问题描述
我对编程不熟悉。我正在寻找C ++中的基数排序实现,我发现了这一点 代码在这里。
void countSort(string a[],int size,size_t k)
{
string *b = NULL; int *c = NULL;
b = new string[size];
c = new int[257];
for (int i = 0; i <257; i++){
c[i] = 0;
}
for (int j = 0; j <size; j++){
c[k < a[j].size() ? (int)(unsigned char)a[j][k] + 1 : 0]++;
//a[j] is a string
}
for (int f = 1; f <257; f++){
c[f] += c[f - 1];
}
for (int r = size - 1; r >= 0; r--){
b[c[k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0] - 1] = a[r];
c[k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0]--;
}
for (int l = 0; l < size; L++){
a[l] = b[l];
}
// avold memory leak
delete[] b;
delete[] c;
}
void radixSort(string b[],int r)
{
size_t max = getMax(b,r);
for (size_t digit = max; digit > 0; digit--){
countSort(b,r,digit - 1);
}
}
所以我的问题是这些行的作用:
c[k < a[j].size() ? (int)(unsigned char)a[j][k] + 1 : 0]++;
b[c[k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0] - 1] = a[r];
c[k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0]--;
那是MSD还是LSD基数排序?
谢谢。
解决方法
这是不必要的紧凑代码的精巧示例,因此很难阅读代码。
对其进行分析有助于将其分开:
// what a mess...
c[k < a[j].size() ? (int)(unsigned char)a[j][k] + 1 : 0]++;
首先取出c
的订阅参数:
// determine index for c
const int iC = k < a[j].size() ? (int)(unsigned char)a[j][k] + 1 : 0;
// post-increment c (as it is it could become a pre-increment as well)
c[iC]++;
索引计算包含一个条件:
// determine index for c
const int iC
// check whether k is (not) exceeding the size of a
= k < a[j].size()
// then
? (int)(unsigned char)a[j][k] + 1
// else
: 0;
数组a
是std::string
的数组,其中std::string
本身包含char
的数组。因此,a[j][k]
产生单个char
。 char
可以是有符号的也可以是无符号的–留给编译器处理。因此,(unsigned char)a[j][k]
不会更改该char
的位,而是将它们解释为无符号数。然后(int)(unsigned char)a[j][k]
将其提升为int
。
请注意,如果当前的编译器已对(int)a[j][k]
进行了签名,则这可能与char
不同,因为在这种情况下,将保留值的可能符号。 (这称为sign extension。)因此,整个过程只负责将当前字符转换为(正)索引并最终加1。
实际上,我打算将其余内容留给读者练习,但是后来我看到了:
b[c[k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0] - 1] = a[r];
像上面一样将其分离,结果为:
const int iC = k < a[r].size() ? (int)(unsigned char)a[r][k] + 1 : 0;
const int iB = c[iC - 1]; // What?
b[iB] = a[r];
考虑到iC
可能导致0(尽管我没有检查整个代码是否完全可以),iC - 1
可能导致-1
。因此,c[-1]
将被访问。
例如,这可能是正确的。 c
指向更大数组的指针,但不在数组的开头。因此,负索引将访问有效存储。这里似乎不是这样:
c = new int[257];
并且我看不到对c
的其他任何分配。
这看起来并不值得信赖。充其量,这种情况过于悲观,从不分配0。
我非常确定我可以证明,如果紧凑的代码无助于更轻松地发现其中可能存在的问题,则可以提高可读性。
那么,非紧凑代码会更慢吗? 根据我的经验,其惊人的优化功能并不适用于现代编译器。
我曾经读过一篇关于优化和Static single assignment form的文章。
同样,当我调试C ++代码时,我不时在Visual Studio调试器监视窗口中看到所有有趣的$$
变量(它绝对不包含任何名为$$
的变量)。
因此,我相信编译器也会在内部做类似的事情。 –明确地这样做以提高可读性应该不会对性能产生最小的影响。
如果真的有疑问,我仍然可以检查汇编器输出。 (例如,Compiler Explorer是一个好地方。)
顺便说一句。 c = new int[257];
?
