` Bin Packing Algorithm to find out best use of material (Michael Ware) - Icetips Article
Icetips - Templates, Tools & Utilities for Clarion Developers

Templates, Tools and Utilities
for Clarion Developers

Icetips Article

Back to article list   Search Articles     Add Comment     Printer friendly     Direct link  

Clarion 6.1: Bin Packing Algorithm to find out best use of material
2005-02-02 -- Michael Ware
 
Newsgroups: comp.lang.clarion > I'm looking for an algorithm or another solution to solve my following > problem: > > I've got a queue with requested lengths: > .... > And I've got a queue with what is in stock: >.... > Is there a known algorithm that does something similar to achieve this or > can somebody help me in the right direction. Nice exercise! I love questions like this, it gives me a chance to exercise an area of my brain that doesn't usually get a good work out. I spent my lunch break today playing with some code. The attached compiles, but I don't have time to run it test it or debug it. For all I know it will just drop into an endless loop and you will have to reboot. It could definitely use some tuning and clean up. But it should be good enough to give you an idea of one approach. Assumptions I made were as follows: 1. You start with two Q's one for the requests and one for the stock on hand. 2. In both Q's the Length field is unique. i.e. you will NOT have two request records for the same length. 3. If ANY combination results in Zero waist for a particular stock it should be used. 4. The queue sizes are reasonable (~10 to 500 records). #s 1 & 2 are required for the code to work. #3 is a question for a mathematician I can't answer. #4 is required for the code to process in a reasonable amount of time. -Mike (p.s. if you do use this, you can pay me back by e-mailing me the changes and cleanup you do to get it working.) FindCuts PROCEDURE ! Declare Procedure WAIST LONG StockQ QUEUE,PRE(SQ) L LONG C LONG END RequestQ QUEUE,PRE(RQ) L LONG C LONG END CutsQ QUEUE,PRE(CQ) STOCKL LONG L1 LONG C1 LONG L2 LONG C2 LONG L3 LONG C3 LONG Waist LONG END FOUNDL1 LONG FOUNDL2 LONG FOUNDL3 LONG FOUNDC1 LONG FOUNDC2 LONG FOUNDC3 LONG CODE ! Process all stock longest to shortest FREE(CUTSQ) SORT(STOCKQ,-SQ:L) DO ResetStockQ ! Clears cutQ, & Positions to first stock LOOP WAIST = SQ:L ! Initialize Waist to this stock lenth ! Find the best combo for this stock ! lenght. DO FirstLength IF ~FOUNDL1 ! Either Found a combo with longer stock ! or no stock left that can fill a request ! or no requests left to fill IF CQ:L1 DO AddCut ! Found a combo with longer stock CYCLE ! ADD IT AND START OVER. ELSE BREAK END END IF WAIST < CQ:Waist ! if a less waistfull CQ:StockL = SQ:L ! combination was CQ:Waist = WAIST ! found, move the CQ:L1 = FOUNDL1 ! current values into CQ:L2 = FOUNDL2 ! the CutQ CQ:L3 = FOUNDL3 CQ:C1 = FOUNDC1 CQ:C2 = FOUNDC2 CQ:C3 = FOUNDC3 END GET(StockQ,POINTER(StockQ)+1) ! Get the next stock size ! If we are out of stock to try or ! the last stock size we used had ! no waist, add the cut and start over IF ERRORCODE() OR WAIST = 0 ! DO AddCut END END ! try the next stock size ! Best