import R data frame into org-mode table

import R data frame into org-mode table

[Rob Stewart]

I have an R script that generates a data frame, that I export to a CSV
file. The data looks something like this:

    "","Variant","Xaxis","N","mean","sd","se"
    "1","line1",10,5,111.11,9.33,3.11
    "1","line1",20,5,112.11,9.13,3.14
    "1","line1",30,5,113.11,9.43,3.10
    "1","line2",10,5,101.11,8.33,2.11
    "1","line2",20,5,100.11,8.13,2.12
    "1","line2",30,5,108.11,8.03,2.10

I have an ongoing org-mode document in which I'd like to report this
data. I'd like to add a table that looks something like this:

    | Variant | X Axis | Y Axis | N | Mean | Standard Deviation |
Standard Error |
    |----------------------------------------------------------------------------|

The data will change frequently over the coming months. I'd like
org-mode to auto-magically read a CSV file, to construct my org-table
on-the-fly. I'd hoping to avoid copying and pasting results each time.

Is there some org-mode or emacs+ESS magic I can use to populate my
empty table, with source data from my CSV file?

--
Rob

Re: import R data frame into org-mode table

[Andreas Leha]

Rob Stewart <robstewart57@gmail.com> writes:

> I have an R script that generates a data frame, that I export to a CSV
> file. The data looks something like this:
>
>     "","Variant","Xaxis","N","mean","sd","se"
>     "1","line1",10,5,111.11,9.33,3.11
>     "1","line1",20,5,112.11,9.13,3.14
>     "1","line1",30,5,113.11,9.43,3.10
>     "1","line2",10,5,101.11,8.33,2.11
>     "1","line2",20,5,100.11,8.13,2.12
>     "1","line2",30,5,108.11,8.03,2.10
>
> I have an ongoing org-mode document in which I'd like to report this
> data. I'd like to add a table that looks something like this:
>
>     | Variant | X Axis | Y Axis | N | Mean | Standard Deviation |
> Standard Error |
>     |----------------------------------------------------------------------------|
>
> The data will change frequently over the coming months. I'd like
> org-mode to auto-magically read a CSV file, to construct my org-table
> on-the-fly. I'd hoping to avoid copying and pasting results each time.
>
> Is there some org-mode or emacs+ESS magic I can use to populate my
> empty table, with source data from my CSV file?
>
> --
> Rob

Definitely there is:

--8<---------------cut here---------------start------------->8---
#+begin_src R :results table :colnames yes
  read.csv('test.csv')
#+end_src

#+results:
| X | Variant | Xaxis | N |   mean |   sd |   se |
|---+---------+-------+---+--------+------+------|
| 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
| 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
| 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
| 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
| 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
| 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |

--8<---------------cut here---------------end--------------->8---

Regards,
Andreas

Re: import R data frame into org-mode table

[Thomas S. Dye]

Rob Stewart <robstewart57@gmail.com> writes:

> I have an R script that generates a data frame, that I export to a CSV
> file. The data looks something like this:
>
>     "","Variant","Xaxis","N","mean","sd","se"
>     "1","line1",10,5,111.11,9.33,3.11
>     "1","line1",20,5,112.11,9.13,3.14
>     "1","line1",30,5,113.11,9.43,3.10
>     "1","line2",10,5,101.11,8.33,2.11
>     "1","line2",20,5,100.11,8.13,2.12
>     "1","line2",30,5,108.11,8.03,2.10
>
> I have an ongoing org-mode document in which I'd like to report this
> data. I'd like to add a table that looks something like this:
>
>     | Variant | X Axis | Y Axis | N | Mean | Standard Deviation |
> Standard Error |
>     |----------------------------------------------------------------------------|
>
> The data will change frequently over the coming months. I'd like
> org-mode to auto-magically read a CSV file, to construct my org-table
> on-the-fly. I'd hoping to avoid copying and pasting results each time.
>
> Is there some org-mode or emacs+ESS magic I can use to populate my
> empty table, with source data from my CSV file?
>
> --
> Rob

Hi Rob,

Andreas' post answers your specific question.

