Dennis, I think I will go the normalized way, since the objects we are talking about are in fact sub-objects, and I can keep a list of them.
Do you (or anyone else) have any experience with those large DBs in terms of SELECT execution times? ms, seconds or minutes? Since I would mainly use field numbers in the WHERE clause I´d create an index on them I guess? The client first wanted to store that data in an XML file, but I was able to convince him to use a DB. The Client is not allways right, but he´s allways the client ;) Thanks again, André -----Original Message----- From: Dennis Cote [mailto:[EMAIL PROTECTED] Sent: Monday, April 10, 2006 10:23 PM To: sqlite-users@sqlite.org Subject: Re: [sqlite] How to handle large amount of data? André Goliath wrote: >Denis, >I didn´t knew references where possible in Sqlite 3.2, >thanks for pointing that out! > > The reference keyword in the column declaration doesn't really do anything in SQLite. I use it more as a comment for human readers so they know which column to use when joining tables. >Do you have more specific information on how that is implemented in SQLite? >Would you think the amount of memory needed would be acceptable for an >field_data with >1.000.000 objects * 21 fields * 1.000 chars ? >It is quite unlikely to reach that DB size, >but I have to expect the unexpected ;) > > > Well, that would be a large table, about 21 GB plus overhead, say 25-30 GB. But if that's the amount of data you have to store, you will have to store it somewhere. It doesn't take anymore more space to store the data in a normalized table. As Jay said, the table is never loaded into memory, so memory should not be an issue. HTH Dennis Cote