On 8/1/2023 11:32 AM, Philip Kaludercic wrote:
> In that case we should certainly apply a patch along the lines of what
> you suggested, but I would prefer the mapconcat approach, if anything
> because of the micro-optimisation of avoiding the overhead for the
> generic seq call and having the create and apply closure over and over
> again.  Not sure if this counts as a strong preference, but it certainly
> is a preference.

How about this? For extra micro-optimization, I put the 'mapconcat' call 
outside of the loop. Now we won't concat the same list of strings for 
every file/dir that we look at.