Never a mistake is stupid... unless you do the same twice. We learn from mistakes, to do improved versions... in more spectacular failures!Originally Posted by audioinstinct
That's a compiler problem. Never used FPC and not sure if Delphi is able to catch that kind of errors at compile time, I think it warns about uninitialized variables.
If you ever play with pointers, you are going to see weirder things for sure.
You can make same mistake with Delphi. It also took me quite some time to learn that.
Nof if I'm not mistake this is left in in Objective Pascal as backward compatibility to the verry first days of Objective Pascal when classes still didn't have capability to automatically initialize its own members (internal fields, variables, etc.).
I think I once sow a code that actually makes use of this to delay the initialization of some class members in order to speed up the creation process of the class itself.
But this is pactically not used nowadays becoouse you would usually want for your class to be the decendant of TObject class which provides you with some basic and comonly used functionality like class name property with which you can determine the class type at runtime without the use of extended RTL, then quckly creating of new instances of that particular class etc.
So before I write the code for this method I was thinking I'd ask for a bit of advice here. I don't want to rewrite it later, and my approach is pretty brutal, I get this feeling that there just must be a more efficient way...
First, I will explain the concepts of Hidden Pairs, in Sudoku theory. This is a really simple one.
It means that in a house(row, column or box) TWO candidates both have an incidence of two in that house, and both candidates reside in the same two cells. Currently, there may be other candidates too in these cells.
What does this mean? Well, think about it... We know that both candidates must be somewhere inside that house and we know that each candidate can only occur once.
Let's say we're in row #3, and we have the following situation:
Candidates for position #4: 3 ,4, 6 ,7. Candidates for position # 7: 3 ,4,5,6,8. There is no other cell in this row that contains 3 or 6 as a candidate.
Let's assume that the cell at position #4 is not a 3 or a 6. What does this mean? Well it means that both 3 and 6 would have only one possible place in this row, and that is at position #7. Since a cell can only have one value at a time, we can conclude that the cell at position #4 must be either a 3 or a 6. We can say the same about the cell at position #7.
Now my question is, what do you think would be the most efficient way of finding these? What I have done certainly does work, but I am sure there is an easier or more efficient way. This is how it works right now:
As you can see it is a mass amount of loops going on here, and it looks pretty ugly.Code:Procedure Find_Hidden_Pairs; var i,j:smallint; // i always marks a row and j always marks a column in my code box:smallint; //boxes are numbered from 1-9 from left to right, up to down. count1,count2:smallint; //These are the variables for the possible candidates. candidate:smallint; //This will be used in a short loop to determine which candidates can be eliminated. TargetCell1,TargetCell2:Cell; //These will be the cells where an elimination will occur Begin TargetCell1:=Cell.create(1,1); TargetCell2:=Cell.create(1,1); //It doesn't matter that yet they have the same properties, I just want to initialize them. for count1:=1 to 9 do begin //Check for Rows for i:=1 to 9 do begin if ( Candidate_Count(count1,i,'row')=2 ) then //Function Candidate_Count(candidate,housenumber:smallint;housetype:string):smallint returns how many candidates exist in a house. begin for count2:=1 to 9 do begin if (count1<>count2) and (Candidate_Count(count2,i,'row')=2) then //We don't want the program to find a 3 that pairs with 3... begin TargetCell1.x:=0; TargetCell2.x:=0; for j:=1 to 9 do begin if ( Candidate_Exists(i,j,count1) )and (Candidate_Exists(i,j,count2) ) then begin {Function Candidate_Exists(row,column,candidate:smallint):boolean Checks if a cell has a possible candidate. This is necessary because it is possible that 2 candidates only have 2 instances but they reside in more than 2 cells. Such as in a row: position#1: 1,3 position#2: 5,3 position#3: 4,1. And the other cells don't have 3 or 1 as candidates, now obviously these don't form pairs} if TargetCell1.x=0 then begin TargetCell1.x:=i; TargetCell1.y:=j; end else begin TargetCell2.x:=i; TargetCell2.y:=j; end; end; end; if TargetCell2.x<>0 then begin for candidate:=1 to 9 do begin if (candidate<>count1) and (candidate<>count2) then begin Grid[TargetCell1.x,TargetCell1.y].Eliminate(candidate); Grid[TargetCell2.x,TargetCell2.y].Eliminate(candidate); end; end; end; end; end; end; end; //End of Row Checks {Here would come the column checks and box checks, I won't include because it follows the exact same procedure the box is only slightly different... sooo....} //Check for Columns //Check for Boxes end; TargetCell1.done; TargetCell2.done; End;
Right now it starts with candidate 1 on each row and if it finds a row where it occurs twice it checks the row for another candidate from 1-9 that occurs twice and if they happen to form a pair then it eliminates the other possibilities, if they don't happen to form a pair, then it will search for another second candidate, and so on, and then it will go to the next row, and if it doesn't find any rows for that candidate then it goes to columns and boxes and then it starts all over again with a starting candidate for 2, and goes on until the pair 9,8 is checked for box 9.
