{- Copyright (c) 2008-2011 John Goerzen All rights reserved. For license and copyright information, see the file LICENSE -} {- | Module : Data.BinPacking Copyright : Copyright (C) 2008-2011 John Goerzen SPDX-License-Identifier: BSD-3-Clause Stability : provisional Portability: portable Tools for packing into bins Written by John Goerzen, jgoerzen\@complete.org This module is designed to solve this type of problem: Given a bunch of objects of varying sizes, what is the best possible way to pack them into fixed-size bins? This can be used, for instance, by the datapacker program to pack files onto CDs or DVDs; by manufacturing environments to pack physical items into physicl bins; etc. A description of bin packing algorithms can be found at . -} module Data.BinPacking (BinPacker, BinPackerError(..), packByOrder, packLargeFirst ) where import Data.List import Control.Monad.Error {- | Potential errors returned as Left values by 'BinPacker' functions. Calling 'show' on this value will produce a nice error message suitable for display. -} data (Num size, Ord size, Show size, Show obj) => BinPackerError size obj = BPTooFewBins [(size, obj)] -- ^ Ran out of bins; attached value is the list of objects that do not fit | BPSizeTooLarge size (size, obj) -- ^ Bin size1 exceeded by at least the given object and size | BPOther String -- ^ Other error deriving (Eq, Read) instance (Num size, Ord size, Show size, Show obj) => Show (BinPackerError size obj) where show (BPTooFewBins _) = "Too few bins" show (BPSizeTooLarge binsize (objsize, obj)) = "Size " ++ show objsize ++ " greater than bin size " ++ show binsize ++ " at " ++ show obj show (BPOther x) = x {- | Let us use this as part of the Either monad -} instance (Num size, Ord size, Show size, Show obj) => Error (BinPackerError size obj) where strMsg = BPOther {- | The primary type for bin-packing functions. These functions take a list of size of bins. If every bin is the same size, you can pass @repeat binSize@ to pass an infinite list of bins if the same size. Any surplus bins will simply be ignored. > [size] is the sizes of bins > [(size, obj)] is the sizes and objects > result is Either error or results -} type BinPacker = forall size obj. (Num size, Ord size, Show size, Show obj) => [size] -- The sizes of bins -> [(size, obj)] -- The sizes and objects -> Either (BinPackerError size obj) [[(size, obj)]] -- Either error or results {- | Pack objects into bins, preserving order. Objects will be taken from the input list one by one, and added to each bin until the bin is full. Work will then proceed on the next bin. No attempt is made to optimize allocations to bins. This is the simplest and most naive bin-packing algorithm, but may not make very good use of bin space. -} packByOrder :: BinPacker packByOrder _ [] = Right [] -- Ran out of sizes packByOrder [] remainder = Left (BPTooFewBins remainder) packByOrder (thisbinsize:otherbins) sizes = let fillBin _ [] = Right [] fillBin accumsize ((s, o):xs) | s > thisbinsize = Left $ BPSizeTooLarge thisbinsize (s, o) | s + accumsize > thisbinsize = Right [] | otherwise = do next <- fillBin (accumsize + s) xs return $ (s, o) : next in do thisset <- fillBin 0 sizes next <- packByOrder otherbins (drop (length thisset) sizes) return (thisset : next) {- | Pack objects into bins. For each bin, start with the largest objects, and keep packing the largest object from the remainder until no object can be found to put in the bin. This is substantially more efficient than 'packByOrder', but requires sorting the input. -} packLargeFirst :: BinPacker packLargeFirst bins sizes = packLargeFirst' bins (sortBy fstSort sizes) where fstSort a b = compare (fst a) (fst b) packLargeFirst' :: BinPacker packLargeFirst' _ [] = Right [] -- Ran out of sizes packLargeFirst' [] remainder = Left (BPTooFewBins remainder) packLargeFirst' (thisbinsize:otherbins) sizes = let fillBin _ [] = Right [] fillBin accumsize sizelist = case break (\x -> (fst x) + accumsize <= thisbinsize) sizelist of (_, []) -> if accumsize == 0 then Left $ BPSizeTooLarge thisbinsize (head sizelist) else Right [] (nonmatches, ((s, o):matchxs)) -> do next <- fillBin (accumsize + s) (nonmatches ++ matchxs) return $ (s, o) : next in do thisset <- fillBin 0 sizes next <- packLargeFirst' otherbins (drop (length thisset) sizes) return (thisset : next)