[r-t] Similar compositions

[Alan Burlison]

> Place notation would save space, but lack of memory isn't a problem here.
> Once a checksum has been generated there is no need to keep the composition.
> I already have them saved in a fairly compressed format anyway.

You could use the PN form instead of a calculated hash which would
then hopefully always be unique, and which would also allow you to
easily recover the row format.

> I think the problem with spotting duplicates is really a problem of deciding
> what a duplicate is.  It seems to me there is a spectrum which roughly goes:
> identical -> trivial rearrangement -> slight difference -> different.  Where
> ever you choose to put the line, someone will argue that it should be more
> one way or the other.  That's why I choose to say that two compositions are
> either identical or different.  If a better argument can be made then I'll
> happily change my mind.

How easy is it to automatically spot trivial rearrangements?

-- 
Alan Burlison
--