In your situation, I typically don't run through a csv file, but just
write the R script so the data frame is returned to the Org mode buffer
as the result of the source code block.

As for the translation of column heads, see the ascii package for R,
which has a print() routine that recognizes Org mode as a type of
output, and which has facilities for formatting column names.

Here is an example from some work in progress that uses a session:

#+property: exports results
#+property: colnames yes
#+property: session *adzes*

#+name: lda-mid-z
#+header: :results output org
#+begin_src R 
  y <- edxrf[edxrf$lab=='UH-Hilo',]
  fit <- lda(as.factor(site) ~  rb_ppm + sr_ppm + y_ppm + zr_ppm + nb_ppm,
             data=y, na.action="na.omit", CV=TRUE)
  adzes <- y$site=='unknown'
  cap <- "Assignment of unknowns to sources"
  z <- ascii(cbind(BPBM=y$bpbm[adzes],round(fit$posterior[adzes,],2)),header=T,include.colnames=T,caption=cap)
  print(z,type="org")
  rm(y,fit,adzes,cap,z)
#+end_src

#+RESULTS: lda-mid-z
#+BEGIN_ORG
#+CAPTION: Assignment of unknowns to sources
| BPBM                      | Haleakala | Kilauea | Mauna Kea | Nu`u | Pololu | Waiahole | unknown |
|---------------------------+-----------+---------+-----------+------+--------+----------+---------|
| 1985.134.01 adze          | 0         | 0.1     | 0         | 0    | 0      | 0.64     | 0.27    |
| 50-HA-1349-G-5 adze       | 0         | 0.85    | 0         | 0    | 0      | 0        | 0.14    |
| 50-HA-900-P24-1 microadze | 0         | 0       | 1         | 0    | 0      | 0        | 0       |

...

hth,
Tom


-- 
Thomas S. Dye
http://www.tsdye.com

Re: import R data frame into org-mode table

[Rob Stewart]

Hi Andreas,

On 17 July 2013 23:09, Andreas Leha <andreas.leha@med.uni-goettingen.de> wrote:

> Definitely there is:
>
> --8<---------------cut here---------------start------------->8---
> #+begin_src R :results table :colnames yes
>   read.csv('test.csv')
> #+end_src
>
> #+results:
> | X | Variant | Xaxis | N |   mean |   sd |   se |
> |---+---------+-------+---+--------+------+------|
> | 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
> | 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
> | 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
> | 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
> | 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
> | 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |
>
> --8<---------------cut here---------------end--------------->8---

Great. I've gone ahead and done this. There are two additional requirements:
1) My table needs a caption. Will #+CAPTION: just above the
#+BEGIN_SRC just work? Even better, a label, allowing me to also cross
reference it.
2) My table is too long for 1 page. It spans multiple pages
vertically. According to this StackOverflow answer
http://stackoverflow.com/a/2896850/1526266 , I should instead use
longtable, not tabular. Is there a way to coerce your above snippet to
use longtable, instead of tabular which is the default.

Thanks!

--
Rob

Re: import R data frame into org-mode table

[Cook, Malcolm]

>Hi Andreas,
 >
 >On 17 July 2013 23:09, Andreas Leha <andreas.leha@med.uni-goettingen.de> wrote:
 >
 >> Definitely there is:
 >>
 >> --8<---------------cut here---------------start------------->8---
 >> #+begin_src R :results table :colnames yes
 >>   read.csv('test.csv')
 >> #+end_src
 >>
 >> #+results:
 >> | X | Variant | Xaxis | N |   mean |   sd |   se |
 >> |---+---------+-------+---+--------+------+------|
 >> | 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
 >> | 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
 >> | 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
 >> | 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
 >> | 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
 >> | 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |
 >>
 >> --8<---------------cut here---------------end--------------->8---
 >
 >Great. I've gone ahead and done this. There are two additional requirements:
 >1) My table needs a caption. Will #+CAPTION: just above the
 >#+BEGIN_SRC just work? 

