C / C++ / MFC

Re: <algorithm> sort keeps generating error messages.

26-Mar-21 2:40

To complement what Greg was saying:
- case 3 fails because comp is a structure not a function returning a bool. Try changing it to:

C++

bool comp (const RDA_Names::Asset& lhs, const RDA_Names::Asset& rhs)
{
  return lhs.asset_No < rhs.asset_No;
}

You will have to also make Asset structure public.

As a case 4 you can try this:

C++

sort (asset_Data.begin (), asset_Data.end (), [](auto& l, auto& r)->bool {
  return l.asset_No < r.asset_No; });

where the comparison function is an anonymous lambda.

The next step is to look at how efficient your code will be. If you are sorting only 500 objects and you don't do it too often, it's probably OK. In general however your solution is highly inefficient. Every time sort has to exchange two elements it makes copies of the asset_Name string.

There are various ways to improve it:
- keep a unique_ptr pointer to string and provide a move constructor for the Asset structure.
- create an array of indexes in RDA_Names and sort it inseted of sorting the asset_Data array.

Mircea

modified 26-Mar-21 9:14am.

aschwarz26-Mar-21 10:27

aschwarz

26-Mar-21 10:27

Your first point is right on, and gets me back to Case 2. You're second point is incorrect. What is being sorted is an object of some type. The sort routine ignores the type and whether the object containing the type is private or public. It only works on comparison, I assume signed char comparison if there is no comparison function, and the ability to index and swap. The comparison function has no cognizance of privacy, it is only passed pointers/objects and returns a bool.

At the sort base level, the input object need not be public, and if the input object contains objects, these contained objects need not be public.

Just guessing.
thanks

Re: <algorithm> sort keeps generating error messages.

Mircea Neacsu26-Mar-21 12:22

26-Mar-21 12:22

Quote:
Just guessing.

Well, you didn't guess right Smile | :)

Pretty much all you said is valid for the old C qsort function. std::sort is an entirely different beast. I suggest you carefully read the description at std::sort - cppreference.com[^]

Quote:
The sort routine ignores the type and whether the object containing the type is private or public.

I'm not sure I understand what you mean by "the object containing the type". In C++ objects don't contain types. Besides the sort routine is not ignoring the object type at all. According to C++ reference: "The type of dereferenced RandomIt must meet the requirements of MoveAssignable and MoveConstructible". That requires at least a copy constructor for the object (either user defined or generated by the compiler).

Quote:
The comparison function has no cognizance of privacy, it is only passed pointers/objects and returns a bool.

Of course not, the comparison function needs to access the members of the object in order to compare them. Either it is a member function or a friend function or the members in question are public.

The example on the C++ reference page mentioned above shows all the different methods you can use to invoke the std::sort algorithm.

Mircea

Win32 Structures

Richard Andrew x6414-Mar-21 9:52

Richard Andrew x64

14-Mar-21 9:52

My knowledge of the low level implementation of C++ structures and classes is limited.

But I want to ask someone who might know, is it theoretically possible to derive a class from a Win32 structure and add methods and still be able to pass the new "structure" to a Win32 API?

I do know that adding class data members would be out of the question, because that would change the size of the structure.

But is adding methods to a structure something I could do and still use it in the windows api?

The difficult we do right away...
...the impossible takes slightly longer.

Mircea Neacsu14-Mar-21 10:22

14-Mar-21 10:22

Yes, it is very much possible. You can write something like this:

C++

class MyPen : public LOGPEN {
  public:
    MyPen (/*args*/);
    void use_pen ();
  private:
    int stuff;
}

MyPen the_pen;
HPEN hp = CreatePenIndirect ((LOGPEN*)&the_pen);

Note that you can also extend your object with additional data members. They will be placed after the base class (Windows structure) and they will do no harm.

Mircea

Richard Andrew x6414-Mar-21 10:32

Richard Andrew x64

14-Mar-21 10:32

Thanks for your reply. I'm excited to try this.

Your example is great. But if I wanted to use an array of MyPen's then that would surely cause problems because of the larger size, no?

The difficult we do right away...
...the impossible takes slightly longer.

