C / C++ / MFC

7-Aug-12 1:04

Hello everybody,

i am having n-number of points like p1(x,y), p2(x,y), p3(x,y), .... pn(x,y). All points lie on a single line.(either a curve or a straight line). i would like to find, using those points, whether that line is straight or not straight.
i googled and tried some of the methods given in examples, (straight line equation, finding if collinear) but those are not working fine even for straight line also.

i am using visual studio 2008 with win32 (not MFC).

Any ideas ?

Thanks in advance,
A. Gopinath.

Richard MacCutchan7-Aug-12 1:17

Richard MacCutchan

7-Aug-12 1:17

The problem is one of mathematics not programming; once you find the mathematical solution, converting it to C/C++ is the easy part.

One of these days I'm going to think of a really clever signature.

tagopi7-Aug-12 1:29

7-Aug-12 1:29

Hello Richard,

Yes, i know its a mathematics question, i tried here so that someone can help me.
Thanks for the reply.

Regards,
A. Gopinath.

Amarnath S7-Aug-12 1:18

Amarnath S

7-Aug-12 1:18

One way would be by using slope. Find the slope of the line corresponding to the first two points (y-coord of p2 - y-coord of p1) / (x-coord of p2 - x-coord of p1).
Find the slopes of the lines corresponding to all other points (with respect to the first point), say (y-coord of pk - y-coord of p1) / (x-coord of pk - x-coord of p1), where k varies from 3 to n.

All of these slopes should be within a reasonable tolerance band. And, since all of these lines pass through p1, they all would lie on the same straight line if they fell within the tolerance band.

The above would not work if the x-coordinates of p2 and p1 are the same (because denominator would be zero). In that case, you choose a point whose x-coordinate is unique as the reference point.

Not sure whether this is the most efficient way, but it certainly a way of achieving what you want.

tagopi7-Aug-12 1:30

7-Aug-12 1:30

Hello Amarnath,

yes, i tried this method also, but iam getting like 0.000002 or something like difference which fails the condition checking. rounding off will solve the solution ? is that right method?

Regards,
A. Gopinath.

Amarnath S7-Aug-12 1:45

Amarnath S

7-Aug-12 1:45

You mean, that your slope value is 0.000002? This means that your line is almost horizontal. Your tolerance values would have to be at least 1/1000 or 1/10000 of your "average" slope value. Then, it would indicate that these points are on a straight line. (You need to use a "double" type to store these numbers, and a "float" may not suffice).

By the way, what is the purpose of this exercise? That purpose would in some sense determine the method of straight-line detection.

tagopi7-Aug-12 1:49

7-Aug-12 1:49

Hello,

i will try using "double" type. actually this is related 3D project, need to find straight line or not.

Thanks again,
A. Gopinath.

Eugen Podsypalnikov7-Aug-12 4:14

Eugen Podsypalnikov

7-Aug-12 4:14

- Take the first and last points to build the 3D law of their straightness
- Define the fluctuation radius as "allowed error"
- All other points must lie in the "pipe" =>
=> any point projection onto the law(1) may not be greather as the radius(2) Smile | :)

They sought it with thimbles, they sought it with care;
They pursued it with forks and hope;
They threatened its life with a railway-share;
They charmed it with smiles and soap. Smile | :)

Amarnath S7-Aug-12 1:51

Amarnath S

7-Aug-12 1:51

One more way would be by using the cross product. Take two points, say p1 and p2 as reference. Consider the vector p1-p2. Now, take a point pk, and hence that vector p1-pk. Find the cross product of these two vectors (p1-p2 and p1-pk). That cross product should be near zero (all three components of the cross product) if the points are collinear. Repeat this for all points pk (k belonging to the set 3 to n).

modified 7-Aug-12 8:10am.

Albert Holguin7-Aug-12 1:40

Albert Holguin

7-Aug-12 1:40

Typically this is where you'd post questions regarding algorithms:
http://www.codeproject.com/Forums/326859/Algorithms.aspx[^]

tagopi7-Aug-12 1:50

7-Aug-12 1:50

Thanks Albert,

hereafter i will use this page to ask questions related.

Regards,
A. Gopinath.

pasztorpisti7-Aug-12 5:57

pasztorpisti

7-Aug-12 5:57

With some algorithm you should pick 2 points that are far enough from each other. After this you can create a line object using these 2 points. The line object can be used to easily determine the distance of the other points from this line and you can accept distances less then a defined epsilon value.

