C / C++ / MFC

CPallini wrote:

Why?

switch (c)<br />
{<br />
   case '\n':<br />
   case '\r':<br />
   case '\t':<br />
   case ' ':<br />
..<br />
   break;<br />
...<br />
}

cp9876 wrote:

If it is not what you need and you have to test for multiple characters, it is much more efficient to do:

switch (c)
{
case '\n':
case '\r':
case '\t':
case ' ':
..
break;
...
}

		if (myString[i] == '\n' || myString[i] == '\r' || myString[i] == '\t')
00401040  movzx       ecx,word ptr string L"hello world" (4020F4h)[edx*2] 
00401048  cmp         cx,0Ah 
0040104C  je          test003+4Ah (40105Ah) 
0040104E  cmp         cx,0Dh 
00401052  je          test003+4Ah (40105Ah) 
00401054  cmp         cx,9 
00401058  jne         test003+4Dh (40105Dh) 
		{
			count++;
0040105A  add         esi,1 
		}
		switch(myString[i])
0040105D  movzx       ecx,cx 
00401060  cmp         ecx,9 
00401063  jl          test003+62h (401072h) 
00401065  cmp         ecx,0Ah 
00401068  jle         test003+5Fh (40106Fh) 
0040106A  cmp         ecx,0Dh 
0040106D  jne         test003+62h (401072h) 
		{
		case '\n':
		case '\r':
		case '\t':
			count++;
0040106F  add         esi,1 

cp9876 wrote:

You are right, for a few cases the difference may be negligible.

cp9876 wrote:

My understanding is that the switch statement can be optimized to O(log(N)) operations (using a tree type of search), whereas the simple cascaded if's will be O(N) operations. Obviously these could be the same but the compiler knows it can optimize a switch statement as the cases are all integer constants. I'd be interested to know the results if anyone can be bothered to do some test cases and analyse the code. It may be that the way the case statements are tested using less than operations is indicative that it would use a tree search for more cases (instead of the simple comparisons in the if construct).

CPallini wrote:

In fact there is no difference at all, in the code snippet shown

Eranga Thennakoon wrote:

Is there any special advantage use of switch-case combination than if() condition.

// MyClass.h
enum STATUS { St_A, St_B, St_C };
// ...
MyClass::STATUS MyClass::GetStatus();
// ...

// MyClass.cpp
switch(GetStatus())
{
case MyClass::St_A:
  // ...
case MyClass::St_B:
  // ...
default:
  // Error.
}

Eranga Thennakoon wrote:

if(strString.at(i) == ' ')
{
iSpace ++;
}
if(strString.at(i) == '\t')
{
iTabs ++;
}
if(strString.at(i) == '\r')
{
iRet ++;
}

int nCount = std::count(strString.begin(), strString.end(), ' ');
nCount += std::count(strString.begin(), strString.end(), '\r');
nCount += std::count(strString.begin(), strString.end(), '\t');

jhwurmbach wrote:

But that definitely is the slowest alternative...

#include <fstream><br />
#include <string><br />
#include<iostream><br />
using namespace std;<br />
<br />
int main()<br />
{<br />
	string str;<br />
<br />
	ifstream myFile("ap1.txt");<br />
	if(! myFile)<br />
	{<br />
		cout<<"Error!"<<endl;<br />
		return -1;<br />
	}<br />
	while(! myFile.eof())<br />
	{<br />
		<br />
			getline(myFile, str);<br />
			str;	<br />
			cout<<str<<endl;<br />
	<br />
	}<br />
	myFile.close();<br />
	return 0;<br />
}

Read a Line;
Does it start with Value = ?
Yes...
    Does it have Integer32 next?
    Yes...
        Grab the -44 part, and keep it somehow. A list, email it to your best friend, whatever you want.
No...
    Move on to the next line!