Raul Miller <[email protected]> wrote: > Anyways, that looks interesting (and similar to the paper William > Tanksley, Jr. referred to - though the paper you refer to here was > published 3 years earlier).
That is indeed interesting and similar. The difference is also interesting. The baseline BWT sorts by the entire string looped on itself. The Schindler variant uses a constant (and data-independent) k-length sort. The Lyndon variant (the one I posted) uses a sort that loops around AND whose length depends on the contents of the string. I suspect that Schindler and Lyndon could be combined, though, to create a sort which needed less information in the common case and ran a bit faster. I don't know, I'd have to dig... But I'm interested. > Raul -Wm ---------------------------------------------------------------------- For information about J forums see http://www.jsoftware.com/forums.htm
