on 08/10/2000 00:01, Jay Reeve at jktr@... wrote:

>> BTW, I am not agree with your estimate that taking in account of CASE will
>> slow down 2 times. Look, we get list of words:
>> 
>> "aaa" "aAAb" "aAaa" "Aaaa" "acccf" "AAAaacc"
>> 
>> it can be very long, and we need again to sort it using another 1MB of MAP.
>> So requirements in RAM grow in 2 times,
>> time of work .... well, it really grow linear ...

> My estimate of twice-as-long came from searching twice as much RAM if we
> used longs instead of short integers. My suggestion was to assign "aAaa",
> "Aaaa", and "AAAaacc" to the same list. There would be a small time
> penalty for sorting that list, but it would save using (and searching)
> 35,271,500 bytes of memory to map upper and lower cases separately.

Yes you are right, that we can put this A a
into one list, i.e. into one hash group
but how you will sort that hash group then ?

by qsort() ?   
and again get into [n * log(n)] dependency.
although yes, for total time will be less.


You assume that list will be very short,
I can not assume the same for database with million of records, for example
I can easy get a list with several thousands of the same word.



-- 
Best regards,
Ruslan Zasukhin

-------------------------
Paradigma.

e-mail: ruslan@...
web   : http://www.paradigmasoft.com

To subscribe to the Valentina mail list
send a letter to valentina-on@...