On 12/24/15 10:26 AM, Bernardo Sulzbach wrote: > I think you focused too much on the ordering issue. He or she may > just > want a more versatile alter table for convenience. However, I > don't know if alter table is used at all in production anywhere (why > would it be? the column names and ordering should not be part of the > data). They are useful for prototyping because by having a stronger > alter table command you don't need to drop and create so many times. > In the end, if you spent enough time in the design phase to prepare > all your schemas, you should not have to drop (or alter) any of the > tables at all. It is a feature, a nice one, but too far from > necessary (from my standpoint) to be worthy of the developers' time. >
I am in the process of building an application that has the need to be able to add columns to tables in response to user actions. It is primarily in the stage where the user is customizing the program to their needs, but such customizations might happen after the program has been in use for awhile. The main case is to be able to add a 'Flag' to records to allow the filter records or save the set of records found in a search. For now, the current method of create new, drop and rename, isn't unworkable (and mostly hidden in an abstraction layer), but being able to directly add a field would be nice. -- Richard Damon
