[Rd] C_LogLin (stats/loglin)

[Gabriel Becker]

Kai,

Apologies for the double message, it didn't go to the list last time.

On Thu, Oct 22, 2015 at 9:59 AM, Kai Nitschke <
kai.nitschke at uniklinik-freiburg.de> wrote:

>
> “.Call” calls the C/C++ function “C_LogLin”. But when I am running it line
> by line I get
> the following error on line 23/24:
> Error: object 'C_LogLin' not found
>
> Hence, my questions:
> 1. Where does the original “loglin” function get the “C_LogLin” function
> from (some libraries? where are they stored)?
>

The libraries are in the stats package, built from the c code found in the
/src  directory of the package's source code.


> 2. Why does it not work in the normal R command (line by) line but in the
> implemented “loglin” function?
>

Notice that C_LogLin is not in quotes. That is actually an R object, not a
string, and it is one that is not exported by the stats package.
stats:::C_LogLin would probably work.


> 3. Is it possible to isolate the “C_LogLin” function so I can use it in
> other scripts or programming languages e.g. Matlab?
>


A function called by the .Call interface is going to operate on R's
internal representation of R objects, called SEXPs, It won't work in other
languages without major reworking (I can't say how major without looking
for at the code to see if, e.g., it's just a wrapper that calls another c
function that might be amenable to that). Even then, you'd have to look at
if/how it deals with NAs, etc.

~G

On Thu, Oct 22, 2015 at 9:59 AM, Kai Nitschke <
kai.nitschke at uniklinik-freiburg.de> wrote:

> Hi everyone,
>
> I have a question regarding a C function of the "stats" package in R.
>
> I tried to understand the “loglin” basic function of the “stats” package
> implemented in
> R. The implemented function itself runs without any problem (perhaps see
> sample). When I
> ran it line by line it stopped at the lines 23-24 of the loglin-function;
> (the following line):
> z <- .Call(C_LogLin, dtab, conf, table, start, nmar, eps, iter)
>
> “.Call” calls the C/C++ function “C_LogLin”. But when I am running it line
> by line I get
> the following error on line 23/24:
> Error: object 'C_LogLin' not found
>
> Hence, my questions:
> 1. Where does the original “loglin” function get the “C_LogLin” function
> from (some libraries? where are they stored)?
> 2. Why does it not work in the normal R command (line by) line but in the
> implemented “loglin” function?
> 3. Is it possible to isolate the “C_LogLin” function so I can use it in
> other scripts or programming languages e.g. Matlab?
>
> Thank you in advance for your help,
>  Kai
>
> --------------------------------------------
> #Sample
>
> # Example data
> mytable = array(0,dim=c(3,4,5))
> mytable[1,,1] = c(35,26,36,29); mytable[2,,1] = c(21,12,26,27);
> mytable[3,,1] = c(9,26,23,38);
> mytable[1,,2] = c(43,26,89,26); mytable[2,,2] = c(78,42,32,9);
> mytable[3,,2] = c(34,39,42,51);
> mytable[1,,3] = c(51,72,62,21); mytable[2,,3] = c(6,52,71,32);
> mytable[3,,3] = c(35,36,48,21);
> mytable[1,,4] = c(12,26,53,42); mytable[2,,4] = c(96,75,81,56);
> mytable[3,,4] = c(32,65,75,63);
> mytable[1,,5] = c(21,18,24,33); mytable[2,,5] = c(22,56,36,91);
> mytable[3,,5] = c(32,65,63,14);
>
> # Call of the original function works fine
> loglin(mytable,list(c(1,2),c(1,3)))
>
> # line by line does not work
> table = mytable; margin = list(c(1,2),c(1,3)); start = rep(1,
> length(table)); fit = FALSE; eps = 0.1;
> iter = 20L; param = FALSE; print = TRUE;
> rfit <- fit
> dtab <- dim(table)
> nvar <- length(dtab)
> ncon <- length(margin)
> conf <- matrix(0L, nrow = nvar, ncol = ncon)
> nmar <- 0
> varnames <- names(dimnames(table))
> for (k in seq_along(margin)) {
>         tmp <- margin[[k]]
>         if (is.character(tmp)) {
>             ## Rewrite margin names to numbers
>             tmp <- match(tmp, varnames)
>             margin[[k]] <- tmp
>         }
>         conf[seq_along(tmp), k] <- tmp
>         nmar <- nmar + prod(dtab[tmp])
> }
> ntab <- length(table)
> if (length(start) != ntab ) stop("'start' and 'table' must be same length")
> z <- .Call(C_LogLin, dtab, conf, table, start, nmar, eps, iter)
> ...
>
>
> --
> Kai Nitschke, Dipl.-Psych.
>
> Division of Functional Brain Imaging (FBI)
> Dept. of Neurology, University of Freiburg
>
> Breisacher Str. 64, 79095 Freiburg, Germany
> Phone: +49 (0) 761 270 - 50410
> Email: kai.nitschke at uniklinik-freiburg.de
> URL: http://www.uniklinik-freiburg.de/fbi/people/kai-nitschke.html
>
> ______________________________________________
> R-devel at r-project.org mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>



-- 
Gabriel Becker, PhD
Associate Scientist (Bioinformatics)
Genentech Research

	[[alternative HTML version deleted]]