Re: Social network analysis and Perl

[Roger]

If you only want to write an analysis package to assist your research on Social Network Analysis, then you could try to download one of these software packages from the International network for Social Network Analysis. Afterall, why reinvent the wheel if there is already a freeware to do what you want. :-)

You mentioned that the RSPerl package was not working properly on your system, can you please describe what is happening and why you think it is not working in detail, so we can perhaps help you to rectify the problem.

By the way, the latest R is version 1.8 on http://cran.r-project.org/, perhaps you should consider an upgrade...

[jmerelo]

Most programs are buggy, for Windows or both. They break down with large networks. I could give them a try using Windows automation from Perl. I could also try to make R work from Perl and give it a go go the SNA package with R, which is extremely slow. Not that writing it in Perl would speed it up a bit, but at least, I could handle data in a format I am familiar with. With respect to RSPerl, it's probably a perfectly good thing, but a) it's not on CPAN and b) googling "RSPerl problems" returns more than 200 hits. It's just weird to see that a popular thing such as Social Network Analysis does not have a Perl library, or a Perl interface to an existing library. I'll update R, anyways.

[jmerelo]

I have upgraded to R 1.8. Installation of RSPerl, from R, works fine, but following the not-so-clear instructions for making R work from Perl issues such an amount of errors...

cc1: aviso: cambiando el orden de búsqueda para el directorio del sis
+tema "/usr/local/include"
cc1: aviso:   porque ya ha sido especificado cmo un directorio que no 
+es del sistema
En el fichero incluído de R.xs:5:
RSCommon.h:49:26: Rinternals.h: No existe el fichero o el directorio
RSCommon.h:50:24: Rdefines.h: No existe el fichero o el directorio
En el fichero incluído de R.xs:5:
RSCommon.h:65: error de decodificación antes de "USER_OBJECT_"
RSCommon.h:65: aviso: la definición de datos no tiene tipo o clase de
+ almacenamiento
RSCommon.h:67:24: Rversion.h: No existe el fichero o el directorio
RSCommon.h:68:28: falta un operador binario antes de `('
RSCommon.h:76:29: R_ext/Boolean.h: No existe el fichero o el directori
+o
En el fichero incluído de R.xs:6:

... and so on, up to 200 hundred, that I think it just does not work with Perl >= 5.6. The web page says it works for 5.005 and R 1.3; it keeps working for R 1.8, but it no longer works with Perl.

[Anonymous Monk]

So why don't you write your R script manually and just shell out to rcmd.exe with Perl? Typically I read data from a db with Perl/DBI, write it to a file, dynamically write an R script, shell to R (which opens the file, analyzes it, writes results to text), Perl validates the results, parses file, loads analyzed data back to DB. All is good in the world.

[Roger]

Ummm, could you translate the error messages to English please?

I am *guessing* that you are using version 0.5.7 of RSPerl, which contains C files hardwired to Perl 5.0's XS. Unfortunately the guts of Perl has changed much since then, you just can not expect perl XS to be backward compatible with older versions. And from what I see, the guts of R has also changed dramatically since then...

You could try to fix RSPerl's C files by resolving each error one-by-one. Probably not... Why don't you email an request to the Author of RSPerl and ask him for the upgrade?

Anonymous Monks's suggestion looks good...

[jmerelo]

Never mind the messages, it's just about missing files, and lots of other warnings and stuff... I just wanted to show there are lots of error messages. And yes, that was my hunch. I'd rather write a social network analysis module in Perl myself than fix somebody else's C code. But I'll email him, anyways. Using R from perl is a good idea.