Part of the setup routine for the product I'm working on installs a database update utility. The utility checks the current version of the users database and (if necessary) executes a series of SQL statements that upgrade the database to the current version.
Two key features of this routine:
- Once initiated, it runs without user interaction
- SQL operations preserve the integrity of the users data
The goal is to keep the setup/database routine as simple as possible for the end user (the target audience is non-technical). However, I find that in some cases, these two features are at odds. For example, I want to add a unique index to one of my tables - yet it's possible that existing data already breaks this rule. I could:
- Silently choose what's "right" for the user and discard (or archive) data; or
- Ask the user to understand what a unique index is and get them to choose what data goes where
Neither option sounds appealing to me. I could compromise and not create a unique index at all, but that would suck. I wonder what others do in this situation?