Mircea Neacsu14-Mar-21 10:36

14-Mar-21 10:36

Correct but only if the Windows API call takes an array of (the smaller) structures as parameters. Of top of my head, I cannot think of one that does.

Mircea

11917640 Member 17-Mar-21 3:01

11917640 Member

17-Mar-21 3:01

This is the way some MFC classes are implemented:

C++

typedef struct tagPOINT
{
    LONG  x;
    LONG  y;
} POINT, *PPOINT, NEAR *NPPOINT, FAR *LPPOINT;

class CPoint :
	public tagPOINT
{
    ...
};

No additional members and virtual functions to get binary compatibility with underlying structure.

Richard Andrew x6417-Mar-21 9:20

Richard Andrew x64

17-Mar-21 9:20

Hi, thanks very much for your post.

Do I understand correctly that non-virtual functions would ruin binary compatibility?

The difficult we do right away...
...the impossible takes slightly longer.

Zoom image on mouse position

11917640 Member 17-Mar-21 19:19

11917640 Member

17-Mar-21 19:19

Non-virtual functions don't change a class instance size. You can use this technique, as in existing Microsoft classes like CPoint, CRect, adding instance non-virtual functions and static functions, if necessary.

Gopi Nath12-Mar-21 0:20

Gopi Nath

12-Mar-21 0:20

Hello,

I am displaying an image in a static control (in a dialog) and trying to stretch according to the zoom factor on mouse position. Zoom on mouse position is working fine, but the image is not fitting inside the static control, going all over the dialog.

Below is the code.

C++

// source image
	Gdiplus::Bitmap *gpBitmap = new Gdiplus::Bitmap(mImagePath);

	int origW = gpBitmap->GetWidth(); // 5298 - original width
	int origH = gpBitmap->GetHeight(); // 2728 - original height

	// destination rectangle
	RECT rcVPC;
	GetClientRect(hWnd, &rcVPC); // 501 X 255 - size of the static control
	HDC hDc = GetDC(hWnd);

	float perW = (float)rcVPC.right / (float)origW;
	float perH = (float)rcVPC.bottom / (float)origH;
		
	int newW = origW * perW * mVPZoomFactor; // starting factor is 1.0 and multiplying with 1.2 with each zoom
	int newH = origH * perH * mVPZoomFactor;

	Gdiplus::Bitmap *bmpResized = new Gdiplus::Bitmap(newW, newH, gpBitmap->GetPixelFormat());
	Gdiplus::Graphics *graphics = new Gdiplus::Graphics(bmpResized);
	graphics->DrawImage(gpBitmap, 0, 0, newW, newH);
	
	int zoomWidth = (int)(0.5 + newW * mVPZoomFactor);
	int zoomHeight = (int)(0.5 + newH * mVPZoomFactor);

	//Find the position "factor"
	double dXFactor = (double)rcVPC.right / stX; // stX, stY - mouse position on zoom
	double dYFactor = (double)rcVPC.bottom / stY;

	//Find the origin
	int left = stX - zoomWidth / dXFactor;
	int right = stY - zoomHeight / dYFactor;

	HBRUSH   hB = GetSysColorBrush(COLOR_3DDKSHADOW);

	FillRect(hDc, &rcVPC, hB);

	HDC hdcMem = CreateCompatibleDC(hDc);

	bmpResized->GetHBITMAP(0, &hBmpVPCanvas);

	HBITMAP oldBmp = (HBITMAP)SelectObject(hdcMem, hBmpVPCanvas);

	//use StretchBlt
	StretchBlt(hDc, left, right, zoomWidth, zoomHeight, hdcMem, 0, 0, newW, newH, SRCCOPY);

	SelectObject(hdcMem, oldBmp);

	DeleteDC(hdcMem);

	delete gpBitmap;

Can anyone help me, what i am doing wrong?

Regards,
Gopinath.

Re: Zoom image on mouse position

Richard MacCutchan12-Mar-21 1:28

Re: Zoom image on mouse position

12-Mar-21 1:28

Gopi Nath wrote:
what i am doing wrong?

Probably mixing GDI and GDI+; you should only use one of these, preferably GDI+. Zooming an image is quite simple in GDI+ as you just need to specify the target rectangle's position and size.

