Front page | perl.dbi.users |
Postings from September 2004
(I think it was RC7 - the one released end of august this year) Originally I had an interesting problem - when I'd prepare a query that called a function taking nvarchar2's as arguments it would work on the first execution, but in subsequent executions Oracle would pass NULL's for arguments to my functions. I opened a TAR with Oracle about it and was able to make it happen with plain OCI. They discovered the problem was I wasn't setting OCI_ATTR_CHARSET_FORM on my bind variables. I put this code in and it worked. So I went back to DBD::Oracle to see how to do it. (Since our app is in perl) I see I need to use 1.16, which is an RC, but 1.15 has zero support for this. After some reading I tried it out, setting the csform value on the bind_param call. When I ran my test app I'd get "DBD::Oracle::st execute failed: ORA-01460: unimplemented or unreasonable conversion requested (DBD ERROR: error possibly near <*> indicator at char 51 in 'selec..." - Odd. I put in a bunch of debugging code into DBD and found it too was setting that attribute so I'm a bit confused as to why it is happening. I cannot seem to make any progress on this. Useful Information: This seems to happen on both perl 5.8.0 and 5.8.3 Client is Fedora Core 2 or RedHat 8 x86. Oracle is 9.2.0.5 running on Solaris 64 bit. Oracle Lib (linux) are 9.2.0.4 DB is in AL16UTF16 $NLS_LANG is American_America.WE8ISO8859P1 Steps to reproduce: Load Schema into an AL16UTF16 (default in 9.2) DB: http://www.jefftrout.com/~threshar/oracle/brokennvarchar.sql Grab perl script: http://www.jefftrout.com/~threshar/oracle/brokennvarchar.pl Modify the DBI connect string accordingly. then run it with the -broken argument. (The -broken, etc. arguments used to be used to show my original bug in its various forms) ./brokennvarchar.pl -broken You will likely see output along these lines: DBD::Oracle::st execute failed: ORA-01460: unimplemented or unreasonable convers ion requested (DBD ERROR: error possibly near <*> indicator at char 51 in 'selec t brokennvarchar.get_id_broken(:p1, :p2) from <*>dual') [for statement ``select brokennvarchar.get_id_broken(?, ?) from dual'' with params: :p1='thresharpd2', : p2='potato']) at brokennvarchar.pl line 95. execute failed: ORA-01460: unimplemented or unreasonable conversion requested (D BD ERROR: error possibly near <*> indicator at char 51 in 'select brokennvarchar .get_id_broken(:p1, :p2) from <*>dual') DBD::Oracle::st fetch failed: ERROR no statement executing (perhaps you need to call execute first) [for statement ``select brokennvarchar.get_id_broken(?, ?) f rom dual'' with params: :p1='thresharpd2', :p2='potato']) at brokennvarchar.pl l ine 105. Failed to fetch ERROR no statement executing (perhaps you need to call execute f irst) And finally, a small C program using OCI to show it working properly: http://www.jefftrout.com/~threshar/oracle/brokennvarchar.c make -f $ORACLE_HOME/rdbms/demo/demo_rdbms.mk build EXE=brokennv archar OBJS=brokennvarchar.o ./brokennvarchar -b you should see it successfully get ID "100" 5 times. (If you run with -w you'll see it work the first time and fail the rest. That was my original problem with the nvarchar's but now it affects the regular varchar's. sigh) Any pointers you could give me would be terrific! -- Jeff Trout <jeff@jefftrout.com> http://www.jefftrout.com/ http://www.stuarthamm.net/Thread Next