Re: Migrate from 8i to 9i with different NLS_LENGTH_SEMANTICS

Hi Kenneth, Pat, and Mark!

I thank you all very much for your help, it was great, for it put me on the right way. I got it all finished last night and could send my solution to my customer.

That's the way I did it:

I took a regular export file from the 8iDB and made an import with "show=y", as Kenneth had suggested. Output was that pseudo-SQL code of 8,000 lines as already mentioned. I wrote a little Java program that did the concatination of the broken lines and put a semi-colon at the end of what probably was a complete line. I could delete all the code that did not define tables, and that was many thousands of lines. In the remaining code there were some syntax errors left, due to some formatting failures of my Java program, which is no SQL parser, of course. All in all the work of a few hours.

I ran the SQL against a DB with Unicode character set and "char" length semantics. That built me a DB structure with the length semantics desired. (Remember that a simple imp ignores the "char" length semantics with source data of "byte" length semantics.) After that, I started a regular import with rows and grants and indexes and constraints and -- right Mark! -- "ignore=y". Done.

Though, next time I'll have a look on TOAD... (thanks for the URL, Pat.)

Bye,

Tosta. Received on Tue Aug 10 2004 - 05:13:53 CDT