Gopi Nath14-Mar-21 20:34

Gopi Nath

14-Mar-21 20:34

Thank you Richard, tried with GDI+ only with some changes, now, i am able to get the required output.

Regards,
Gopinath.

Select TRACE functions based on both a variable and a #define?

arnold_w11-Mar-21 5:31

Re: Select TRACE functions based on both a variable and a #define?

11-Mar-21 5:31

I'm working on an embedded ARM-project (STM32F) and I'm programming in C programming language and compiling with Eclipse/GCC. In the project there is TRACE functionality implemented that, if enabled, logs various things over UART using queues and DMA:

C++

#define TRACE(grp, lvl, ...) traceToPrintStream(grp, __FILE__, __FUNCTION__, __LINE__, lvl, __VA_ARGS__)

It can be used, e.g. like this:

C++

TRACE("regCallbacks", T_D, "The user pressed the button");
TRACE("regCallbacks", T_D, "Write to reg %d. First written byte: 0x%X", regAddr, regValues[0]);

I need a different trace that is not queued, does not use DMA and does not include the file name, function name and line number, but the other persons in this project do not agree to throw out the existing TRACE, so I need to make something in parallel. I want to be able to enable my TRACE in two ways, either by #defines:

C++

#define ENABLE_USER_EVENTS_LOGGING
#define ENABLE_COMMUNICATOIN_BUS_LOGGING

or by setting a varible during runtime:

C++

Bool_t enableCommunicationBusLogging;
Bool_t enableUserEventsLogging;

I need a different grouping (the grp parameter) than the others, so the example TRACEings above would have to be augmented with my grouping as well:

C++

TRACE_EX("regCallbacks", USER_EVENTS, T_D, "The user pressed the button");
TRACE_EX("regCallbacks", COMMUNICATION_BUS, T_D, "Write to reg %d. First written byte: 0x%X", regAddr, regValues[0]);

The rules I would like to apply are:
1. If my TRACE is enabled using the macros (hardcode enabled), then the other TRACE can not be used (hardcode disabled).
2. If my TRACE is not enabled using the macros, then both TRACEs can be enabled/disabled during runtime independently of each other. However, they are enabled per groups!

Does anybody know how I can solve this? I found a great website (C Preprocessor tricks, tips, and idioms · pfultz2/Cloak Wiki · GitHub[^]) that shows some really neat macro-tricks, but I'm not sure how to apply it my case.

Richard MacCutchan11-Mar-21 6:11

Re: Select TRACE functions based on both a variable and a #define?

11-Mar-21 6:11

As I understand it (assuming the traceToPrintStream function is fixed) you need a slightly different macro for your TRACE_EX, something like:

C++

#define TRACE_EX(grp, lvl, ...) traceToPrintStream(grp, "", "", 0, lvl, __VA_ARGS__)

If you want to enable the traces at runtime then you need them always enabled at compile time, otherwise you will need to rebuild the code every time. So to allow the traces to be based on switches you need something like:

C++

#define TRACE_EX(switch, grp, lvl, ...) \
if (switch) \
{ \
    traceToPrintStream(grp, "", "", 0, lvl, __VA_ARGS) \
}

where switch is the name of the boolean that controls the tracing.

arnold_w11-Mar-21 7:07

Re: Select TRACE functions based on both a variable and a #define?

11-Mar-21 7:07

Ok, so this is what I have now:

C++

#define USER_EVENTS_LOGGING          0x00000001
#define COMMUNICATOIN_BUS_LOGGING    0x00000002
uint32_t enabledMyGrpBitFlags = 0;

#if defined(ENABLE_USER_EVENTS_LOGGING)
#define TRACE_EX(grp, myGrp, lvl, ...)   if (myGrp & USER_EVENTS_LOGGING)       printf_(__VA_ARGS__)
#elif defined(ENABLE_COMMUNICATOIN_BUS_LOGGING)
#define TRACE_EX(grp, myGrp, lvl, ...)   if (myGrp & COMMUNICATOIN_BUS_LOGGING) printf_(__VA_ARGS__)
#else
#define TRACE_EX(grp, myGrp, lvl, ...)                                                 \
	do {                                                                               \
		if (myGrp & enabledMyGrpBitFlags) {                                            \
			printf_(__VA_ARGS__);                                                      \
		}                                                                              \
		traceToPrintStream(grp, __FILE__, __FUNCTION__, __LINE__, lvl, __VA_ARGS__);   \
    } while (0)
#endif