There should be a recursive way or I don't know, anything that won't cause it to go through 5 or 6 for loops just to check one row for 2 candidates. It seems way overkill. Anyone has an idea how can I make this more simple?
Last edited by audioinstinct; 05-01-2015 at 01:52 PM.
In order to reduce the number of loops you are doing I recomend you start using sets for storing the posible candidates instead of arrays. Why? Becouse this way you alow yourself to the use of set operators. In your case you are interested in intersection of two sets as this returns you another set that conatins only those elements that are present in all sets. Yes you can find duplicate elements in multiple sets with a single command.
Check this quick code to see an example of how to do this:
I belive that using of sets would come quite usefull even in other approaches.Code:type TCandidates = set of 1..9; ... procedure TForm2.Button1Click(Sender: TObject); var Candidates1, Candidates2, Candidates3: TCandidates; Duplicates: TCandidates; Candidate: Integer; begin Candidates1 := [3,4,6,7]; Candidates2 := [3,4,5,6,8]; Candidates3 := [4,6]; Memo1.Lines.Add('Intersection between Candidate1 and Candidates2'); Duplicates := Candidates1 * Candidates2; for Candidate in Duplicates do begin Memo1.Lines.Add(IntToStr(Candidate)); end; Memo1.Lines.Add('Intersection between Candidate1, Candidates2 and Candidates3'); Duplicates := Candidates1 * Candidates2 * Candidates3; for Candidate in Duplicates do begin Memo1.Lines.Add(IntToStr(Candidate)); end; end;
EDIT: In order to be able to use for in loops you need to use Delphi 2005 or newer or FPC 2.4.2 or newer.
EDIT2: And for all those of you who still might be using older Delphi version or other pascal compilers which doesn't support for in loops you first need to declare colection of all posible items in your set and then declare set of that colection like so:
And then you need to loop through your colection of posible items and test for each on if it is present in your set like so:Code:type TCandidate = 1..9; TSetOfCandidates = set of TCandidate;
Yes this is quite messy and performance ineficient especially if you are working with sets that can have large number of posible elements.Code:procedure TForm3.Button1Click(Sender: TObject); var Candidates: TSetOfCandidates; Candidate: TCandidate; begin Candidates := [1,5,7]; for Candidate := Low(Candidate) to High(Candidate) do begin if Candidate in Candidates then Memo1.Lines.Add(IntToStr(Candidate)); end; end;
Just imagine you have to loop through a set that can conatin all 16 bit unsigned integer values (all 65536 of them) and you have to test for each to see if it is present in your desired set. Doing something like this could take ages.
Last edited by SilverWarior; 05-01-2015 at 05:38 PM.
Thanks!! That's a very helpful idea. Indeed, when I'm going to rewrite the XY-wing procedure for example, this will be great. I had a seperate function just so I could calculate the common candidates of two cells, instead with sets this will become really simple, so yeah it definitely will come in handy later on too.In order to reduce the number of loops you are doing I recomend you start using sets for storing the posible candidates instead of arrays. Why? Becouse this way you alow yourself to the use of set operators. In your case you are interested in intersection of two sets as this returns you another set that conatins only those elements that are present in all sets. Yes you can find duplicate elements in multiple sets with a single command.