Yes

> Even better, a label, allowing me to also cross
 >reference it.
 >2) My table is too long for 1 page. It spans multiple pages
 >vertically. According to this StackOverflow answer
 >http://stackoverflow.com/a/2896850/1526266 , I should instead use
 >longtable, not tabular. Is there a way to coerce your above snippet to
 >use longtable, instead of tabular which is the default.
 >

Looks like your answer is here: http://orgmode.org/manual/Tables-in-LaTeX-export.html

 >Thanks!
 >
 >--
 >Rob

Re: import R data frame into org-mode table

[John Hendy]

On Mon, Jul 29, 2013 at 10:35 AM, Cook, Malcolm <MEC@stowers.org> wrote:
>>Hi Andreas,
>  >
>  >On 17 July 2013 23:09, Andreas Leha <andreas.leha@med.uni-goettingen.de> wrote:
>  >
>  >> Definitely there is:
>  >>
>  >> --8<---------------cut here---------------start------------->8---
>  >> #+begin_src R :results table :colnames yes
>  >>   read.csv('test.csv')
>  >> #+end_src
>  >>
>  >> #+results:
>  >> | X | Variant | Xaxis | N |   mean |   sd |   se |
>  >> |---+---------+-------+---+--------+------+------|
>  >> | 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
>  >> | 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
>  >> | 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
>  >> | 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
>  >> | 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
>  >> | 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |
>  >>
>  >> --8<---------------cut here---------------end--------------->8---
>  >
>  >Great. I've gone ahead and done this. There are two additional requirements:
>  >1) My table needs a caption. Will #+CAPTION: just above the
>  >#+BEGIN_SRC just work?
>
> Yes
>
>> Even better, a label, allowing me to also cross
>  >reference it.
>  >2) My table is too long for 1 page. It spans multiple pages
>  >vertically. According to this StackOverflow answer
>  >http://stackoverflow.com/a/2896850/1526266 , I should instead use
>  >longtable, not tabular. Is there a way to coerce your above snippet to
>  >use longtable, instead of tabular which is the default.
>  >
>
> Looks like your answer is here: http://orgmode.org/manual/Tables-in-LaTeX-export.html

Sort of. Depends on the Org-mode version, and someone will have to
chime in on the updated status of various parts of the manual. For 8+,
I believe the syntax is different:

#+attr_latex: :environment longtable

Also, there have been many threads in the past about how to add
#+attr_latex lines to results output (mostly graphics/files)
successfully. If you simply take the above and add =#+attr_latex:
stuff= above the =#+results= line, babel won't recognize it and will
just create a new results block. If those on this email are already
well aware of this... my apologies for being redundant, but it causes
enough confusion that I figured I'd leave another bread crumb trail :)
Here's an example of a time that came up on the list:
- http://lists.gnu.org/archive/html/emacs-orgmode/2012-07/msg00237.html

You need to use any attributes (that includes #+begin/end_center, and
any #+attr_backend lines) *in combination with* a named source block.
So the full solution should look something like this:

#+name: export-table
#+begin_src R :results table :colnames yes
read.csv('test.csv')
#+end_src

#+RESULTS: export-table
#+attr_latex: :environment longtable
| | |

I see above that =:results output org= was used. There was some
discussion about this for use with the new exporter as I wasn't
getting great results:
- https://lists.gnu.org/archive/html/emacs-orgmode/2013-03/msg01582.html
- http://www.mail-archive.com/emacs-orgmode@gnu.org/msg69748.html

If it's working for you, don't worry about it. If not, you may find
some help in those threads. I've taken to using the ascii package for
output org-mode tables from R.
- http://orgmode.org/worg/org-contrib/babel/examples/ascii.html

Works great for me.


