Hmm, at first I thought that's a great idea. But then reality hit: in this way, I would need somehow to juggle two different database versions existing at the same time while I read data from the old DB (ODB), process it and add it to the new DB (NDB). I guess I would need to rename one database and that could blow.

Whereas what my situation is: ODB exists and is running. An update is coming and a NDB must be created based on ODB's data plus some added columns for certain tables. I dump the ODB to file. I erase ODB. I create NDB with no data. I read the ODB's dumpfile, process it (to add some values for the new columns) and insert it into NDB.

So, perhaps I can exploit your idea by creating my version of mysqldump based on DBI which is friendlier to my aims.

Earlier answer: Yep, that's a good way: SHOW COLUMNS and then ask DBI to SELECT and get back all results as Perl data structure. Actually no need for SHOW COLUMNS, just a SELECT will contain a hash with column names/values. After your suggestion I saw DBI, add fields to existing table? which shows how to add columns with ALTER TABLE.

bw, bliako