C / C++ / MFC

Luc Pattyn2-Aug-10 3:57

Luc Pattyn

2-Aug-10 3:57

char c;
int commas=0;
while((c=*str++)!=0) if (c==',') commas++;

Luc Pattyn [Forum Guidelines] [Why QA sucks] [My Articles] Nil Volentibus Arduum

Please use <PRE> tags for code snippets, they preserve indentation, and improve readability.

image processing toolkits | batch image processing

C^hris L^osinger2-Aug-10 4:14

C^hris L^osinger

2-Aug-10 4:14

const char *p = "a,b,c,de,f,";
int c = 0;
for (;*p;c+=(','==*p++));

Cedric Moonen2-Aug-10 4:33

Cedric Moonen

2-Aug-10 4:33

Ahh, the joys of C++. Did you make it so complex on purpose, or is all your code written like that ? Poke tongue | ;-P

Cédric Moonen
Software developer

Charting control [v3.0]
OpenGL game tutorial in C++

Iain Clarke, Warrior Programmer

Iain Clarke, Warrior Programmer3-Aug-10 2:09

3-Aug-10 2:09

That was classic Kernighan and Richie C!

Sheesh, youngsters of today...

Whippersnapper!

Iain.

I am one of "those foreigners coming over here and stealing our jobs". Yay me!

Nemanja Trifunovic2-Aug-10 4:58

Nemanja Trifunovic

2-Aug-10 4:58

[Untested]:

std::string str("a, b, c, d, e");
int count = std::count(str.begin(), str.end(), ',');

or if you insist on C-style strings:

int count = std::count(str, str + strlen(str), ',');

utf8-cpp

Moak2-Aug-10 5:20

Moak

2-Aug-10 5:20

The inline ASM version is still missing, anyone? Big Grin | :-D

Chat in Europe Java | [Coffee]

Now with 24% more Twitter

Emilio Garavaglia2-Aug-10 20:25

Emilio Garavaglia

2-Aug-10 20:25

I think is interesting to compare the solutions.
Note: The "trivial" algorithm is to scan the string and count the ','.

Now:

Superman: Just did that using CRT: no additional data are added, and the strchr function is made walking from ',' to ','.

Aescleal: Made an elegant use of STL algorithms.
But: a constant literal is copied into a temporary std::string (probably with also a dynamic array allocated by the class itself), creates two temporary iterators, passing them to the std::count algorithm that just do the loop.

Luc Pattyn: Use a more concise loop.

Chris Losinger: same. Very elegant, but more cryptic.

Nemanja Trifunovich: Same a Aescleal, but also propose to use std::cout with const char* instead of std::string::iterator-s, eliminating the need of the conversions.

Moral of the story: I'm starting to believe that STL - and in particular "strings" are over-evaluated.
The more coincise and elegant (together) is probably the second Nemanja proposal.
But one question makes me wonder:
Suppose I'm a coder asked to do this task, and suppose I know the algoritm, but i'm not expert in the language libraries: what should be fastest for me?

- Read the STL documentation -probaly in strict alphabetic order- trying to guess is there can be something that can help me in counting the ',' (I'm luky the <algorithm> header come as first and the count function is one of the firsts ... estimated time: 15 minutes - but supposing I already know the "iterator"/"container"/"algorithm" model)

- Just write

unsigned count_chars(const char* s, char d)
{
    unsigned c=0;
    for(unsigned int i=0; s[i]; ++i)
        if(s[i]==d) ++c;
    return c;
}

actual time 2 minutes! And the best of efficiency (no spurious allocation and copies)

<sarcasm>Do yo understand the beauty of STL ?!?</sarcasm>

(NOTE: using indexes, instead of incrementing pointers, is actually more efficient when translated into todays processor's code)

2 bugs found.
> recompile ...
65534 bugs found.
D'Oh! | :doh:

Moak2-Aug-10 22:26

Moak

2-Aug-10 22:26

An interesting comparison, Emilio!

