AW: AW: [R] constructing specially ordered factor

Khamenia, Valery V.Khamenia at
Tue Oct 5 09:02:44 CEST 2004

> Please follow the posting guide and do your homework before 
> posting, 

1. my last homework in university was done a lot of years ago.

2. I always try to follow posting guide.

>      An object of the same type of 'x'. but if an element is equal to
>      one with a smaller index, it is removed.
> so the order is preserved, by definition.

Here stated when element is removed. There is no explicit statement,
that the order is preserved. If one writes its own implementation
with reodered output, it still matches the docs. Or?

> BTW it uses hashing for `acceleration', not something as slow 
> as sorting.

BTW, do you mean that current hash-based implementation brings *clearly* 
better performance than any O(n*log(n)) sort based algorithm? 
If I have correctly understood src/main/unique.c then current 
hash function is niether minimal perfect hash function nor even 
minimal hash function. In addition, as might be expected, 
current hash function uses full pass through the string to get 
a hash key. 

So, in particular, can anyone clearly show that the current 
hash-based algorithm will be quicker then sort-based 
algorithm if the input has:

1. a lot of strings;
2. strings are very long; 
3. strings are quite unsimilar


Hm, I don't believe you are ready to prove smth. like that. 

P.S. Sorry for broken email-reference. I am bound to Outlook 
Express now.


More information about the R-help mailing list