C/C ++ string processing (4): STD: vector and STD: stringbuilder

C/C ++ string processing (4): STD: vector and STD: stringbuilder

Xu Shiwei
2008-3-28

Introduction

STD: stringbuilder is implemented based on STD: vector. Therefore, although STD: vector is discussed in this article, all conclusions are equally effective for STD: stringbuilder.

Implementation Overview

In short, STD: vector is a dynamic array that manages a linear, dynamically increasing memory.

How to accelerate STD: vector? Use Vector: Reserve

Before we estimate the size of a vector, we should call reserve (size) to pre-allocate the memory (size here is the estimated number of vector elements) before inserting data ).

Avoid inserting/deleting data from a vector (BEGIN ).

That is to say, we should try to use vector: insert (end (),...) Or vector: push_back/pop_back to add/delete data. Instead of using vector: insert (begin (),...) Operation. The push_front operation is not provided for the vector. There is only one reason: inserting/deleting data from the vector is inefficient.

If you need to insert a large amount of data at the beginning of the container (BEGIN), you can select STD: deque (which has the advantage of random vector access and efficient list insertion) or STD: list.

When can't STD: vector be used for STD: vector?

When the container needs to hold massive data and the number of elements is unpredictable, STD: vector cannot be used. All linear memory-based data structures (such as STD: vector, STD: string) encounter performance bottlenecks when handling massive data volumes.

Memory fragmentation

A typical problem with linear memory-based data structures (such as STD: vector, STD: string) Is that memory fragments are easily generated. After a large number of operations on STD: vector or STD: String, memory fragments will be severe.

STD: vector and allocator

We know that the prototype of STD: vector is:

template <class DataT, class AllocT = std::allocator<DataT> >
class vector; 

So do we need to change alloct to GC Allocator like MAP/multimap, set/Multiset, list/slist, hashmap/hashmultimap, hashset/hashmultiset, and deque?

The answer is: no. GC allocator is helpful for improving small memory allocation. However, the data structure in the dynamic linear memory is invalid. In addition to STD: vector, a typical data structure is STD: string (STD: basic_string ).

Recommended reading

• STD: deque-a Data Structure between STD: vector and STD: List, which can be randomly located, the performance of massive data is still very stable (in fact, its push_back/push_front performance is almost constant: O (1), unlike STD: vector, with the increase of elements, insert speed decreases sharply ).
• STD: List-two-way linked list.