My two cents. I think someone who has encountered this kind of problem before (data manipulation, string parsing) will need the equal amount of time for either a handcoded C-string solution (e.g. Luc's) or using a library solution (e.g. Aescleal's). Let's say 1 minute. Someone who does it the first time, would probably do the same as when I write something in a new language, you Google for examples and then use the first code snippet you like (probably it's not the best solution but a working one). Let's say 5 minutes.

Cheers,
M

PS: The beauty of STL glows in less trivial examples. The iterator design pattern in particular is a nice idea in my opinion, it gives you a decoupling of data object and algorithm, e.g. you can use the same algorithm with a wide range of objects... and only have to learn it once = time saved.

Chat in Europe Java | [Coffee]

Now with 24% more Twitter

Emilio Garavaglia3-Aug-10 20:17

Emilio Garavaglia

3-Aug-10 20:17

Moak wrote:
The beauty of STL glows in less trivial examples

Oh yes ... but reality is different: it is a matter of fact that vector (sequential indexed), list (sequential) and map (indexed) don't have same interfaces, and that the most of stl algorithm that are "non trivial" works with "sequential indexed" (aka "random access") iterators.

Two siple example:

std::list<int> alist;
std::vector<int> avector;

alist.remove(10); //remove all 10s
avector.erase(std::remove(avector.begin(),avector.end(),10),avector.end());

Until someone doesn't demonstrate me how can I come to write the second example without having the knowledge of at least 20 pages of documentation (yes, I want a 8 y.o. baby coding that!), I will consider "you can use the same algorithm with a wide range of objects... and only have to learn it once = time saved." a pure marketing lye.

(The fake is that vector doesn't have a remove function, hence is not dual to list, and std::remove doesn't remove: it just swaps things around. You cannot know that unless someone tells you)

The pattern itself can be a good idea (iterators as a bridge between collections and algorithms) but the way is implemented is far from being obvious.
That's the point i was considering.

2 bugs found.
> recompile ...
65534 bugs found.
D'Oh! | :doh:

COM interface disappears from ROT [modified]

Moak3-Aug-10 22:22

Moak

3-Aug-10 22:22

Please have a breakfast before making angry morning postings. Smile | :)

Chat in Europe Java | [Coffee]

Now with 24% more Twitter

Phil J Pearson2-Aug-10 1:09

2-Aug-10 1:09

I have two applications, both of which I have written.
AppA was written in C++ using MFC. During startup it registers a CSingleDocTemplate and creates a document. The document provides a COM interface and registers itself in the ROT:

LPUNKNOWN punk = GetInterface(&IID_Welder); // doesn't AddRef
::RegisterActiveObject(punk, CLSID_Interface, ACTIVEOBJECT_WEAK, &m_dwRegister);

AppA revokes the entry in the ROT when it is destroyed:

if (m_dwRegister)
  ::RevokeActiveObject(m_dwRegister, NULL);

The destructor is only called when AppA terminates.

AppB is written in C# and is a client of the interface provided by AppA:

private void Connect()
  {
  welderInterface = (Welder.Interface)Marshal.GetActiveObject("Welder.Interface");
  Marshal.GetIUnknownForObject(welderInterface);	// AddRef's it ...
  }

private void Disconnect()
  {
  if (welderInterface != null)
    {
    Marshal.ReleaseComObject(welderInterface);
    welderInterface = null;
    }
  }

(error handling omitted for clarity)

That all works perfectly the first time AppB runs after AppA has started; AppB can call Connect and Disconnect as often as it wants and use the interface between. However, if AppB terminates and runs again (AppA still running) the interface is no longer available - Marshal.GetActiveObject throws an exception from MK_E_UNAVAILABLE. The only way to make the interface available again is to restart AppA - not an acceptable solution!

Please can someone suggest how I might fix this?

Phil

The opinions expressed in this post are not necessarily those of the author, especially if you find them impolite, inaccurate or inflammatory.

modified on Monday, August 2, 2010 7:16 AM

Code-o-mat2-Aug-10 1:18

Code-o-mat

2-Aug-10 1:18

My guess is that when you do this:

Phil J Pearson wrote:
Marshal.ReleaseComObject(welderInterface);

your object's reference count reaches zero and it gets destroyed. If this is your COM object, try initializing its reference counter to 1 (i guess it is zero now), or AddRef it after creation and Release when your AppA exits.

