a string class

Started by
6 comments, last by Dragun 24 years, 7 months ago
I'm fairly sure that the actual character data in classes such as CString and the STL basic_string are dynamic arrays. This allows them to be used in the classic char* string sense while allowing growth. I wouldn't suggest really long arrays because they will never be long enough for someone out there, and will eat a great deal more memory than they need in most cases. Linked lists are messy to do things like printing and char* conversion with.

Why not just use or use as a base class CString (if you don't mind MFC, it's prettier, and it isn't derived from CObject so you don't need most of the Foundation classes so MFC-haters, don't try screaming that old line *grin*) or the STL basic_string (it's fine, it's dynamic, and it's platform independent) ?

-fel

~ The opinions stated by this individual are the opinions of this individual and not the opinions of her company, any organization she might be part of, her parrot, or anyone else. ~
Advertisement
I did a linked list that allowed the string to be dynamic in length. This makes it easy to delete and modify it.
I can't help but think that a linked list implementation is going to be slow, especially considering how often you are going to use the class. With a dynamic array, you can pick a size that is reasonable and be fairly sure that the array won't have to be resized very often, so you only take a speed hit occasionally. A linked list implementation would concatenate faster, but for almost everything else, the array will be faster.
Beg Pardon? The only way a dynamic array implementation would require more memory than a linked list implementation for anything is when reallocating the buffer of the array to change its size.

In a singly-linked list, you are required to store an extra pointer with every item. With strings, assuming a pure linked list implementation, that's a pointer for every character, or a 400% increase in size (1 byte of string requiring 4 bytes of pointer) over an array of the same size?! Using a doubly-linked list (why? singly linked with pointer reversal is good enough for most applications ) will disimprove that to an 800% increase.

Using dynamic arrays, you are limited to the capacity of malloc(), which is governed by the size_t variable. Testing under linux, it worked for values up to 2147483647 before compiling with warnings... which is more space than I think most people have memory (by just a bit )

As for efficiency, it really depends what you're planning on doing. Concatenating is done in constant time using linked lists, but most other operations are done in comparable (or better - substr becomes a memcpy ) time by the array. Sure, there's a slight internal memory loss just after increasing the size of the array, but with an efficient scheme this could be minimized (if it's even a problem) without needing to call realloc too often. (Again, depends on the string operations required)
I would recommend arrays, all the way!

Sorry for going on so long... I've just had this discussion too many times before :P

White Fire

Err - Yup - If you want to write your own I would use malloc(), realloc and free(). I cant see how the linked list would help too much - if anything would be slower as surely you would have to basically call next-> to get each subsequent character? Also would be bigger.
Making your own string class is not a good idea - like fel said, use MFC or STL instead.... but if you do it anyway:

Make sure you implement copy-on-write, or else make sure you pass your strings by reference only.

Also, make your dynamic memory setup "granular" - that is, make it smart enough to allocate 64 bytes at once rather than resizing 64 different times, lest you'll be summon the memory fragmentation demon...

Mason McCuskey
Spin Studios
www.spin-studios.com

Founder, Cuttlefish Industries
The Cuttlefish Engine lets anyone develop great games for iPad, iPhone, Android, WP7, the web, and more!
I'm creating my own string class and wondering what is the most efficient way to approach this. Should I either have a huge character array or a linked list to store the string? The array approach would be faster but takes up more space than it should. The latter would take up the minimal space but it would be slower to manipulate the strings. Any thoughts would be of help...
I realise this topic is long over, but the antagonist in me couldn't leave it alone

A string in c is no more or less than a pointer to char (Probably an array). A string class would then be a wrapper for *char, so you could implement ideas like arbitrary-length strings, fully optimised substring searches, and just about any other string operation you can think of, and provide a simple external interface for these functions.

Necessary? Hardly, but may save time in the long run.

Enjoy,

White Fire

This topic is closed to new replies.

Advertisement