Alain,

Thanks for the comments. I may have to check out the SWAP further, but it 
looks as though it generates very efficient ML, whereas I thought it just 
invoked a runtime FN that used RAM vars. Maybe that has changed since I 
last looked at it.

>> I'm pretty sure that this your trim FNs, which use a RAM counter rather
>> than a register var, will be slower.
>
>No doubt, but in fact, I must report this as a 'bug' in the latest
>beta. The RAM variable is not required with the official Release 5.
>
I think you misunderstood me. I was referring to

WHILE rTrimStr[rTrimStr[0]] = _" "
rTrimStr[0]--
WEND

rTrimStr[0] is in RAM, so reading and decrementing it each loop will be 
significantly slower than decrementing a register var and pluggin the 
result in at the end. I admit I tried it that way myself, because I like 
its simplicity, too.

 0"0
 =J= a  y
  "