On 2/17/06, Ragnar <gnari@xxxxxxx> wrote: > Say again ? > Let us say you have 1 billion rows, where the > column in question contains strings like > baaaaaaaaaaaaaaa....aaa > baaaaaaaaaaaaaaa....aab > baaaaaaaaaaaaaaa....aac > ... > not necessarily in this order on disc of course > > The minimum value would be keyed as 00000001h, > the next one as 00000002h and so on. > > Now insert new value 'aaaaa' > > Not only will you have to update 1 billion records, > but also all the values in your map. > > please explain No comment on the usefulness of the idea overall.. but the solution would be to insert with the colliding value of the existing one lesser than it.. It will falsly claim equal, which you then must fix with a second local sort which should be fast because you only need to sort the duplicates/false dupes. If you insert too much then this obviously becomes completely useless.
