RS-DBI using embedded Perl DBI

All,

I've attached my implementation of David James RS-DBI interface using Perl DBI  (R Package).  I've found it to be quite quick.  The nice thing is you can connect to any database perl can.

The only big difference in S code is, you now have to specify the type of database your connecting to as well as host and port if applicable.  

Great. Saikat, David and I did the same sort of thing with JDBC.  It
is a good idea to leverage the existing code for the various
databases.  In S-JDBC interface, we were using the SJava package to
actually make the connection to Java. You appear to be using direct C
code to connect to Perl rather than the RSPerl package which provides
a way for embedding Perl in R. I do recognize some of the code that
seems to be taken from that package, so perhaps you are using it but I
don't thing directly, if it all.  You might find that using that
package simplifies the code a bit and makes it easier for others to
experiment with.

I've only tested this with Oracle. An example is:

library(RSPerlDBI)
ora <- dbManager("PerlDBI")
con <- dbConnect(ora, user = "query", password = "query", dbname = "rduxu",dbtype = "Oracle")
result <- quickSQL(con, "select * from ALL_TABLES")   

Unfortunetly if theres any Perl DBI errors it kills R.  Any ideas Duncan?

Yep. The problem is that some errors in Perl jumps straight to exit()
and _exit() rather than signalling an exception and returning control
to the caller, in this case the C code invoked from R.  The RSPerl
package has the same difficulty and I haven't found time to fix it
yet. A simple solution is to provide an alternative implementation of
Perl's fatal error handler routine. Then, when one links against the
Perl library to create RSPerlDBI.so, the routine in your C code will
be linked and used. This can be a bit tricky at times depending on
linkers, how Perl was compiled, etc. But, if your routine is called,
then it should throw and S error and the execution stack will return.
That's the basic approach I use in embedding unless I have more
specific needs.


BTW, I am on the r-sig-db mailing list, so there's no need mail me
separately.

Feel free to make suggestions, etc...

Thanks,

Jake

-------------------------
T Jake Luciani
Software Developer
Optoelectronics Unit
Agere Systems
jake at agere.com

_______________________________________________________________

Duncan Temple Lang                duncan at research.bell-labs.com
Bell Labs, Lucent Technologies    office: (908)582-3217
700 Mountain Avenue, Room 2C-259  fax:    (908)582-3340
Murray Hill, NJ  07974-2070       
         http://cm.bell-labs.com/stat/duncan