fit cut was found, ! this adds it to the cut Queue ! and reduces the Stock and ! request Queue counts accordingly AddCut ROUTINE ADD(CUTSQ) ! Add cut SQ:L = CQ:StockL GET(StockQ,SQ:L) SQ:C -= 1 ! Reduce Stock PUT(StockQ) RQ:L = CQ:L1 GET(RequestQ,RQ:L) RQ:C -= CQ:C1 ! Reduce requests PUT(RequestQ) IF CQ:L2 RQ:L = CQ:L2 GET(RequestQ,RQ:L) RQ:C -= CQ:C2 ! Reduce requests PUT(RequestQ) END IF CQ:L3 RQ:L = CQ:L3 GET(RequestQ,RQ:L) RQ:C -= CQ:L3 ! Reduce requests PUT(RequestQ) END DO ResetStockQ ! Clear cut queue ResetStockQ ROUTINE CQ:STOCKL =0 CQ:L1 =0 CQ:C1 =0 CQ:L2 =0 CQ:C2 =0 CQ:L3 =0 CQ:C3 =0 CQ:Waist =0 GET(StockQ,1) ! Position to first Stock CQ:Waist = SQ:L ! Set Maximum Waist FirstLength ROUTINE DATA StartW LONG BestW1 LONG TestL LONG TestC LONG MaxCnt LONG HoldL2 LONG HoldC2 LONG HoldL3 LONG HoldC3 LONG CODE FOUNDL1 = 0 FOUNDC1 = 0 FOUNDL2 = 0 FOUNDC2 = 0 FOUNDL3 = 0 FOUNDC3 = 0 StartW = WAIST SORT(RequestQ,-RQ:L) ! Sort by requested length desc. GET(REQUESTQ,1) LOOP MaxCnt = CHOOSE( INT(StartW/RQ:L)>RQ:C,RQ:C,INT(StartW/RQ:L)) IF MaxCnt >= 1 ! Request fits in this stock TestL = RQ:L LOOP TestC = MaxCnt TO 1 BY -1 IF TestL*TestC > WAIST THEN CYCLE. WAIST = WAIST - (TestL*TestC) DO SECONDLENGTH IF WAIST < BestW1 BestW1 = WAIST FOUNDL1 = Testl FOUNDC2 = TestC HoldL2 = FOUNDL2 HoldC2 = FOUNDC2 HoldL3 = FOUNDL3 HoldC3 = FOUNDC3 END END END GET(REQUESTQ,POINTER(REQUESTQ)+1) IF ERRORCODE() THEN BREAK. END FOUNDL2 = HoldL2 FOUNDC2 = HoldC2 FOUNDL3 = HoldL3 FOUNDC3 = HoldC3 WAIST = BestW1 SECONDLENGTH ROUTINE DATA StartQ LONG StartW LONG BestW2 LONG TestL LONG TestC LONG MaxCnt LONG HoldL3 LONG HoldC3 LONG CODE StartQ = POINTER(RequestQ) StartW = WAIST BestW2 = WAIST FOUNDL2 = 0 FOUNDC2 = 0 FOUNDL3 = 0 FOUNDC3 = 0 LOOP IF BestW2 = 0 THEN BREAK. GET(REQUESTQ,POINTER(REQUESTQ)+1) IF ERRORCODE() THEN BREAK. MaxCnt = CHOOSE( INT(StartW/RQ:L)>RQ:C,RQ:C,INT(StartW/RQ:L)) IF MaxCnt<1 THEN CYCLE. ! Length exceeds waist TestL = RQ:L LOOP TestC = MaxCnt TO 1 BY -1 IF (TestL*TestC) > WAIST THEN CYCLE. WAIST = StartW - (TestL*TestC) DO ThirdLENGTH IF WAIST < BestW2 FOUNDL2 = TestL FOUNDC2 = TestC HoldL3 = FoundL3 HoldC3 = FoundC3 BestW2 = WAIST IF BestW2=0 THEN BREAK. END END END WAIST = BestW2 FoundL3 = HoldL3 FoundC3 = HoldC3 GET(RequestQ,StartQ) ThirdLENGTH ROUTINE DATA StartQ LONG StartW LONG BestW3 LONG TestL LONG TestC LONG MaxCnt LONG CODE StartQ = POINTER(RequestQ) StartW = WAIST BestW3 = WAIST FoundL3 = 0 FoundC3 = 0 LOOP IF BestW3 = 0 THEN BREAK. GET(REQUESTQ,POINTER(REQUESTQ)+1) IF ERRORCODE() THEN BREAK. MaxCnt = CHOOSE( INT(StartW/RQ:L)>RQ:C,RQ:C,INT(StartW/RQ:L)) IF MaxCnt < 1 THEN CYCLE. ! Length exceeds waist TestL = RQ:L LOOP TestC = MaxCnt TO 1 BY -1 WAIST = StartW - (TestL*TestC) IF WAIST < BestW3 AND WAIST > 0 FoundL3 = TestL FoundC3 = TestC BestW3 = WAIST IF BestW3=0 THEN BREAK. END END END WAIST = BestW3 GET(RequestQ,StartQ)


Today is April 25, 2024, 1:20 am
This article has been viewed 35119 times.
Google search has resulted in 40 hits on this article since January 25, 2004.



Back to article list   Search Articles   Add Comment   Printer friendly

Login

User Name:

Password: