks
   E.g. [[]] gets recognised as a link (when there's text inside the
   brackets). This is bad for R code at least, and more generally
   could be argued to be inappropriate. Is it difficult to get org to
   ignore text in code blocks? [DED]
   
   I believe Carsten addressed this recently on the mailing list with
   the comment that it was indeed a difficult issue.  I believe this
   may be one area where we could wait for an upstream (org-mode) fix.

   [Dan] Carsten has fixed this now in the core.

** DONE with :results replace, non-table output doesn't replace table output
   And vice versa. E.g. Try this first with table and then with len(table) [DED]
#+begin_src python :var table=sandbox :results replace
table
#+end_src

| 1 |         2 | 3 |
| 4 | "schulte" | 6 |
: 2

Yes, this is certainly a problem.  I fear that if we begin replacing
anything immediately following a source block (regardless of whether
it matches the type of our current results) we may accidentally delete
hand written portions of the user's org-mode buffer.

I think that the best solution here would be to actually start
labeling results with a line that looks something like...

#+results: name

This would have a couple of benefits...
1) we wouldn't have to worry about possibly deleting non-results
   (which is currently an issue)
2) we could reliably replace results even if there are different types
3) we could reference the results of a source-code block in variable
   definitions, which would be useful if for example we don't wish to
   re-run a source-block every time because it is long-running.

Thoughts?  If no-one objects, I believe I will implement the labeling
of results.

** DONE extra quotes for nested string
Well R appears to be reading the tables without issue...

these *should* be quoted
#+srcname: ls
#+begin_src sh :results replace
ls
#+end_src

| "COPYING"          |
| "README.markdown"  |
| "block"            |
| "examples.org"     |
| "existing_tools"   |
| "intro.org"        |
| "org-babel"          |
| "rorg.org"         |
| "test-export.html" |
| "test-export.org"  |

#+srcname: test-quotes
#+begin_src ruby :var tab=ls
tab[1][0]
#+end_src

: README.markdown

#+srcname: test-quotes
#+begin_src R :var tab=ls
as.matrix(tab[2,])
#+end_src

: README.markdown

** DONE simple ruby arrays not working

As an example eval the following.  Adding a line to test

#+tblname: simple-ruby-array
| 3 | 4 | 5 |

#+srcname: ruby-array-test
#+begin_src ruby :var ar = simple-ruby-array :results silent
ar.first.first
#+end_src

** DONE space trailing language name
fix regexp so it works when there's a space trailing the language name

#+srcname: test-trailing-space
#+begin_src ruby 
:schulte
#+end_src

** DONE Args out of range error
   
The following block resulted in the error below [DED]. It ran without
error directly in the shell.
#+begin_src sh
cd ~/work/genopca
for platf in ill aff ; do
    for pop in CEU YRI ASI ; do
        rm -f $platf/hapmap-genos-$pop-all $platf/hapmap-rs-all
        cat $platf/hapmap-genos-$pop-* > $platf/hapmap-genos-$pop-all
        cat $platf/hapmap-rs-* > $platf/hapmap-rs-all
    done
done
#+end_src
  
 executing source block with sh...
finished executing source block
string-equal: Args out of range: "", -1, 0

the error =string-equal: Args out of range: "", -1, 0= looks like what
used to be output when the block returned an empty results string.
This should be fixed in the current version, you should now see the
following message =no result returned by source block=.

** DONE ruby arrays not recognized as such

Something is wrong in [[file:lisp/org-babel-script.el]] related to the
recognition of ruby arrays as such.

#+begin_src ruby :results replace
[1, 2, 3, 4]
#+end_src

| 1 | 2 | 3 | 4 |

#+begin_src python :results replace
[1, 2, 3, 4]
#+end_src

| 1 | 2 | 3 | 4 |
** REJECTED elisp reference fails for literal number
   That's a bug in Dan's elisp, not in org-babel.
#+srcname: elisp-test(a=4)
#+begin_src emacs-lisp 
(message a)
#+end_src





: 30


* Tests
Evaluate all the cells in this table for a comprehensive test of the
org-babel functionality.

*Note*: if you have customized =org-babel-default-header-args= then some
of these tests may fail.

#+TBLNAME: org-babel-tests
| functionality           | block                      | arg |    expected |     results | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| basic evaluation        |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| emacs lisp              | basic-elisp                |     |           5 |           5 | pass |
| shell                   | basic-shell                |     |           6 |           6 | pass |
| ruby                    | basic-ruby                 |     |   org-babel |   org-babel | pass |
| python                  | basic-python               |     | hello world | hello world | pass |
| R                       | basic-R                    |     |          13 |          13 | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| tables                  |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| emacs lisp              | table-elisp                |     |           3 |           3 | pass |
| ruby                    | table-ruby                 |     |       1-2-3 |       1-2-3 | pass |
| python                  | table-python               |     |           5 |           5 | pass |
| R                       | table-R                    |     |         3.5 |         3.5 | pass |
| R: col names in R       | table-R-colnames           |     |          -3 |          -3 | pass |
| R: col names in org     | table-R-colnames-org       |     |         169 |         169 | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| source block references |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| all languages           | chained-ref-last           |     |       Array |       Array | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| source block functions  |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| emacs lisp              | defun-fibb                 |     |       fibbd |       fibbd | pass |
| run over                | Fibonacci                  |   0 |           1 |           1 | pass |
| a                       | Fibonacci                  |   1 |           1 |           1 | pass |
| variety                 | Fibonacci                  |   2 |           2 |           2 | pass |
| of                      | Fibonacci                  |   3 |           3 |           3 | pass |
| different               | Fibonacci                  |   4 |           5 |           5 | pass |
| arguments               | Fibonacci                  |   5 |           8 |           8 | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| bugs and tasks          |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| simple ruby arrays      | ruby-array-test            |     |           3 |           3 | pass |
| R number evaluation     | bug-R-number-evaluation    |     |           2 |           2 | pass |
| multi-line ruby blocks  | multi-line-ruby-test       |     |           2 |           2 | pass |
| forcing vector results  | test-forced-vector-results |     |       Array |       Array | pass |
| deeply nested arguments | deeply-nested-args-bug     |     |           8 |           8 | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| sessions                |                            |     |             |             | pass |
|-------------------------+----------------------------+-----+-------------+-------------+------|
| set ruby session        | set-ruby-session-var       |     |        :set |        :set | pass |
| get from ruby session   | get-ruby-session-var       |     |           3 |           3 | pass |
| set python session      | set-python-session-var     |     |         set |         set | pass |
| get from python session | get-python-session-var     |     |           4 |           4 | pass |
| set R session           | set-R-session-var          |     |         set |         set | pass |
| get from R session      | get-R-session-var          |     |           5 |           5 | pass |
#+TBLFM: $5='(if (= (length $3) 1) (progn (message (format "running %S" '(sbe $2 (n $3)))) (sbe $2 (n $3))) (sbe $2))::$6='(if (string= $4 $5) "pass" (format "expected %S but was %S" $4 $5))
#+TBLFM: $5=""::$6=""


The second TBLFM line (followed by replacing '[]' with '') can be used
to blank out the table results, in the absence of a better method.

** basic tests

#+srcname: basic-elisp
#+begin_src emacs-lisp :results silent
(+ 1 4)
#+end_src


#+srcname: basic-shell
#+begin_src sh :results silent
expr 1 + 5
#+end_src


#+srcname: date-simple
#+begin_src sh :results silent
date
#+end_src

#+srcname: basic-ruby
#+begin_src ruby :results silent
"org-babel"
#+end_src


#+srcname: basic-python
#+begin_src python :results silent
'hello world'
#+end_src


#+srcname: basic-R
#+begin_src R :results silent
b <- 9
b + 4
#+end_src


** read tables

#+tblname: test-table
| 1 | 2 | 3 |
| 4 | 5 | 6 |

#+tblname: test-table-colnames
| var1 | var2 | var3 |
|------+------+------|
|    1 |   22 |   13 |
|   41 |   55 |   67 |

#+srcname: table-elisp
#+begin_src emacs-lisp :results silent :var table=test-table
(length (car table))
#+end_src


#+srcname: table-ruby
#+begin_src ruby :results silent :var table=test-table
table.first.join("-")
#+end_src


#+srcname: table-python
#+begin_src python :var table=test-table
table[1][1]
#+end_src

#+srcname: table-R(table=test-table)
#+begin_src R
mean(mean(table))
#+end_src

#+srcname: table-R-colnames(table=test-table-colnames)
#+begin_src R :results silent
sum(table$var2 - table$var3)
#+end_src

#+srcname: R-square(x=default-name-doesnt-exist)
#+begin_src R :colnames t
x^2
#+end_src

This should return 169. The fact that R is able to use the column name
to index the data frame (x$var3) proves that a table with column names
(a header row) has been recognised as input for the R-square function
block, and that the R-square block has output an elisp table with
column names, and that the colnames have again been recognised when
creating the R variables in this block.
#+srcname: table-R-colnames-org(x = R-square(x=test-table-colnames))
#+begin_src R
x$var3[1]
#+end_src




** references

Lets pass a references through all of our languages...

Lets start by reversing the table from the previous examples

#+srcname: chained-ref-first
#+begin_src python :var table = test-table
table.reverse()
table
#+end_src

#+resname: chained-ref-first
| 4 | 5 | 6 |
| 1 | 2 | 3 |

Take the first part of the list

#+srcname: chained-ref-second
#+begin_src R :var table = chained-ref-first
table[1]
#+end_src

#+resname: chained-ref-second
| 4 |
| 1 |

Turn the numbers into string

#+srcname: chained-ref-third
#+begin_src emacs-lisp :var table = chained-ref-second
(mapcar (lambda (el) (format "%S" el)) table)
#+end_src

#+resname: chained-ref-third
| "(4)" | "(1)" |

and Check that it is still a list

#+srcname: chained-ref-last
#+begin_src ruby :var table=chained-ref-third
table.class.name
#+end_src


** source blocks as functions

#+srcname: defun-fibb
#+begin_src emacs-lisp :results silent
(defun fibbd (n) (if (< n 2) 1 (+ (fibbd (- n 1)) (fibbd (- n 2)))))
#+end_src


#+srcname: fibonacci
#+begin_src emacs-lisp :results silent :var n=7
(fibbd n)
#+end_src







** sbe tests (these don't seem to be working...)
Testing the insertion of results into org-mode tables.

#+srcname: multi-line-output
#+begin_src ruby :results replace
"the first line ends here


     and this is the second one

even a third"
#+end_src

#+resname:
: the first line ends here\n\n\n     and this is the second one\n\neven a third

#+srcname: multi-line-error
#+begin_src ruby :results replace
raise "oh nooooooooooo"
#+end_src

#+resname:
: oh nooooooooooo

| the first line ends here... | -:5: warning: parenthesize argument(s) for future version... |
#+TBLFM: $1='(sbe "multi-line-output")::$2='(sbe "multi-line-error")

** forcing results types tests

#+srcname: test-trivial-vector
#+begin_src emacs-lisp :results vector silent
8
#+end_src

#+srcname: test-forced-vector-results
#+begin_src ruby :var triv=test-trivial-vector :results silent
triv.class.name
#+end_src

** sessions

#+srcname: set-ruby-session-var
#+begin_src ruby :session :results silent
var = [1, 2, 3]
:set
#+end_src

#+srcname: get-ruby-session-var
#+begin_src ruby :session :results silent
var.size
#+end_src

#+srcname: set-python-session-var
#+begin_src python :session
var=4
'set'
#+end_src

#+srcname: get-python-session-var
#+begin_src python :session
var
#+end_src

#+srcname: set-R-session-var
#+begin_src R :session
a <- 5
'set'
#+end_src

#+srcname: get-R-session-var
#+begin_src R :session
a
#+end_src


* Sandbox
  :PROPERTIES:
  :CUSTOM_ID: sandbox
  :END:
To run these examples evaluate [[file:lisp/org-babel-init.el][org-babel-init.el]]

** org-babel.el beginning functionality

#+begin_src sh  :results replace
date
#+end_src

#+resname:
: Sun Jul  5 18:54:39 EDT 2009

#+begin_src ruby
Time.now
#+end_src

#+resname:
: Sun Jul 05 18:54:35 -0400 2009

#+begin_src python
"Hello World"
#+end_src

#+resname:
: Hello World


** org-babel-R

#+begin_src R :results replace
a <- 9
b <- 16
a + b
#+end_src

#+resname:
: 25

#+begin_src R
hist(rgamma(20,3,3))
#+end_src



** org-babel plays with tables
Alright, this should demonstrate both the ability of org-babel to read
tables into a lisp source code block, and to then convert the results
of the source code block into an org table.  It's using the classic
"lisp is elegant" demonstration transpose function.  To try this
out...

1. evaluate [[file:lisp/org-babel-init.el]] to load org-babel and friends
2. evaluate the transpose definition =\C-c\\C-c= on the beginning of
   the source block
3. evaluate the next source code block, this should read in the table
   because of the =:var table=previous=, then transpose the table, and
   finally it should insert the transposed table into the buffer
   immediately following the block

*** Emacs lisp

#+begin_src emacs-lisp :results silent
(defun transpose (table)
  (apply #'mapcar* #'list table))
#+end_src


#+TBLNAME: sandbox
| 1 |       2 | 3 |
| 4 | schulte | 6 |

#+begin_src emacs-lisp :var table=sandbox :results replace
(transpose table)
#+end_src


#+begin_src emacs-lisp
'(1 2 3 4 5)
#+end_src

#+resname:
| 1 | 2 | 3 | 4 | 5 |

*** Ruby and Python

#+begin_src ruby :var table=sandbox :results replace
table.first.join(" - ")
#+end_src

#+resname:
: 1 - 2 - 3

#+begin_src python :var table=sandbox
table[0]
#+end_src


#+begin_src ruby :var table=sandbox :results replace
table
#+end_src

#+resname:
: [[1, 2, 3], [4, "schulte", 6]]


| 1 |         2 | 3 |
| 4 | "schulte" | 6 |

#+begin_src python :var table=sandbox :results replace
len(table)
#+end_src

: 2

| "__add__" | "__class__" | "__contains__" | "__delattr__" | "__delitem__" | "__delslice__" | "__doc__" | "__eq__" | "__format__" | "__ge__" | "__getattribute__" | "__getitem__" | "__getslice__" | "__gt__" | "__hash__" | "__iadd__" | "__imul__" | "__init__" | "__iter__" | "__le__" | "__len__" | "__lt__" | "__mul__" | "__ne__" | "__new__" | "__reduce__" | "__reduce_ex__" | "__repr__" | "__reversed__" | "__rmul__" | "__setattr__" | "__setitem__" | "__setslice__" | "__sizeof__" | "__str__" | "__subclasshook__" | "append" | "count" | "extend" | "index" | "insert" | "pop" | "remove" | "reverse" | "sort" |

*** (sandbox table) R

#+TBLNAME: sandbox_r
| 1 |       2 | 3 |
| 4 | schulte | 6 |

#+begin_src R :results replace
x <- c(rnorm(10, mean=-3, sd=1), rnorm(10, mean=3, sd=1))
x
#+end_src

| -3.35473133869346 |
|    -2.45714878661 |
| -3.32819924928633 |
| -2.97310212756194 |
| -2.09640758369576 |
| -5.06054014378736 |
| -2.20713700711221 |
| -1.37618039712037 |
| -1.95839385821742 |
| -3.90407396475502 |
|  2.51168071590226 |
|  3.96753011570494 |
|  3.31793212627865 |
|  1.99829753972341 |
|  4.00403686419829 |
|  4.63723764452927 |
|  3.94636744261313 |
|  3.58355906547775 |
|  3.01563442274226 |
|   1.7634976849927 |

#+begin_src R var tabel=sandbox_r :results replace
tabel
#+end_src

| 1 |         2 | 3 |
| 4 | "schulte" | 6 |

*** shell
Now shell commands are converted to tables using =org-table-import=
and if these tables are non-trivial (i.e. have multiple elements) then
they are imported as org-mode tables...

#+begin_src sh :results replace
ls -l
#+end_src

| "total"      | 208 | ""    | ""    |    "" |   "" | "" | ""                |
| "-rw-r--r--" |   1 | "dan" | "dan" |    57 | 2009 | 15 | "block"           |
| "-rw-r--r--" |   1 | "dan" | "dan" | 35147 | 2009 | 15 | "COPYING"         |
| "-rw-r--r--" |   1 | "dan" | "dan" |   722 | 2009 | 18 | "examples.org"    |
| "drwxr-xr-x" |   4 | "dan" | "dan" |  4096 | 2009 | 19 | "existing_tools"  |
| "-rw-r--r--" |   1 | "dan" | "dan" |  2207 | 2009 | 14 | "intro.org"       |
| "drwxr-xr-x" |   2 | "dan" | "dan" |  4096 | 2009 | 18 | "org-babel"         |
| "-rw-r--r--" |   1 | "dan" | "dan" |   277 | 2009 | 20 | "README.markdown" |
| "-rw-r--r--" |   1 | "dan" | "dan" | 11837 | 2009 | 18 | "rorg.html"       |
| "-rw-r--r--" |   1 | "dan" | "dan" | 61829 | 2009 | 19 | "#rorg.org#"      |
| "-rw-r--r--" |   1 | "dan" | "dan" | 60190 | 2009 | 19 | "rorg.org"        |
| "-rw-r--r--" |   1 | "dan" | "dan" |   972 | 2009 | 11 | "test-export.org" |


** silent evaluation

#+begin_src ruby
:im_the_results
#+end_src

: :im_the_results

#+begin_src ruby :results silent
:im_the_results
#+end_src

#+begin_src ruby :results replace
:im_the_results_
#+end_src

: :im_the_results_


** (sandbox) referencing other source blocks
Doing this in emacs-lisp first because it's trivial to convert
emacs-lisp results to and from emacs-lisp.

*** emacs lisp source reference
This first example performs a calculation in the first source block
named =top=, the results of this calculation are then saved into the
variable =first= by the header argument =:var first=top=, and it is
used in the calculations of the second source block.

#+SRCNAME: top
#+begin_src emacs-lisp
(+ 4 2)
#+end_src

#+begin_src emacs-lisp :var first=top :results replace
(* first 3)
#+end_src

: 18

This example is the same as the previous only the variable being
passed through is a table rather than a number.

#+begin_src emacs-lisp :results silent
(defun transpose (table)
  (apply #'mapcar* #'list table))
#+end_src

#+TBLNAME: top_table
| 1 |       2 | 3 |
| 4 | schulte | 6 |

#+SRCNAME: second_src_example
#+begin_src emacs-lisp :var table=top_table
(transpose table)
#+end_src

#+begin_src emacs-lisp :var table=second_src_example :results replace
(transpose table)
#+end_src

| 1 |         2 | 3 |
| 4 | "schulte" | 6 |
*** ruby python
Now working for ruby

#+srcname: start
#+begin_src ruby
89
#+end_src

#+begin_src ruby :var other=start :results replace
2 * other
#+end_src

and for python

#+SRCNAME: start_two
#+begin_src python
98
#+end_src

#+begin_src python :var another=start_two :results replace
another*3
#+end_src

*** mixed languages
Since all variables are converted into Emacs Lisp it is no problem to
reference variables specified in another language.

#+SRCNAME: ruby-block
#+begin_src ruby
2
#+end_src

#+SRCNAME: lisp_block
#+begin_src emacs-lisp :var ruby-variable=ruby-block
(* ruby-variable 8)
#+end_src

#+begin_src python :var lisp_var=lisp_block
lisp_var + 4
#+end_src

: 20

*** R

#+srcname: first_r
#+begin_src R :results replace
a <- 9
a
#+end_src

: 9

#+begin_src R :var other=first_r :results replace
other + 2
#+end_src

: 11


** (sandbox) selective export

For exportation tests and examples see (including exportation of
inline source code blocks) [[file:test-export.org]]


** (sandbox) source blocks as functions

#+srcname: default
#+begin_src emacs-lisp :results silent
5
#+end_src

#+srcname: triple
#+begin_src emacs-lisp :var n=default :results replace
(* 3 n)
#+end_src

: 15

#+begin_src emacs-lisp :var result=triple(n=3, m=98) :results replace
result
#+end_src

: 294

The following just demonstrates the ability to assign variables to
literal values, which was not implemented until recently.

#+begin_src ruby :var num="eric" :results replace
num+" schulte "
#+end_src

: "eric schulte "


** (sandbox) inline source blocks

This is an inline source code block src_ruby{1 + 6}.  And another
source block with text output src_emacs-lisp{"eric"}.

This is an inline source code block with header
arguments.  src_ruby[:var n=fibbd( n = 0 )]{n}


** (sandbox) integration w/org tables

#+begin_src emacs-lisp :results silent
(defun fibbd (n) (if (< n 2) 1 (+ (fibbd (- n 1)) (fibbd (- n 2)))))
#+end_src

#+srcname: fibbd
#+begin_src emacs-lisp :var n=4 :results silent
(fibbd n)
#+end_src

#+begin_src emacs-lisp :results silent
(mapcar #'fibbd '(0 1 2 3 4 5 6 7 8))
#+end_src

Something is not working here.  The function `sbe ' works fine when
called from outside of the table (see the source block below), but
produces an error when called from inside the table.  I think there
must be some narrowing going on during intra-table emacs-lisp
evaluation.

| original | fibbd |
|----------+-------|
|        0 |     1 |
|        1 |     1 |
|        2 |     2 |
|        3 |     3 |
|        4 |     5 |
|        5 |     8 |
|        6 |    13 |
|        7 |    21 |
|        8 |    34 |
|        9 |    55 |
#+TBLFM: $2='(sbe "fibbd" (n $1))

silent-result

#+begin_src emacs-lisp :results silent
(sbe 'fibbd (n "8"))
#+end_src


* Buffer Dictionary
 LocalWords:  DBlocks dblocks org-babel el eric fontification