Check this quick code to see an example of how to do this:
I belive that using of sets would come quite usefull even in other approaches.Code:type TCandidates = set of 1..9; ... procedure TForm2.Button1Click(Sender: TObject); var Candidates1, Candidates2, Candidates3: TCandidates; Duplicates: TCandidates; Candidate: Integer; begin Candidates1 := [3,4,6,7]; Candidates2 := [3,4,5,6,8]; Candidates3 := [4,6]; Memo1.Lines.Add('Intersection between Candidate1 and Candidates2'); Duplicates := Candidates1 * Candidates2; for Candidate in Duplicates do begin Memo1.Lines.Add(IntToStr(Candidate)); end; Memo1.Lines.Add('Intersection between Candidate1, Candidates2 and Candidates3'); Duplicates := Candidates1 * Candidates2 * Candidates3; for Candidate in Duplicates do begin Memo1.Lines.Add(IntToStr(Candidate)); end; end;
EDIT: In order to be able to use for in loops you need to use Delphi 2005 or newer or FPC 2.4.2 or newer.
EDIT2: And for all those of you who still might be using older Delphi version or other pascal compilers which doesn't support for in loops you first need to declare colection of all posible items in your set and then declare set of that colection like so:
And then you need to loop through your colection of posible items and test for each on if it is present in your set like so:Code:type TCandidate = 1..9; TSetOfCandidates = set of TCandidate;
Yes this is quite messy and performance ineficient especially if you are working with sets that can have large number of posible elements.Code:procedure TForm3.Button1Click(Sender: TObject); var Candidates: TSetOfCandidates; Candidate: TCandidate; begin Candidates := [1,5,7]; for Candidate := Low(Candidate) to High(Candidate) do begin if Candidate in Candidates then Memo1.Lines.Add(IntToStr(Candidate)); end; end;
Just imagine you have to loop through a set that can conatin all 16 bit unsigned integer values (all 65536 of them) and you have to test for each to see if it is present in your desired set. Doing something like this could take ages.
@Phibermon @AirPas
Please keep the conversation about the Phibermons engine in seperate thread as that is off-topci for this one which is about making sudoku solver. Let us keep PGD forums organized shall we.
I can move all posts about your Phibermons engine to another thread. Just say where OK.
Removed, it's not importantI'll post about it when I'm ready to. AirPas please feel free to get in touch if you want some more details
When the moon hits your eye like a big pizza pie - that's an extinction level impact event.
I apologize for nitpicking but it must be stressed that 'Objective Pascal' (A pascal equivalent to Objective-C for using Apple Objective-C interfaces) is something quite different to 'Object Pascal' (What everybody is thinking about)
--
I was going to write a Monte Carlo Sudoku solver for you to demonstrate the principal (The basic premise of a Monte Carlo simulation is to converge upon an answer to a problem by using random sampling (or more generally to determine the probability distribution of an unknown value which in turn can be used to estimate the true value))
However a search online provides an excellent approach to this technique that you can investigate : http://www.lptmc.jussieu.fr/user/talbot/sudoku.html
Before you get your feet too wet, a good learning exercise for the Monte Carlo method is to write a program to estimate the value of Pi by using either the ratio of a circle to a square with a length equal to the diameter of the circle or the more interesting technique which is to code a virtual buffon's needle simulation and measure the ratio of needle samples that cross lines to those that fall between.
One of the most interesting things about Monte Carlo simulations is that they become more accurate (or converge more quickly depending on the class of problem) the more 'random' your random generator is. By using a true random source of a uniform distribution you'll converge on answers more quickly or provide more accurate estimates than if you use the standard pseudo random generator provided in programming libraries. I had fascinating results (Buffon's needle Pi estimator) using a stream of true atomic random numbers as seeds for a pseudo generator. I even gained a whole decimal place of accuracy by feeding in random mouse movements (ensuring you get a good uniform distribution of samples before you re-seed - distribution of random generators and pseudo random generators in general can differ wildly, many are *not* uniform. I know that the FPC random functions are uniformly distributed but I can't vouch for Delphi)
It's truly mind blowing when you get it and it's a lot simpler than it sounds. Plus how cool is it to say you solve sudoku puzzles using the same technique that's used to model nuclear explosions?
Last edited by phibermon; 05-01-2015 at 06:35 PM.
When the moon hits your eye like a big pizza pie - that's an extinction level impact event.
Thanks, when I'll have more time I'm going to look at it. Problem is that I don't know shit about C, and these variable names like "is" and "jb" are just too confusing, they say nothing about what it means so it makes it even more difficult for me to understand what the program does.
Hey, thanks for the replies and all the help, guys. I had to take a pause from continuing the program, I am having some difficult exams, I will continue when I'm done with them and keep this thread updated.
Peace.
Reply With Quote
Bookmarks