Good luck!
John


>
>  >Thanks!
>  >
>  >--
>  >Rob
>
>

Re: import R data frame into org-mode table

[Nick Dokos]

John Hendy <jw.hendy@gmail.com> writes:

> On Mon, Jul 29, 2013 at 10:35 AM, Cook, Malcolm <MEC@stowers.org> wrote:
>>>Hi Andreas,
>>  >
>>  >On 17 July 2013 23:09, Andreas Leha <andreas.leha@med.uni-goettingen.de> wrote:
>>  >
>>  >> Definitely there is:
>>  >>
>>  >> --8<---------------cut here---------------start------------->8---
>>  >> #+begin_src R :results table :colnames yes
>>  >>   read.csv('test.csv')
>>  >> #+end_src
>>  >>
>>  >> #+results:
>>  >> | X | Variant | Xaxis | N |   mean |   sd |   se |
>>  >> |---+---------+-------+---+--------+------+------|
>>  >> | 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
>>  >> | 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
>>  >> | 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
>>  >> | 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
>>  >> | 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
>>  >> | 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |
>>  >>
>>  >> --8<---------------cut here---------------end--------------->8---
>>  >
>>  >Great. I've gone ahead and done this. There are two additional requirements:
>>  >1) My table needs a caption. Will #+CAPTION: just above the
>>  >#+BEGIN_SRC just work?
>>
>> Yes

I don't think so: the caption is a property of the table, not of the
code block, so putting it before the code block does not work (at least
in my limited 8.x tests). You have to put it before the table - see
example below.

>>
>>> Even better, a label, allowing me to also cross
>>  >reference it.
>>  >2) My table is too long for 1 page. It spans multiple pages
>>  >vertically. According to this StackOverflow answer
>>  >http://stackoverflow.com/a/2896850/1526266 , I should instead use
>>  >longtable, not tabular. Is there a way to coerce your above snippet to
>>  >use longtable, instead of tabular which is the default.
>>  >
>>
>> Looks like your answer is here: http://orgmode.org/manual/Tables-in-LaTeX-export.html
>
> Sort of. Depends on the Org-mode version, and someone will have to
> chime in on the updated status of various parts of the manual. For 8+,
> I believe the syntax is different:
>
> #+attr_latex: :environment longtable
>
> Also, there have been many threads in the past about how to add
> #+attr_latex lines to results output (mostly graphics/files)
> successfully. If you simply take the above and add =#+attr_latex:
> stuff= above the =#+results= line, babel won't recognize it and will
> just create a new results block. If those on this email are already
> well aware of this... my apologies for being redundant, but it causes
> enough confusion that I figured I'd leave another bread crumb trail :)
> Here's an example of a time that came up on the list:
> - http://lists.gnu.org/archive/html/emacs-orgmode/2012-07/msg00237.html
>

This is no longer the case: Eric S. fixed this a couple of months ago,
so that even an unlabeled #+RESULTS block will get the recomputed
results without creating a new block. The following exports fine to
latex/pdf with current org:

--8<---------------cut here---------------start------------->8---

* foo

#+begin_src R :results table :colnames yes :exports results
  read.csv('test.csv')
#+end_src

#+NAME: foo
#+CAPTION: My table
#+ATTR_LATEX: :environment longtable
#+RESULTS:
| X | Variant | Xaxis | N |   mean |   sd |   se |
|---+---------+-------+---+--------+------+------|
| 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
| 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
| 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
| 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
| 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
| 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |


* bar

See Table \ref{foo}.
--8<---------------cut here---------------end--------------->8---

even satisfying the OP's desire for labeling the table so that the xref
will work (although there may be better methods to xref: I haven't kept
up with the recent discussions on the ML).

Of course, naming the code block should still be considered best
practice IMO.

-- 
Nick

Re: import R data frame into org-mode table

[Cook, Malcolm]

