[Orgmode] Re: [Org-Babel] and R... non-numeric cells

[Dan Davison]

Sébastien Vauban <address@hidden>
writes:

> Hello,
>
> For a report I'm writing, I've been helped by a colleague of mine (let's call
> him Albert) for the R graphics generation.
>
> Here's an extract of my doc:
>
> #+TBLNAME: investissement-2010-2013
> #+ATTR_LaTeX: align=lSSSS
> |                        | \s{Année 2010} | \s{Année 2011} | \s{Année 2012} | 
> \s{Année 2013} |
> |------------------------+----------------+----------------+----------------+----------------|
> | RFO                    |     2596376.30 |     1500000.00 |      500000.00 | 
>      500000.00 |
> | RFO réseau structurant |     3804467.00 |     6534066.00 |     3804467.00 | 
>           0.00 |
> | Équipements            |     1000000.00 |      150000.00 |       50000.00 | 
>       50000.00 |
> |------------------------+----------------+----------------+----------------+----------------|
> | Total (HTVA)           |     7400843.30 |     8184066.00 |     4354467.00 | 
>      550000.00 |
> #+TBLFM: 
> @5$2=vsum(@address@hidden);%.2f::@5$3=vsum(@address@hidden);%.2f::@5$4=vsum(@address@hidden);%.2f::@5$5=vsum(@address@hidden);%.2f
>
> whose graphical representation is:
>
> #+srcname: barplot-investment(ptable = investissement-2010-2013)
> #+begin_src R :file 1-01-investissement-2010-2013.png :exports none :session
> source("mcplot.R", local=TRUE)
> ## select the last row only, exclude first column, scale: unit = 1M
> alldata <- as.matrix(ptable[2:4, -1]) / 1000000
> axisLabels <- c("Année", "Montant HTVA (M€)")
> mcStackedBarplot(alldata, "Investissements", c(2010:2013), 
> ptable[-nrow(ptable),1], legend.location="topright")
> #+end_src
>
> That works perfectly for him (on Ubuntu 9.04, R 2.7.1, Emacs 22.2.1, Org 6.35)
>
> Not for me... on Ubuntu 10.04, R 2.10.1, Emacs 23.1.1, Org 7.01, ESS 5.10: I
> get the message
>
>     *Error in as.matrix(ptable[2:4, -1])/1e+06 :
>      non-numeric argument to binary operator*

Hi Seb,

The problem is that Org has not automatically recognised that you want
the first row to be treated as column names. The reason is that your
table contains multiple horizontal separator lines. So for this table,
you must use ':colnames yes' to tell Org that your first row contains
column names. See sections 14.8.2.12, 14.8.2.13 and 14.8.2.14 in the
manual.

As Erik pointed out, using str is a very good way to investigate your
data structures. Notice below that your column names have become the
first row of the data frame, and as a consequence the columns of the
data frame have been force to character rather than numeric:

--8<---------------cut here---------------start------------->8---
#+srcname: barplot-investment(ptable = investissement-2010-2013)
#+begin_src R :XXX-file 1-01-investissement-2010-2013.png :exports none 
:session :results output
str(ptable)
#+end_src

#+results: barplot-investment
: 
: 'data.frame': 5 obs. of  5 variables:
:  $ V1: chr  "" "RFO" "RFO réseau structurant" "Équipements" ...
:  $ V2: chr  "\\s{Année 2010}" "2596376.3" "3804467.0" "1000000.0" ...
:  $ V3: chr  "\\s{Année 2011}" "1500000.0" "6534066.0" "150000.0" ...
:  $ V4: chr  "\\s{Année 2012}" "500000.0" "3804467.0" "50000.0" ...
:  $ V5: chr  "\\s{Année 2013}" "500000.0" "0.0" "50000.0" ...
--8<---------------cut here---------------end--------------->8---


However, if we use :colnames yes then we have

--8<---------------cut here---------------start------------->8---
#+srcname: barplot-investment(ptable = investissement-2010-2013)
#+begin_src R :XXX-file 1-01-investissement-2010-2013.png :exports none 
:session :results output :colnames yes
str(ptable)
#+end_src