An O(n) algorithm to determine the 2 endpoints (assuming that the points are in a line) before your create the line object:
Initialize an empty bounding AABB and then expand it by adding all points. Declare six point variables and each of them should store the last point that was used to expand the AABB in +x,-x,+y,-y,+z,-z directions (note: the first point should fill all the six variables).
When this is done choose a dimension in which the size of the AABB is the biggest (x, y, or z). If the biggest dimension is zero (or near zero) then all points are located at the same spot.
Lets say its x: in this case the two endpoints (from the previously mentioned 6 point variables) are the points that were stored when the AABB was expanded last time in -x and +x directions.

C++ middleware with GUI in C

CPallini7-Aug-12 21:17

CPallini

7-Aug-12 21:17

That it's simple: if all the points are on a straight line then use two of them to compute the straight line equation.
Once you have such equation, test if the other points lie on the line too.

Veni, vidi, vici.

Member 83997696-Aug-12 20:02

Member 8399769

6-Aug-12 20:02

Hi,

I am new to this forum but i hope i am at the right place. I am working on a project which has to be upgraded from the base project written in C++. We have decided to use emwin from segger as GUI which is written in C. The big heck is we have our base project middle ware still in C++.

Please suggest me whether its a wise idea to keep middleware in C++ and go ahead with GUI in C. The GUI library is written with Extern C so we will not have problem calling the library function. But how can we get the data from C++ middleware application.

Thanks
Om

Richard MacCutchan6-Aug-12 20:59

Richard MacCutchan

6-Aug-12 20:59

Member 8399769 wrote:
But how can we get the data from C++ middleware application.

This question is not clear, what data, where is it held and where is it going? On a more basic point there is no problem using C-based libraries from C++ programs; after all the entire Windows API, including the GDI section, is written in C and used extensively from C++ programs.

One of these days I'm going to think of a really clever signature.

Member 83997696-Aug-12 21:58

Member 8399769

6-Aug-12 21:58

Thanks for the response.

My question was any body can provide some sample code in C++ from where we can access GUI which is written in c. I have GUI button click call back function also written in c language itself. how can we pass that to a c++ application.?

Richard MacCutchan6-Aug-12 22:46

Richard MacCutchan

6-Aug-12 22:46

Member 8399769 wrote:
some sample code in C++ from where we can access GUI which is written in c.

How can we provide sample code to a GUI that we do not have? Use the documentation for your library to access the various functions that you are using.

Member 8399769 wrote:
I have GUI button click call back function also written in c language itself. how can we pass that to a c++ application.?

The same way you would pass it to a C function. The syntax for both languages is exactly the same.

One of these days I'm going to think of a really clever signature.

JohnAspras6-Aug-12 21:21

JohnAspras

6-Aug-12 21:21

What i would do first is create wrapper classes around the C widgets.
and use C++ from then on...

wrapping will take a bit more time to develop but it will be worth the effort
since then on you can use the wrapper classes not only in the particular but any other project after that..

Software_Developer6-Aug-12 22:42

Software_Developer

6-Aug-12 22:42

For wrapping C header files into C++ you do this:

#ifdef __cplusplus
extern "C" {
#endif
/*List of C functions  */
#ifdef __cplusplus
}
#endif
#endif

More extern "C" [here] and [here].

[List] of GUI libraries.

pasztorpisti7-Aug-12 11:53

pasztorpisti

7-Aug-12 11:53

I tend to find that microcontroller guys are usually gravitating towards C from C++ and I feel something like that in your words. Rewriting a project from C++ to C is usually a downgrade especially in terms of software design if you don't have a strong reason to do that. I was also skeptic about c++ for a long time (I came from assembly, and C, pascal) but believe me that C++ allows you to write much more easy to maintain/clear and bug free code than C. Many ppl think that the only way to communicate between C and C++ compilation units is by calling from C++ to C. This is not true. You can put a function into a cpp file and can declare it as extern "C". In this calse you can call this C++ function from a C file but you can't use C++ types in the function declaration (like bool or pointers/refs to C++ classes), but inside the function body you can use pure C++ with classes/templates...

My advice: If you have an option to choose a C++ gui library than choose that and continue writing the project in C++. If you want to go on with your current C library then you should create wrapper classes around the stuff your gui has (buttons, editboxes, labels, dialogs, etc...) and then go on writing the project in C++ and use the gui library as C++ through its wrapper classes.

Any more questions? Smile | :)