>-----Original Message-----
 >From: John Hendy [mailto:jw.hendy@gmail.com]
 >Sent: Monday, July 29, 2013 11:03 AM
 >To: Cook, Malcolm
 >Cc: Rob Stewart; Andreas Leha; emacs-orgmode@gnu.org
 >Subject: Re: [O] import R data frame into org-mode table
 >
 >On Mon, Jul 29, 2013 at 10:35 AM, Cook, Malcolm <MEC@stowers.org> wrote:
 >>>Hi Andreas,
 >>  >
 >>  >On 17 July 2013 23:09, Andreas Leha <andreas.leha@med.uni-goettingen.de> wrote:
 >>  >
 >>  >> Definitely there is:
 >>  >>
 >>  >> --8<---------------cut here---------------start------------->8---
 >>  >> #+begin_src R :results table :colnames yes
 >>  >>   read.csv('test.csv')
 >>  >> #+end_src
 >>  >>
 >>  >> #+results:
 >>  >> | X | Variant | Xaxis | N |   mean |   sd |   se |
 >>  >> |---+---------+-------+---+--------+------+------|
 >>  >> | 1 | line1   |    10 | 5 | 111.11 | 9.33 | 3.11 |
 >>  >> | 1 | line1   |    20 | 5 | 112.11 | 9.13 | 3.14 |
 >>  >> | 1 | line1   |    30 | 5 | 113.11 | 9.43 |  3.1 |
 >>  >> | 1 | line2   |    10 | 5 | 101.11 | 8.33 | 2.11 |
 >>  >> | 1 | line2   |    20 | 5 | 100.11 | 8.13 | 2.12 |
 >>  >> | 1 | line2   |    30 | 5 | 108.11 | 8.03 |  2.1 |
 >>  >>
 >>  >> --8<---------------cut here---------------end--------------->8---
 >>  >
 >>  >Great. I've gone ahead and done this. There are two additional requirements:
 >>  >1) My table needs a caption. Will #+CAPTION: just above the
 >>  >#+BEGIN_SRC just work?
 >>
 >> Yes
 >>
 >>> Even better, a label, allowing me to also cross
 >>  >reference it.
 >>  >2) My table is too long for 1 page. It spans multiple pages
 >>  >vertically. According to this StackOverflow answer
 >>  >http://stackoverflow.com/a/2896850/1526266 , I should instead use
 >>  >longtable, not tabular. Is there a way to coerce your above snippet to
 >>  >use longtable, instead of tabular which is the default.
 >>  >
 >>
 >> Looks like your answer is here: http://orgmode.org/manual/Tables-in-LaTeX-export.html
 >
 >Sort of. Depends on the Org-mode version, and someone will have to
 >chime in on the updated status of various parts of the manual. For 8+,
 >I believe the syntax is different:
 >
 >#+attr_latex: :environment longtable
 >
 >Also, there have been many threads in the past about how to add
 >#+attr_latex lines to results output (mostly graphics/files)
 >successfully. If you simply take the above and add =#+attr_latex:
 >stuff= above the =#+results= line, babel won't recognize it and will
 >just create a new results block. If those on this email are already
 >well aware of this... my apologies for being redundant, but it causes
 >enough confusion that I figured I'd leave another bread crumb trail :)
 >Here's an example of a time that came up on the list:
 >- http://lists.gnu.org/archive/html/emacs-orgmode/2012-07/msg00237.html
 >
 >You need to use any attributes (that includes #+begin/end_center, and
 >any #+attr_backend lines) *in combination with* a named source block.
 >So the full solution should look something like this:
 >
 >#+name: export-table
 >#+begin_src R :results table :colnames yes
 >read.csv('test.csv')
 >#+end_src
 >
 >#+RESULTS: export-table
 >#+attr_latex: :environment longtable
 >| | |
[Cook, Malcolm] 

Indeed, thanks, and, my worked out example follows using emacs 2.18.9 and org-mode version 8.0.6

#+LATEX: \listoftables                                                                                                                                                                                                                       
#+LaTeX_HEADER: \usepackage{longtable}                                                                                                                                                                                                       
#+name: longtabletest                                                                                                                                                                                                                        
#+CAPTION: test of longtable caption                                                                                                                                                                                                         
#+begin_src R :results value :colnames yes                                                                                                                                                                                                   
data.frame(num=1:260,alpha=rep(LETTERS,10))                                                                                                                                                                                                  
#+end_src                                                                                                                                                                                                                                    
                                                                                                                                                                                                                                             
#+RESULTS:                                                                                                                                                                                                                                   
#+attr_latex: :environment longtable                                                                                                                                                                                                         
| num | alpha |                                                                                                                                                                                                                              
|-----+-------|                                                                                                                                                                                                                              
|   1 | A     |                                                                                                                                                                                                                              
|   2 | B     |       
...