Is it possible to remove the run-time checks:

C++

if (myGrp & USER_EVENTS_LOGGING) 
if (myGrp & COMMUNICATOIN_BUS_LOGGING)

so that the TRACE_EX macro would become empty when the myGrop isn't matching? In other words, let's say only ENABLE_USER_EVENTS_LOGGING (not ENABLE_COMMUNICATOIN_BUS_LOGGING) is defined, then I would like the following line to "disappear" after the preprocessor is finished:

C++

TRACE_EX("regCallbacks", COMMUNICATION_BUS, T_D, "Write to reg %d. First written byte: 0x%X", regAddr, regValues[0]);

Is that possible somehow?

Richard MacCutchan11-Mar-21 23:11

Re: Select TRACE functions based on both a variable and a #define?

11-Mar-21 23:11

The run time checks are only effective at the time the code runs, so there is no way that the macro can be changed at that point. And the macro itself is evaluated, and its code generated in the preprocessor, long before the code is running.

arnold_w12-Mar-21 0:54

Re: Select TRACE functions based on both a variable and a #define?

12-Mar-21 0:54

I want to be able to compile in 2 different modes:

1. Normal mode: The 2 types of TRACE can be enabled/disabled during runtime independently of each other.
2. Hardcoded my TRACE only mode: Only my TRACE is enabled (1 or more myGrp TRACE groups), it's impossible to enable/disable anything TRACE related during runtime.

So, when compiling in mode 2 above, it should be possible for the macro to evaluate the passed myGrp parameter to determine how the TRACE_EX should be expanded. So if we only enable

C++

#define ENABLE_USER_EVENTS_LOGGING

then the preprocessor should be able to determine that this line should be removed:

C++

TRACE_EX("regCallbacks", COMMUNICATION_BUS, T_D, "Write to reg %d. First written byte: 0x%X", regAddr, regValues[0]);

Richard MacCutchan12-Mar-21 1:22

Re: Select TRACE functions based on both a variable and a #define?

12-Mar-21 1:22

I have tried what I think you are asking for but the preprocessor rejects it. As far as I know you can only do it the way I already suggested.

arnold_w12-Mar-21 1:51

Re: Select TRACE functions based on both a variable and a #define?

12-Mar-21 1:51

On the following webpage C Preprocessor tricks, tips, and idioms · pfultz2/Cloak Wiki · GitHub[^] it is shown how an IIF (immediate if) statement for macros can be implemented. But as soon as I pass an expression, as oppose to either 0 or 1, to it, I can't get it to compile either:

C++

#define USER_EVENTS_LOGGING          0x00000001
#define COMMUNICATION_BUS_LOGGING    0x00000002

#if defined(ENABLE_USER_EVENTS_LOGGING) && defined(ENABLE_COMMUNICATION_BUS_LOGGING)
#define HARDCODE_ENABLED_MY_GRP_BIT_FLAGS      (USER_EVENTS_LOGGING | COMMUNICATION_BUS_LOGGING)
#elif !defined(ENABLE_USER_EVENTS_LOGGING) && defined(ENABLE_COMMUNICATION_BUS_LOGGING)
#define HARDCODE_ENABLED_MY_GRP_BIT_FLAGS      (COMMUNICATION_BUS_LOGGING)
#elif defined(ENABLE_USER_EVENTS_LOGGING) && !defined(ENABLE_COMMUNICATION_BUS_LOGGING)
#define HARDCODE_ENABLED_MY_GRP_BIT_FLAGS      (USER_EVENTS_LOGGING)
#else
#define HARDCODE_ENABLED_MY_GRP_BIT_FLAGS      (0)
#endif

IIF(1)(true, false)                                                               // Compiles fine
IIF((0 < (HARDCODE_ENABLED_MY_GRP_BIT_FLAGS & USER_EVENTS_LOGGING)))(true, false) // Doesn't compile!

Richard MacCutchan12-Mar-21 2:43