#+results: barplot-investment
: 
: 'data.frame': 4 obs. of  5 variables:
:  $ X              : chr  "RFO" "RFO réseau structurant" "Équipements" "Total 
(HTVA)"
:  $ X.s.Année.2010.: num  2596376 3804467 1000000 7400843
:  $ X.s.Année.2011.: num  1500000 6534066 150000 8184066
:  $ X.s.Année.2012.: num  500000 3804467 50000 4354467
:  $ X.s.Année.2013.: num  500000 0 50000 550000
--8<---------------cut here---------------end--------------->8---


And in this case you probably want ':rownames yes', which will mean you
no longer need the -1 in the indexing expression in your R code.

--8<---------------cut here---------------start------------->8---
#+srcname: barplot-investment(ptable = investissement-2010-2013)
#+begin_src R :XXX-file 1-01-investissement-2010-2013.png :exports none 
:session :results output :colnames yes :rownames yes
str(ptable)
options(width=100)
head(ptable)
#+end_src

#+results: barplot-investment
#+begin_example

'data.frame':   4 obs. of  4 variables:
 $ X.s.Année.2010.: num  2596376 3804467 1000000 7400843
 $ X.s.Année.2011.: num  1500000 6534066 150000 8184066
 $ X.s.Année.2012.: num  500000 3804467 50000 4354467
 $ X.s.Année.2013.: num  500000 0 50000 550000
                       X.s.Année.2010. X.s.Année.2011. X.s.Année.2012. 
X.s.Année.2013.
RFO                            2596376         1500000          500000          
500000
RFO réseau structurant         3804467         6534066         3804467          
     0
Équipements                    1000000          150000           50000          
 50000
Total (HTVA)                   7400843         8184066         4354467          
550000
#+end_example
--8<---------------cut here---------------end--------------->8---


Dan

> As 1M is numeric, the non-numeric operand must be
> =as.matrix(ptable[2:4, -1])=... Verification:
>
>> ptable
>                       V1              V2              V3              V4
> 1                        \\s{Année 2010} \\s{Année 2011} \\s{Année 2012}
> 2                    RFO       2596376.3       1500000.0        500000.0
> 3 RFO réseau structurant       3804467.0       6534066.0       3804467.0
> 4            Équipements       1000000.0        150000.0         50000.0
> 5           Total (HTVA)       7400843.3       8184066.0       4354467.0
>                V5
> 1 \\s{Année 2013}
> 2        500000.0
> 3             0.0
> 4         50000.0
> 5        550000.0
>
>> as.matrix(ptable[2:4, -1])
>   V2          V3          V4          V5        
> 2 "2596376.3" "1500000.0" "500000.0"  "500000.0"
> 3 "3804467.0" "6534066.0" "3804467.0" "0.0"     
> 4 "1000000.0" "150000.0"  "50000.0"   "50000.0" 
>
> The numerics are written between double quotes... Why!?
>
> I had temporarily patched the above problem in my document by updating the 
> line
> with the assignment:
>
> #+srcname: barplot-investment-sva(ptable = investissement-2010-2013)
> #+begin_src R :file 1-01-investissement-sva-2010-2013.png :exports none 
> :session
> source("mcplot.R", local=TRUE)
> ## select the last row only, exclude first column, scale: unit = 1M
> alldata <- matrix(as.numeric(as.matrix(ptable[2:4, -1])), nrow=3, ncol=4) / 
> 1000000
> axisLabels <- c("Année", "Montant HTVA (M€)")
> mcStackedBarplot(alldata, "Investissements", c(2010:2013), ptable[2:4,1], 
> legend.location="topright")
> #+end_src
>
> and I even just noticed that, instead of complexifying the expression, I can
> simplify it, in my case, as my =ptable= is numeric already:
>
>> ptable[2:4, -1]
>          V2        V3        V4       V5
> 2 2596376.3 1500000.0  500000.0 500000.0
> 3 3804467.0 6534066.0 3804467.0      0.0
> 4 1000000.0  150000.0   50000.0  50000.0
>
> But I still don't understand what is reponsible of a different treatment of
> string and numerics between our 2 machines.
>
> Any idea?
>
> Best regards,
>   Seb