I thought I'd follow up on the linked list to permutation
vector problem a bit:

The original problem arose from a client requirement. An
operating system data structure is a linked list of buffers, arranged
in LRU fashion. Hence, the N most recently used buffers are the
first N entries on the chain. Buffers are associated with files
though another mechanism, that gives a list of buffer indices for
each file. What I wanted to do was to provide a list of the N
most popular files being accessed at the moment. The idea was to
permute the linked list into a list sorted into LRU order,
then do some selection within other class information.

Nobody came up with a Gee-Whiz Swell algorithm using scan or
other linear(ish) time function.

So, I decided to brute force it on my PC to see how it would run
if compiled. Here's the APL 2000 version:

r{<-}({rho}links){rho}{neg}1
:for i :in {iota}{rho}links
r[hd]{<-}i
hd{<-}links[hd]
:endfor

On a 100k element vector, this takes 12.23 seconds on my
Belchfire 3000  486DX4-100 under the interpreter.

When compiled with APEX, it takes 0.14 seconds, or a factor of
about 87 times faster. And, you can read the code!

Bob