Re: [O] Bug: Python SRC exec tuple fails [7.9.3f (release_7.9.3f-17-g7524ef <at> MY-PATH/)]

[Roland Donat]

Andreas Röhler <andreas.roehler <at> easy-emacs.de> writes:

> 
> Am 08.05.2013 22:50, schrieb Roland Donat:
> >>> Yes, you're right Andreas. It "fails" to show the accented characters 
if
> > you
> >>> try to print the entire tuple.
> >>> It fails too if you evaluate a[0][0] in your interpreter. You should 
see
> > :
> >>>>>> a[0][0]
> >>> '\xc3\xa9'
> >>> But print a[0][0] gives the expected answer 'é'
> >>>
> >>> So, based on your successful experience consisting in returning a[0]
[0]
> > in
> >>> the orgmode source block, we can assume that org-babel use the python
> > print
> >>> function to display results in org buffer, aren't we?
> >>>
> >>> Another strange behaviour, when you evaluate the src_block test given 
in
> >>> example, you get :
> >>> | \303\251 | a        |
> >>> | a        | \303\240 |
> >>>
> >>> Whereas I was expecting to get the same code than in the python
> > interpreter,
> >>> that is :
> >>> | \xc3\xa9 | a          |
> >>> | a        | '\xc3\xa0' |
> >>>
> >>> In addition, when I try to save my buffer, Emacs doesn't recognize the
> >>> encoding of characters \303\251 and \303\240 and asks me to choose an
> >>> encoding. Then, I enter utf-8 and nothing happens BUT when I quit and
> > reopen
> >>> my file : the characters are printed correctly.... Too strange for
> > me....
> >>>
> >>> Cheers,
> >>>
> >>> Roland.
> >>
> >> so what about that:
> >>
> >> a = ( ( "é", "a" ), ( "a", "à" ) )
> >> for i, j in a:
> >>       print i, j
> >>
> >> BTW previous post was sent prematurely..
> >>
> >> Andreas
> >>
> >>
> >
> > Yep, using a couple of for loops will work but the result won't return 
as a
> > table which is a requirement for me.
> >
> > To precise the context a littre more, I have basically 2 source blocks :
> > 1) the famous python block which must return a table
> > 2) a R block used to post-process the previous table
> >
> > Well, thanks for your help.
> > I think I spent too much time on this so I'm thinking about changing my
> > approach. For example, put the result of the first step into a file and 
then
> > process the file in step 2.
> >
> > Best regards,
> >
> > Roland.
> 
> Just playing a little bit with your example, what about this:
> 
> #+begin_src python :results output :preamble # -*- coding: utf-8 -*-
> a = ( ( "é", "a" ), ( "a", "à" ) )
> for i, j in a:
>      print("|%s | %s|" % (i, j))
> #+end_src
> 
> 
Yes Andreas! It works just fine for the python block. But when the python 
result arrives as input of my R post
processing code, R receives the following string "| é | a |\n| a | à |\n" 
whereas a data.frame is
expected. Indeed, theoretically, when a org-table is used as input of a R 
source block, the
org-table is automatically converted into a data.frame which is very 
convenient to handle data.

Here is my buffer :

#+name: pytab
#+begin_src python :results output raw :preamble # -*- coding: utf-8 -*-
a = ( ( "é", "a" ), ( "a", "à" ) )
for i, j in a:
    print "| {0} | {1} |".format( i, j )
#+end_src

#+TBLNAME: pytab
| é | a |
| a | à |

#+name: Rpostproc
#+begin_src R :results output :session :preamble # -*- coding: utf-8 -*- 
:var tab=pytab
print( tab )
#+end_src

#+RESULTS: Rpostproc
: [1] "| é | a |\n| a | à |\n"

In fact, converting the results of my python block into a string looking 
like an org-table is what I
used to do before I had to use a R block to post process results. 

I assume that org-babel asks for a re-evaluation of "pytab" source block 
when "pytab" is used as argument of another source block. 

The problem stems from the fact that it's the direct evaluation of "pytab" 
(a string) which is used and
not the org-table converted result as shown in the buffer.  

Well, I could just convert the R string into a data.frame but it's very 
complete with my real data
(non trivial separator problems).

A solution to this problem could be to force org-babel to take as argument 
of the R code what is
really shown in the buffer and not the direct result of the python block. 
Another solution could be
to stop the re-evaluation of the R arguments.

Anyway, thanks to spend time on this Andreas!