Rectpack is a collection of heuristic algorithms for solving the 2D knapsack problem, also known as the bin packing problem. In essence packing a set of rectangles into the smallest number of bins.
Download the package or clone the repository, and then install with:
python setup.py install
or use pypi:
pip install rectpack
Packing rectangles into a number of bins is very simple:
from rectpack import newPacker
rectangles = [(100, 30), (40, 60), (30, 30),(70, 70), (100, 50), (30, 30)]
bins = [(300, 450), (80, 40), (200, 150)]
packer = newPacker()
# Add the rectangles to packing queue
for r in rectangles:
packer.add_rect(*r)
# Add the bins where the rectangles will be placed
for b in bins:
packer.add_bin(*b)
# Start packing
packer.pack()
Once the rectangles have been packed the results can be accessed in several ways:
# Obtain number of bins used for packing
nbins = len(packer)
# Index first bin
abin = packer[0]
# Bin dimmensions (bins can be reordered during packing)
width, height = abin.width, abin.height
# Number of rectangles packed into first bin
nrect = len(packer[0])
# Second bin first rectangle
rect = packer[1][0]
# rect is a Rectangle object
x = rect.x # rectangle bottom-left x coordinate
y = rect.y # rectangle bottom-left y coordinate
w = rect.width
h = rect.height
But in the majority of cases you will only need a rectangle list:
# Full rectangle list
all_rects = packer.rect_list()
for rect in all_rects:
b, x, y, w, h, rid = rect
# b - Bin index
# x - Rectangle bottom-left corner x coordinate
# y - Rectangle bottom-left corner y coordinate
# w - Rectangle width
# h - Rectangle height
# rid - User asigned rectangle id or None
Lastly all the dimmension (bins and rectangles) must be integers or decimals to avoid collisions caused by floating point rounding. If your data is floating point use float2dec to convert float values to decimals (see float below)
A more detailed description of API calls:
-
class newPacker([, mode][, bin_algo][, pack_algo][, sort_algo][, rotation])
Return a new packer object- mode: Mode of operations
- PackingMode.Offline: The set of rectangles is known beforehand, packing won't start until pack() is called.
- PackingMode.Online: The rectangles are unknown at the beginning of the job, and will be packed as soon as they are added.
- bin_algo: Bin selection heuristic
- PackingBin.BNF: (Bin Next Fit) If a rectangle doesn't fit into the current bin, close it and try next one.
- PackingBin.BFF: (Bin First Fit) Pack rectangle into the first bin it fits (without closing)
- PackingBin.BBF: (Bin Best Fit) Pack rectangle into the bin that gives best fitness.
- PackingBin.Global: For each bin pack the rectangle with the best fitness until it is full, then continue with next bin.
- pack_algo: One of the supported packing algorithms (see list below)
- sort_algo: Rectangle sort order before packing (only for offline mode)
- SORT_NONE: Rectangles left unsorted.
- SORT_AREA: Sort by descending area.
- SORT_PERI: Sort by descending perimeter.
- SORT_DIFF: Sort by difference of rectangle sides.
- SORT_SSIDE: Sort by shortest side.
- SORT_LSIDE: Sort by longest side.
- SORT_RATIO: Sort by ration between sides.
- rotation: Enable or disable rectangle rotation.
- mode: Mode of operations
-
packer.add_bin(width, height[, count])
Add empty bin or bins to a packer- width: Bin width
- height: Bin height
- count: Number of bins to add, 1 by default. It's possible to add infinie bins with count=float("inf")
-
packer.add_rect(width, height[, rid])
Add rectangle to packing queue- width: Rectangle width
- height: Rectangle height
- rid: User assigned rectangle id
-
packer.pack():
Starts packing process (only for offline mode). -
packer.rect_list():
Returns the list of packed rectangles, each one represented by the tuple (b, x, y, w, h, rid) where:- b: Index for the bin the rectangle was packed into
- x: X coordinate for the rectangle bottom-left corner
- y: Y coordinate for the rectangle bottom-left corner
- w: Rectangle width
- h: Rectangle height
- rid: User provided id or None
This library implements three of the algorithms described in [1] Skyline, Maxrects, and Guillotine, with the following variants:
-
MaxRects
- MaxRectsBl
- MaxRectsBssf
- MaxRectsBaf
- MaxRectsBlsf
-
Skyline
- SkylineBl
- SkylineBlWm
- SkylineMwf
- SkylineMwfl
- SkylineMwfWm
- SkylineMwflWm
-
Guillotine
- GuillotineBssfSas
- GuillotineBssfLas
- GuillotineBssfSlas
- GuillotineBssfLlas
- GuillotineBssfMaxas
- GuillotineBssfMinas
- GuillotineBlsfSas
- GuillotineBlsfLas
- GuillotineBlsfSlas
- GuillotineBlsfLlas
- GuillotineBlsfMaxas
- GuillotineBlsfMinas
- GuillotineBafSas
- GuillotineBafLas
- GuillotineBafSlas
- GuillotineBafLlas
- GuillotineBafMaxas
- GuillotineBafMinas
I recommend to use the default algorithm unless the packing is too slow, in that case switch to one of the Guillotine variants for example GuillotineBssfSas. You can learn more about the algorithms in [1].
Rectpack is thoroughly tested, run the tests with:
python setup.py test
or
python -m unittest discover
If you need to use floats just convert them to fixed-point using a Decimal type, be carefull rounding up so the actual rectangle size is always smaller than the conversion. Rectpack provides helper funcion float2dec for this task, it accepts a number and the number of decimals to round to, and returns the rounded Decimal.
from rectpack import float2dec, newPacker
float_rects = [...]
dec_rects = [(float2dec(r[0], 3), float2dec(r[1], 3)) for r in float_rects]
p = newPacker()
...
[1] Jukka Jylang - A Thousand Ways to Pack the Bin - A Practical Approach to Two-Dimensional Rectangle Bin Packing (2010)
[2] Huang, E. Korf - Optimal Rectangle Packing: An Absolute Placement Approach (2013)