The problem is that I can't produce a canonical representation of the entries in my database. Often times some entries are subsets of others, but considered equal. It's possible for an entry to be a subset of two larger entries that aren't equal themselves and still be equal to each of the larger entries.
On 9/28/06, Igor Tandetnik <[EMAIL PROTECTED]> wrote:
Ron Stevens <sqlite-Y9FGH9USQxS1Z/[EMAIL PROTECTED]> wrote: > I have a custom function that compares two text values and returns 1 > if they're equal based on an algorithm that's not strictly text > comparison and 0 if they don't match. I want to do a query that > groups all rows that match using my custom function into a single > group. Is that possible? What you really need is a custom function that produces a "canonical representation" for all the strings in the same equivalence class under your equivalence relation. That is, a function CR(s) such as for every s1 and s2, s1 ~ s2 if and only if CR(s1) = CR(s2). Here '~' is your equivalence relation, '=' is the usual byte-wise comparison. E.g. if '~' is a case-insensitive comparison, then CR(s) could return s converted to all lowercase (or all uppercase). Once you have such a function, you can simply GROUP BY CR(fieldName). Igor Tandetnik ----------------------------------------------------------------------------- To unsubscribe, send email to [EMAIL PROTECTED] -----------------------------------------------------------------------------
