Re: Ada Shootout program for K-Nucleotide (patches)

[Georg Bauhaus <rm.dash-bauhaus@futureapps.de>]

Ole-Hjalmar Kristensen schrieb:

> astra06:~/ada/div-ada> uname -a
> SunOS astra06 5.10 Generic_137137-09 sun4u sparc SUNW,Sun-Fire-V440
> astra06:~/ada/div-ada> ls -l x.txt
> -rw-------   1 ok145024 clustra  60000000 Sep  1 12:03 x.txt
> astra06:~/ada/div-ada> time ./get_line_test < x.txt
> 0.49u 0.09s 0:00.59 98.3%
> 


Comparing the two programs, they are on a par, almost
exactly.

However, yours is IMHO less messy. It also has

        return Copy & Get_Line;

which, I guess, solves the problem of long lines, even
at the expense of copying.

Can we reduce the copying overhead?
The copying would be have O(f(Item'Length)) with
f(n) ~= s*(s+1)/2 where s = n/BUFSIZ, so it is in O(n**2).
We could, I think, use a storage pool:

- whenever copying is needed, allocate another buffer
  right next to the current buffer. Use the newly allocated
  buffer for the recursive call.

- on reaching of line/file, reinterpret the sequence of
  buffers in the storage pool as a String. Finally, Free.

Can this be done?

(This won't make a Shootout program any smaller, but
seems good for the general case...)