On 28 Jun 2010, at 7:31pm, Pavel Ivanov wrote: >> Such freedom is not suitable for data interchange between two systems. Not >> that SQLite or any other database would change the PK during import-export, >> but they are free to do so as long as the *intramural* integrity is >> preserved. > > Can you point out some documentation supporting this claim? > I've no time now to search internet on this matter but I believe DBMS > changing data that *I stored* in it is fundamentally broken. I > wouldn't advise anybody to use it. DBMS is allowed to change only > internal details which do not appear in CREATE TABLE and INSERT > statements (like ROWID for example).
This /is/ how SQLite behaves. If you allow it to do its own thing with _rowid_, and never refer to it, it feels free to change _rowid_ values any time it likes. If you define a column and it says "Hey, I can save space because I already have _rowid_ which fits those requirements." then since it knows you can access the values in that column it won't ever change them. Makes perfect sense to me. By the way, these cases where SQLite may change _rowid_ values are not mundane. I think the only place I've seen it documented is for VACUUM or for autovacuums. Simon. _______________________________________________ sqlite-users mailing list sqlite-users@sqlite.org http://sqlite.org:8080/cgi-bin/mailman/listinfo/sqlite-users