为什么不int c[257];
?
257个int
值并不使我担心立即超过堆栈大小。
更不用说,数组,尤其是用new
分配的数组,实际上是糟糕的C ++风格,要求使用U.B.。好像尚未发明std::vector…
在我还是学生的时候,我就以某种方式错过了有关Radix排序的课程(尽管我必须承认,我还没有在日常业务中错过这些知识)。 因此,出于好奇,我浏览了Wikipedia,并重新实现了那里的描述。 旨在提供(希望更好)替换问题中发现和公开的OP。
因此,我实现了
- 根据en.wikipedia.org: Radix sort – History上的描述的幼稚方法
- 然后OP展示了我在de.wikipedia.org: Countingsort – Algorithmus上发现的方法(带有计数排序)。
#include <iostream>
#include <sstream>
#include <string>
#include <vector>
/* helper to find max. length in data strings
*/
size_t maxLength(const std::vector<std::string> &data)
{
size_t lenMax = 0;
for (const std::string &value : data) {
if (lenMax < value.size()) lenMax = value.size();
}
return lenMax;
}
/* a naive implementation of radix sort
* like described in https://en.wikipedia.org/wiki/Radix_sort
*/
void radixSort(std::vector<std::string> &data)
{
/* A char has 8 bits - which encode (unsigned) the numbers of [0,255].
* Hence,256 buckets are used for sorting.
*/
std::vector<std::string> buckets[256];
// determine max. length of input data:
const size_t len = maxLength(data);
/* iterate over data for according to max. length
*/
for (size_t i = len; i--;) { // i-- -> check for 0 and post-decrement
// sort data into buckets according to the current "digit":
for (std::string &value : data) {
/* digits after end of string are considered as '\0'
* because 0 is the usual end-marker of C strings
* and the least possible value of an unsigned char.
* This shall ensure that an string goes before a longer
* string with same prefix.
*/
const unsigned char digit = i < value.size() ? value[i] : '\0';
// move current string into the corresponding bucket
buckets[digit].push_back(std::move(value));
}
// store buckets back into data (preserving current order)
data.clear();
for (std::vector<std::string> &bucket : buckets) {
// append bucket to the data
data.insert(data.end(),std::make_move_iterator(bucket.begin()),std::make_move_iterator(bucket.end()));
bucket.clear();
}
}
}
/* counting sort as helper for the not so naive radix sort
*/
void countSort(std::vector<std::string> &data,size_t i)
{
/* There are 256 possible values for an unsigned char
* (which may have a value in [0,255]).
*/
size_t counts[256] = { 0 }; // initialize all counters with 0.
// count how often a certain charater appears at the place i
for (const std::string &value : data) {
/* digits after end of string are considered as '\0'
* because 0 is the usual end-marker of C strings
* and the least possible value of an unsigned char.
* This shall ensure that an string goes before a longer
* string with same prefix.
*/
const unsigned char digit = i < value.size() ? value[i] : '\0';
// count the resp. bucket counter
++counts[digit];
}
// turn counts of digits into offsets in data
size_t total = 0;
for (size_t &count : counts) {
#if 0 // could be compact (and,maybe,confusing):
total = count += total; // as C++ assignment is right-associative
#else // but is the same as:
count += total; // add previous total sum to count
total = count; // remember new total
#endif // 0
}
// an auxiliary buffer to sort the input data into.
std::vector<std::string> buffer(data.size());
/* Move input into aux. buffer
* while using the bucket offsets (the former counts)
* for addressing of new positions.
* This is done backwards intentionally as the offsets
* are decremented from end to begin of partitions.
*/
for (size_t j = data.size(); j--;) { // j-- -> check for 0 and post-decrement
std::string &value = data[j];
// see comment for digit above...
const unsigned char digit = i < value.size() ? value[i] : '\0';
/* decrement offset and use as index
* Arrays (and vectors) in C++ are 0-based.