> The problem with computers is that they do what you tell them to do and not what you want them to do. <
> "It doesn't work, fix it" does not qualify as a bug report. <
> Amazing what new features none of the programmers working on the project ever heard of you can learn about when reading what the marketing guys wrote about it. <

Phil J Pearson2-Aug-10 1:51

2-Aug-10 1:51

No, that's not it. as I said:
1. The destructor is only called when AppA exits (checked in the debugger).
2. AppB can call Connect and Disconnect many times without problem. The problem only occurs when AppB exits. There's nothing special done with the interface on exit.

Phil

The opinions expressed in this post are not necessarily those of the author, especially if you find them impolite, inaccurate or inflammatory.

Code-o-mat2-Aug-10 1:59

Code-o-mat

2-Aug-10 1:59

How did you check that your interface isn't deleted when AppB exits? I mean,

Phil J Pearson wrote:
(checked in the debugger).

explain this, please.

Phil J Pearson2-Aug-10 2:07

2-Aug-10 2:07

Logging code and breakpoints in the doc destructor and in the doc OnFinalRelease. Only called when AppA exits - not AppB.

Phil

The opinions expressed in this post are not necessarily those of the author, especially if you find them impolite, inaccurate or inflammatory.

Phil J Pearson2-Aug-10 2:19

2-Aug-10 2:19

Oops! Sorry! I take back my earlier rejection of your reply.
It seems that the interface was being destroyed even though the document that implements it was not.
I AddRef'd the interface in the doc constructor and Release'd it again in the destructor as you suggested and the problem is solved.

Thank you!

Phil

The opinions expressed in this post are not necessarily those of the author, especially if you find them impolite, inaccurate or inflammatory.

Code-o-mat2-Aug-10 2:24

Code-o-mat

2-Aug-10 2:24

It happens...great it works now. Thumbs Up | :thumbsup:

ClistCtrl

alexander 19831-Aug-10 23:59

alexander 1983

1-Aug-10 23:59

i have a question i hava a ClistCtrl control in a modal dialog box (report type). Anyone knows how to make ClsiCtrl keep the values entered after closing the dialog box?when reopening the dialog the control to display the data entered before. Thank you.

«_Superman_»2-Aug-10 0:14

«_Superman_»

2-Aug-10 0:14

This is already done for you and depends on the life time of the dialog object.
Also update the list control in a callable method other than OnInitDialog.

So each time you call DoModal, it will be on the same object of the dialog class.

«_Superman_»

I love work. It gives me something to do between weekends.

Microsoft MVP (Visual C++)

Polymorphism in C

David Crow2-Aug-10 2:49

David Crow

2-Aug-10 2:49

alexander 1983 wrote:
Anyone knows how to make ClsiCtrl keep the values entered after closing the dialog box?when reopening the dialog the control to display the data entered before.

Keep the values in some sort of container (e.g., list, array).

"One man's wage rise is another man's price increase." - Harold Wilson

"Fireproof doesn't mean the fire will never come. It means when the fire comes that you will be able to withstand it." - Michael Simmons

"Man who follows car will be exhausted." - Confucius

bleedingfingers2-Aug-10 3:18

bleedingfingers

2-Aug-10 3:18

What others said is right. You have to store the data - the data that's supposed to be shown - elsewhere and in order to show that data in the control, you have to feed that, may be through the dialog object, may be something like this.

CMyDialog : public CDialog
{
public:
CMyDialog(CListCtrlData *pd):m_pd(pd)...
BOOL OnInitDialog(...)
{
//...
// fill the list ctrl with the data
//...
}
private:
CListCtrlData *m_pd;
};

CListCtrlData gDataA;

void SomeFunction()
{
CMyDialog dlg(&gDataA);
dlg.DoModal();
}

CListCtrlData gDataB;

void SomeOtherFunction()
{
CMyDialog dlg(&gDataB);
dlg.DoModal();
}

If the data is supposed to be editable, just make sure that your data class has appropriate methods to update and call these methods from the relevant event handlers for the control.

byte till it megahurtz