Interface to R

R is a computer language widely used in statistics. It includes a huge library of statistical and graphical functions which can be called from APL. For example, you can call R to do linear and non-linear modelling, classical statistical tests, time-series analysis, classification, clustering and many other tasks.

You can call the R library in two ways:

• If you are running Dyalog or APLX or any other COM-capable APL, you can call R through the COM interface.
  • This method only works under Windows
• If you are running APLX Version 5 or later, you can call R directly through APLX's plug-in architecture
  • This method works under Windows, Macintosh and Linux and is more tightly integrated, but requires APLX Version 5.

The two methods are described below.

Using COM Server (Dyalog)

Installation

1. Install R statistical environment from http://www.r-project.org

2. Install R package rcproxy from http://cran.rakanu.com/bin/windows/contrib/2.8/rscproxy_1.2-0.zip

3. Install DCOM Server from http://cran.r-project.org/contrib/extra/dcom/

Sample APL Session

     )load r2apl
d:\r2apl saved Mon Apr 06 22:10:59 2009
     rinit ⍝ initialize connection to R COM server
     )obs  ⍝ new object R is created
R
      +x←10?100         ⍝ random numbers in APL
14 76 46 54 22 5 68 94 39 52
      'y'rput x         ⍝ send them to R
      1 rexec 'mean(y)' ⍝ calculate mean value in R
47
      +/x÷⍴x            ⍝ the same in APL
47
      y←rget 'y'        ⍝ get variable y from R to APL
      x≡y               ⍝ compare with original x
1
      y←?100 6⍴2*30     ⍝ generate another random set
      y←(+/y÷2*30)÷6
      'y'rput y         ⍝ put it in R
      rexec 'hist(y)'   ⍝ call R function to plot histogram

And R windows with histogram plot pops up:

Dyalog APL functions

rinit

Initialize APL connection to R COM Server.

     ∇ rinit
[1]    '#.R'⎕WC'OleClient' 'StatConnectorSrv.StatConnector'
[2]    #.R.Init('R')
     ∇

rexec

Execute R expression and possibly return the result to APL

     ∇ a←{r}rexec exp
[1]    ⍝ Execute R expression and possibly return the result
[2]    ⍝R: R expression ('char')
[3]    ⍝L: if 1, then return the result
[4]    ⍎(0=⎕NC'r')/'r←0'
[5]    a←⍬
[6]    :If r
[7]        a←#.R.Evaluate(exp)
[8]    :Else
[9]        #.R.EvaluateNoReturn(exp)
[10]   :EndIf
     ∇

rput

Put APL array into R

     ∇ {m}rput a;x1
[1]    ⍝ Put APL array into R
[2]    ⍝R-name of variable to export from APL
[3]    ⍝L-name of variable to create in R
[4]    ⍎(0=⎕NC'm')/'m←''x''' ⍝ Default name is x
[5]    m←,⊂m
[6]    a←,⊂a
[7]    #.R.SetSymbol(m,a) ⍝ put var a into R with name m
     ∇

rget

Get R variable (matrix or vector) to APL

     ∇ r←rget name;rho;Re;Im;⎕ML
[1]   ⍝ Get R variable (matrix or vector) to APL
[2]   ⍝R-name of variable to import ('char')
[3]    ⎕ML←3
[4]    r←#.R.GetSymbol(name)
     ∇

Get workspace here

You can download a copy of the workspace here

• Dyalog

  • r2apl.dws

• APLX (Equivalent functions to the ones listed above)

  • r2aplx.aws

Using R from APLX Version 5

APLX includes the ability to call code written in other languages like .NET, Java, Ruby and R through an external plug-in mechanism. This allows a tighter integration of APL and R than you can achieve with the COM interface. For example, it allows support for:

• complex R data types (data frames, factors, time series, etc),
• NA (not-available) values
• attributes

• access to the text form of a variable via ⎕DS

Sample APLX Session

      ⍝ Load the plugin library
      r←'r' ⎕new 'r'         
      r                       
[r:R]
      r.⎕DS
R version 2.9.0 (2009-04-17)

      ⍝ First some simple stuff...      
      r.x←10?10              ⍝ APL data assigned to R variable 'x'
      r.x
2 8 5 6 3 1 7 10 4 9
      r.⎕EVAL 'mean(x)'      ⍝ Evaluate arbitrary expression
5.5
      r.mean (⊂10?10)        ⍝ No need to use an R variable
5.5
      r.y←?100 6⍴2*30        ⍝ Generate another random set
      r.y←(+/r.y÷2*30)÷6
      r.⎕EVAL 'hist(y)'      ⍝ Call R function to plot histogram 
[r:histogram]                ⍝ (Shows the same histogram as in the 
                             ⍝  Dyalog example above)

      ⍝ Let's create an R time series...    
      ts←r.ts (⊂12?100)           
      ts
[r:ts]
      ts.⎕DS                 ⍝ Examine text form
Time Series:
Start = 1
End = 12
Frequency = 1
 [1] 93 94 45 53 84  1 63 75 56 54 25 98

      ⍝ R data's attributes can be accessed and changed
      ⍝ using the ∆XXX syntax...  
      ts.attributes.⎕DS      ⍝ Get the attributes
$tsp
[1]  1 12  1

$class
[1] "ts"

      ts.∆tsp
1 12 1
      ts.∆tsp←2008 (2008+11÷12) 12   ⍝ Change the start/end time and frequency
      ts.⎕DS
     Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
2008  93  94  45  53  84   1  63  75  56  54  25  98

      ⍝ An example of a data frame...
      ⍝ Get the built-in data set giving eruption data for the Old Faithful geyser
      r.faithful                      
[r:frame]                              

      ⍝ Examine the data as text
      200 ↑r.faithful.⎕DS
    eruptions waiting
1       3.600      79
2       1.800      54
3       3.333      74
4       2.283      62
5       4.533      85
6       2.883      55
7       4.700      88
8       3.600      85
9
      ⍝ Get a named column from the data frame
      eruptions←(r.faithful).$$ 'eruptions'
      10↑eruptions
3.6 1.8 3.333 2.283 4.533 2.883 4.7 3.6 1.95 4.35

      ⍝ Plot a histogram of the eruption durations in minutes 
      r.eruptions←eruptions    
      r.⎕EVAL 'hist(eruptions,seq(0,6,0.2),prob=TRUE)'                                                     
[r:histogram]
      ⊣ r.⎕EVAL 'lines(density(eruptions,bw=0.1))'

      ⍝ For eruptions of more than 3 minutes' duration,
      ⍝ apply Kolmogorov-Smirnov test to check whether duration
      ⍝ obeys a normal distribution...

      r.long←(eruptions > 3)/eruptions
      ks←r.⎕EVAL 'ks.test (long, "pnorm", mean=mean(long), sd=sd(long))'
      ks
[r:htest]
      ks.⎕DS                                    

 One-sample Kolmogorov-Smirnov test

data:  long
D = 0.0661, p-value = 0.4284
alternative hypothesis: two-sided

      ks.⎕VAL
 0.06613335935  0.4283591902  two-sided One-sample Kolmogorov-Smirnov test

Author: AlexanderSkomorokhov 2009-04-06

APLX material: SimonMarsden 2009-06-30

CategoryTechnologies