On Fri, Oct 23, 2020 at 2:21 PM David G. Johnston <david.g.johnston@xxxxxxxxx> wrote: > I'm torn here because this would be the first usage of this concept in > PostgreSQL (I think). Yeah, I also have some qualms about this design in the context of Postgres. Particularly because Postgres allows arrays to begin at negative indices. > Tangentially, I noticed that we have a "starts_with" function but no > corresponding "end_with". Ah, interesting. On the other hand, there are both "left" and "right", "lpad" and "rpad", and "ltrim" and "rtrim". And at least ends_with has the fairly elegant alternative of "s LIKE '%suffix'". > It's been a while but there used to be a systemic inertia working against > adding minor useful functions such as these. > > With the new documentation layout I would at least consider updating the > description for the normal functions with an example on how to formulate > an expression that works contra-normally, and in the case where there does > exist such a specialized function, naming it. Supposing you go this route, which of the options would you envision mentioning as the converse of split_part?