~Malcolm

 >
 >I see above that =:results output org= was used. There was some
 >discussion about this for use with the new exporter as I wasn't
 >getting great results:
 >- https://lists.gnu.org/archive/html/emacs-orgmode/2013-03/msg01582.html
 >- http://www.mail-archive.com/emacs-orgmode@gnu.org/msg69748.html
 >
 >If it's working for you, don't worry about it. If not, you may find
 >some help in those threads. I've taken to using the ascii package for
 >output org-mode tables from R.
 >- http://orgmode.org/worg/org-contrib/babel/examples/ascii.html
 >
 >Works great for me.
 >
 >
 >Good luck!
 >John
 >
 >
 >>
 >>  >Thanks!
 >>  >
 >>  >--
 >>  >Rob
 >>
 >>

Re: import R data frame into org-mode table

[Nick Dokos]

"Cook, Malcolm" <MEC@stowers.org> writes:

>
> Indeed, thanks, and, my worked out example follows using emacs 2.18.9 and org-mode version 8.0.6
>

2.18.9?

> #+LATEX: \listoftables
> #+LaTeX_HEADER: \usepackage{longtable}
> #+name: longtabletest
> #+CAPTION: test of longtable caption
> #+begin_src R :results value :colnames yes
> data.frame(num=1:260,alpha=rep(LETTERS,10))
> #+end_src
>
> #+RESULTS:
> #+attr_latex: :environment longtable
> | num | alpha |
> |-----+-------|
> |   1 | A     |
> |   2 | B     |       
> ...

Does this really produce a table caption for you when you export to
latex? And does the attr_latex line stay there at the top of the table
when the code block is evaluated?  For me, the answers are "no" and
"no". I have to rewrite it as follows in order for it to work:

> #+begin_src R :results value :colnames yes
> data.frame(num=1:260,alpha=rep(LETTERS,10))
> #+end_src
>
> #+CAPTION: test of longtable caption
> #+attr_latex: :environment longtable
> #+RESULTS:
> | num | alpha |
> |-----+-------|
> |   1 | A     |
> |   2 | B     |

-- 
Nick

Re: import R data frame into org-mode table

[Cook, Malcolm]

>>
 >> Indeed, thanks, and, my worked out example follows using emacs 2.18.9 and org-mode version 8.0.6
 >>
 >
 >2.18.9?

oops, that was the GTK_ Version number.  My emacs is 24.3.1.

>> #+LATEX: \listoftables
 >> #+LaTeX_HEADER: \usepackage{longtable}
 >> #+name: longtabletest
 >> #+CAPTION: test of longtable caption
 >> #+begin_src R :results value :colnames yes
 >> data.frame(num=1:260,alpha=rep(LETTERS,10))
 >> #+end_src
 >>
 >> #+RESULTS:
 >> #+attr_latex: :environment longtable
 >> | num | alpha |
 >> |-----+-------|
 >> |   1 | A     |
 >> |   2 | B     |
 >> ...
 >
 >Does this really produce a table caption for you when you export to
 >latex?
> And does the attr_latex line stay there at the top of the table
 >when the code block is evaluated?  

Yes and Yes.  At least, I swear it worked in my hands yesterday, but today the answer is No and No.  !!  I can't explain what has changed.

> For me, the answers are "no" and
 >"no". 
> I have to rewrite it as follows in order for it to work:
 >
 >> #+begin_src R :results value :colnames yes
 >> data.frame(num=1:260,alpha=rep(LETTERS,10))
 >> #+end_src
 >>
 >> #+CAPTION: test of longtable caption
 >> #+attr_latex: :environment longtable
 >> #+RESULTS:
 >> | num | alpha |
 >> |-----+-------|
 >> |   1 | A     |
 >> |   2 | B     |

In my hands, both the #+CAPTION and the #+att_latex as you place them are deleted upon source block evaluation. 

Are you sure your placement works for you?

Summary: In my hands today, there is no placement of them that both survives source block re-evaluation and works.  And, they can both be placed immediately above or below the #+RESULTS: line and are respected on export.

(However, another funny interaction, the caption of the longtable has multiple entries, one for each page, in my #+LATEX: \listoftables )

Hmmmm....

 >
 >--
 >Nick
 >
 >