* Hence,this is adjusted respectively (compared to the source of algorithm).
*/
const size_t k = --counts[digit];
// move input element into auxiliary buffer at the determined offset
buffer[k] = std::move(value);
}
/* That's it.
* Move aux. buffer back into data.
*/
data = std::move(buffer);
}
/* radix sort using count sort internally
*/
void radixCountSort(std::vector<std::string> &data)
{
// determine max. length of input data:
const size_t len = maxLength(data);
/* iterate over data according to max. length
*/
for (size_t i = len; i--;) { // i-- -> check for 0 and post-decrement
countSort(data,i);
}
}
/* output of vector with strings
*/
std::ostream& operator<<(std::ostream &out,const std::vector<std::string> &data)
{
const char *sep = " ";
for (const std::string &value : data) {
out << sep << '"' << value << '"';
sep = ",";
}
return out;
}
/* do a test for certain data
*/
void test(const std::vector<std::string> &data)
{
std::cout << "Data: {" << data << " }\n";
std::vector<std::string> data1 = data;
radixSort(data1);
std::cout << "Radix Sorted: {" << data1 << " }\n";
std::vector<std::string> data2 = data;
radixCountSort(data2);
std::cout << "Radix Count Sorted: {" << data2 << " }\n";
}
/* helper to turn a text into a vector of strings
* (by separating at white spaces)
*/
std::vector<std::string> tokenize(const char *text)
{
std::istringstream in(text);
std::vector<std::string> tokens;
for (std::string token; in >> token;) tokens.push_back(token);
return tokens;
}
/* main program
*/
int main()
{
// do some tests:
test({ "Hi","He","Hello","World","Wide","Web" });
test({ });
test(
tokenize(
"Radix sort dates back as far as 1887 to the work of Herman Hollerith on tabulating machines.\n"
"Radix sorting algorithms came into common use as a way to sort punched cards as early as 1923.\n"
"The first memory-efficient computer algorithm was developed in 1954 at MIT by Harold H. Seward.\n"
"Computerized radix sorts had previously been dismissed as impractical "
"because of the perceived need for variable allocation of buckets of unknown size.\n"
"Seward's innovation was to use a linear scan to determine the required bucket sizes and offsets beforehand,"
"allowing for a single static allocation of auxiliary memory.\n"
"The linear scan is closely related to Seward's other algorithm - counting sort."));
}
输出:
Data: { "Hi","Web" }
Radix Sorted: { "He","Hi","Web","World" }
Radix Count Sorted: { "He","World" }
Data: { }
Radix Sorted: { }
Radix Count Sorted: { }
Data: { "Radix","sort","dates","back","as","far","1887","to","the","work","of","Herman","Hollerith","on","tabulating","machines.","Radix","sorting","algorithms","came","into","common","use","a","way","punched","cards","early","1923.","The","first","memory-efficient","computer","algorithm","was","developed","in","1954","at","MIT","by","Harold","H.","Seward.","Computerized","radix","sorts","had","previously","been","dismissed","impractical","because","perceived","need","for","variable","allocation","buckets","unknown","size.","Seward's","innovation","linear","scan","determine","required","bucket","sizes","and","offsets","beforehand,","allowing","single","static","auxiliary","memory.","is","closely","related","other","-","counting","sort." }
Radix Sorted: { "-","sort.","work" }
Radix Count Sorted: { "-","work" }
请注意,对字符串进行了排序以解释字符的数值。 相反,如果要使用英语词典排序,则必须修改数字到存储桶的映射。因此,字符值的顺序可能会更改,并将相应的大写和小写字符映射到同一存储桶。
经常复制字符串(或其他容器)会占用空间和,而且有些事情,我充其量只能避免生产代码。 move semantics是一种降低CPU压力的方法,同时保持代码的干净度和可比性。 这是我试图(据我所知)示例代码中的内容。