I think LCL has many CompareNNNN funcs, but misses func for TPoint. Here it's comparing x/y screen coordinates

function ComparePoints(P1, P2: TPoint): integer;
begin
   if (P1.X=P2.X) and (P1.Y=P2.Y) then exit(0);
   if (P1.Y>P2.Y) then exit(1);
   if (P1.Y<P2.Y) then exit(-1);
   if (P1.X>P2.X) then exit(1) else exit(-1);
end;

Alexey wrote

I think LCL has many CompareNNNN funcs, but misses func for TPoint. Here it's comparing x/y screen coordinates

Why not extend TPoint with an IsEquals() method, so you can do something like

var
   p1: TPoint;
   p2: TPoint;
begin
   // ... populate p1 & p2
   if p1.IsEquals(p2) then
     // do something.

ComparePoints() returning a Integer is not a very friendly return type either, whereas Boolean makes more sense.

Regards,
Graeme

What do you think about:

function ComparePoints(P1, P2: TPoint): TPoint;
begin
  Result.X := (P1.X - P2.X);
  Result.Y := (P1.Y - P2.Y);
end;

Best regards.

Sandro Cumerlato

On Jul 26, 2017 3:56 PM, "Alexey via Lazarus"
wrote:

I think LCL has many CompareNNNN funcs, but misses func for TPoint. Here it's comparing x/y screen coordinates

function ComparePoints(P1, P2: TPoint): integer;
begin
  if (P1.X=P2.X) and (P1.Y=P2.Y) then exit(0);
  if (P1.Y>P2.Y) then exit(1);
  if (P1.Y<P2.Y) then exit(-1);
  if (P1.X>P2.X) then exit(1) else exit(-1);
end;

First start by defining what ComparePoints is supposed to measure or compare.

If "equals" is needed, why not simply use p1=p2 ?

Michael.

Graeme Geldenhuys wrote

ComparePoints() returning a Integer is not a very friendly return type either,

Hm, in my app Int makes sense: i need to compare Points in binary search, and see: >, <, =.

This compare is very specific to your application.

I don't think it would be a good idea to generalize this.

Michael.

IMHO you should consider 5 cases: P1 = P2 and the four quadrants

https://en.m.wikipedia.org/wiki/Quadrant_(plane_geometry)

- Quadrant I should correspond to >
- Quadrant III should correspond to <
- What about quadrants II and IV? You could compare your point to Y = -X line to reduce your study to 3 cases.

Best regards

Sandro Cumerlato

On Jul 26, 2017 6:27 PM, "Alexey via Lazarus"
wrote:

Graeme Geldenhuys wrote

ComparePoints() returning a Integer is not a very friendly return type either,

Hm, in my app Int makes sense: i need to compare Points in binary search, and see: >, <, =.

Michael Van Canneyt wrote

First start by defining what ComparePoints is supposed to measure or compare.

If "compare" doesn't mean "equals", then a better name should be used, otherwise that too is up for incorrect interpretation.

If "equals" is needed, why not simply use p1=p2 ?

If operator overload exists for the types then sure, but as you know I like OOP approaches to things.

Also is TPoint a record or a object or a class?

In fpGUI the equivalent of TPoint and TRect are object types and has many related methods defined inside them. I don't believe p1 = p2 is going to work with the object data type.

https://github.com/graemeg/fpGUI/blob/develop/src/corelib/fpg_base.pas#L160..L191

Regards,
Graeme

Alexey wrote

Hm, in my app Int makes sense: i need to compare Points in binary search, and see: >, <, =.

Ever heard of Manhatten Length? That would be a more logical approach. It is a commonly used length calculation for points.

http://en.wikipedia.org/wiki/Taxicab_geometry

The interface as in fpGUI's TfpgPoint:

https://github.com/graemeg/fpGUI/blob/develop/src/corelib/fpg_base.pas#L164..L165

The implementations:

https://github.com/graemeg/fpGUI/blob/develop/src/corelib/fpg_base.pas#L2021..L2029

Regards,
Graeme

> First start by defining what ComparePoints is supposed to measure or > compare.

yes, of course: compare source editor x/y positions. Ie, it is CompareCoords.

If a metric would be more logical (I don't know why exactly it should be but whatever) the euclidean metric would be even more logical, wouldn't it?

It is a commonly used length calculation for points.

That sounds horrible.
Is length really used like this?
The length of two points???
Metrics are distance functions.

http://en.wikipedia.org/wiki/Taxicab_geometry

Any 2 dimensional metric can be used.
But it may totally not be the thing the TC needs as the metric conditions may not be desirable in his context.

The point is (and Michael made it already) that there is no definition of how to compare points. You have quite a lot of options and the current problem determines which function to use. Therefore it makes no sense to introduce a general ComparePoints-function.

R.

Am 26.07.2017 19:51 schrieb "Graeme Geldenhuys via Lazarus" <

like OOP approaches to things.

Also is TPoint a record or a object or a class?

It's an advanced record with methods and operator overloads: https://svn.freepascal.org/cgi-bin/viewvc.cgi/trunk/rtl/inc/typshrdh.inc?revision=33062&view=markup

In fpGUI the equivalent of TPoint and TRect are object types and has many

related methods defined inside them. I don't believe p1 = p2 is going to work with the object data type.

For object types you'll need to declare global operator overloads to use e.g. the = operator (which have the disadvantage that they don't work with generics... at least not yet :/ ).

Regards,
Sven
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.lazarus-ide.org/pipermail/lazarus/attachments/20170726/9ead2c8a/attachment-0001.html>

Hi Sven,

Thanks for clarifying the TPoint type. I thought it was a record structure, but just wanted to make sure - as I mentioned, different frameworks or libraries might use different types.

Sven Barth wrote

For object types you'll need to declare global operator overloads to use e.g. the = operator

Then it's also good that I don't like the = operator for objects and classes. As I mentioned earlier, I like the more OOP approach where methods of an object/class act on the data the type holds. So TPoint.Equals() makes much more sense to me that p1 = p2. Just my personal opinion I guess. ;-)

Regards,
Graeme

In that case you need some enumerated.
TRelativePosition =
(rpEqual,
rpNorth,rpEast,rpSouth,rpWest,

  rpNorthEast,rpSouthEast,rpSouthWest,rpNorhWest);

+1, 0, -1 does not tell you enough.

Michael.

Exactly.

Michael.

Therefore it makes no sense to introduce a general

ComparePoints-function.

Yes, it is right, func not needed. And my case [compare source editor x/y coords] is rare.