"Brian Johnson" <[EMAIL PROTECTED]> wrote: > I have a few questions about autoincrement and RowID that I couldn't google an > answer > > I have a database for hardware with limited ram (and flash ram) so I'd like to > use as little ram as possible. > > Is there any difference between the hidden field rowid and a field defined as > INTEGER PRIMARY KEY? > > Is it possible to turn off or rename the rowid field? .. I'll answer my own > question here in case someone else googles for this .. according to > http://www.sqlite.org/autoinc.html an INTEGER PRIMARY KEY becomes an alias to > the rowid field and my tests confirm it .. so yes you can rename it this way. > Maybe someone can confirm that the data is not taking up twice the RAM as one > field?
Confirmed. > > Why does a INTEGER PRIMARY KEY field autoincrement when inserting a NULL into > that field as per http://www.sqlite.org/faq.html#q1 and a field defined as int > primary key not work the same way? > Because you might really want to insert a NULL into some other INTEGER field. But inserting a NULL into an INTEGER PRIMARY KEY makes not sense. Also, we can find the largest existing INTEGER PRIMARY KEY in logorithmic time - or constant time if the caching mechanism works. Finding the largest value of an regular INTEGER column is linear time and is thus *much* slower. -- D. Richard Hipp <[EMAIL PROTECTED]>
