text analysis

[CyberPsychotic <fygrave@freenet.bishkek.su>]

Hello folks,
 I have abit practical question, since I know this list is partly devoted
to cryptography matters, I guess this is the right place to ask:

 So I was playing with some crypto-analysis algorythms, and some steps of
them involve such thing as finding out the frequency of some character or
set of characters. So I wonder what would be the optimal (speed, resources
etc)way of coding this. While playing with single character frequency, I
guess the best way would be having unsigned int 256 elements array (I
refer to C coding) and each time, I find certain character, i just
increase the element of the array with this char offset. This seems very
neat to me (except the thing that some chars could be never found in
text). Anyways, when things come to 2 characters set, i have to get 1024
character set, and so on, which looks quite unreasonable to me to allocate
memory for elements, which probably will be never found in text... I was
thinking of other solution and came to two way connected lists (correct
term?)  things, i.e. : i have some structure like: 

struct element {
char value[ELEMENT_LENGTH];
unsigned int frequency;
struct element *previous;
struct element *next;
}
 and could dinamically allocate memory for each new found element, but
this would slow down whole code by the time list of new elements grow up.
(b/c i will have to look thro.. whole list in order to find whether such
element has already been found or not),  so I wonder maybe there another
neat way to complete tasks like this? I would appreciate any ideas, hints,
code examples. (would be very helpful having a look on some code
performing such a task, I surfed the web but didn't find many)...

 Thanks beforehands,
 Fyodor