Re: New to D

[Era Scarecrow via Digitalmars-d-learn]

On Thursday, 27 October 2016 at 13:43:26 UTC, Steven 
Schveighoffer wrote:
It depends on the size of the file and the expectation of 
duplicate words. I'm assuming the number of words is limited, 
so you are going to allocate far less data by duping on demand. 
In addition, you may incur penalties for accessing the string 
directly from the file -- the OS may have swapped out that page 
and have to re-read it from the file itself.

You could also read the entire file into a string and go based 
on that.

 Depends. I recall experimenting early on with Memory mapped 
files (80Mb-300Mb), and it instantly loaded, no time at all. I 
don't think it even read the file or portions until I made 
requests to it (Course in my instance I'd have to create virtual 
records to access everything, still need to re-write and finish 
that project).

 It really depends on the circumstances though, and I suppose 
also understanding when a buffer is shared/reused and when to use 
dup.