Re: [nimh-dev] More stuff

[Joel Uckelman <uckelman@xxxxxxxxx>]

Thus spake "Jon Stewart":
> > MH treats sequences like sets. E.g. 'scan 2 1 1' will show you the scan
> > for 1, followed by the scan for 2. I like it that way. I wish it were
> > also a way to apply set-theoretic operations other than union and negation,
> > though. Say I want to list all messages in foo and bar, e.g., I should be
> > able to 'scan foo&bar'.
> 
> 
> Hmm. Then I guess we will need to calculate the sequence upfront. In that 
> case, it's probably best to have a sorted list of range tuples, [beg, 
> end). Constructing the list isn't too hard. You throw the first specified 
> range in the list and, if there are any others, you go through the list 
> for each one (it's a quadratic algorithm) and either insert the range in 
> its rightful place -- if it's disjoint with existing ranges -- or expand 
> the existing ranges if there are overlaps.
> 
> CPU and memory usage are constant in the common case of only one specified 
> range. If someone is dumb enough to specify such a complicated listing 
> that the quadratic nature gets to be inhibiting, well, that's a good 
> thing. Software should stand up for itself when abused.

A quadratic algorithm isn't bad when n is small. I can't remember the last
time that I gave a sequence containing even two overlaping ranges.

_______________________________________________
nimh-dev mailing list
nimh-dev@xxxxxxxxxxx
http://lists.ellipsis.cx/mailman/listinfo/nimh-dev