binpacking

This package contains greedy algorithms to solve two typical bin packing problems, (i) sorting items into a constant number of bins, (ii) sorting items into a low number of bins of constant size. Here's a usage example

>>> import binpacking
>>>
>>> b = { 'a': 10, 'b': 10, 'c':11, 'd':1, 'e': 2,'f':7 }
>>> bins = binpacking.to_constant_bin_number(b,4) # 4 being the number of bins
>>> print("===== dict\n",b,"\n",bins)
===== dict
 {'a': 10, 'b': 10, 'c': 11, 'd': 1, 'e': 2, 'f': 7}
 [{'c': 11}, {'b': 10}, {'a': 10}, {'f': 7, 'e': 2, 'd': 1}]
>>>
>>> b = list(b.values())
>>> bins = binpacking.to_constant_volume(b,11) # 11 being the bin volume
>>> print("===== list\n",b,"\n",bins)
===== list
 [10, 10, 11, 1, 2, 7]
 [[11], [10], [10], [7, 2, 1]]

Consider you have a list of items, each carrying a weight w_i. Typical questions are

1. How can we distribute the items to a minimum number of bins N of equal volume V?
2. How can we distribute the items to exactly N bins where each carries items that sum up to approximately equal weight?

Problems like this can easily occur in modern computing. Assume you have to run computations where a lot of files of different sizes have to be loaded into the memory. However, you only have a machine with 8GB of RAM. How should you bind the files such that you have to run your program a minimum amount of times? This is equivalent to solving problem 1.

What about problem 2? Say you have to run a large number of computations. For each of the jobs you know the time it will probably take to finish. However, you only have a CPU with 4 cores. How should you distribute the jobs to the 4 cores such that they will all finish at approximately the same time?

The package provides the command line tool "binpacking" using which one can easily bin pack csv-files containing a column that can be identified with a weight. To see the usage enter

$ binpacking -h
Usage: binpacking [options]

Options:
  -h, --help            show this help message and exit
  -f FILEPATH, --filepath=FILEPATH
                        path to the csv-file to be bin-packed
  -V V_MAX, --volume=V_MAX
                        maximum volume per bin (constant volume algorithm will
                        be used)
  -N N_BIN, --n-bin=N_BIN
                        number of bins (constant bin number algorithm will be
                        used)
  -c WEIGHT_COLUMN, --weight-column=WEIGHT_COLUMN
                        integer (or string) giving the column number (or
                    column name in header) where the weight is stored
  -H, --has-header      parse this option if there is a header in the csv-file
  -d DELIM, --delimiter=DELIM
                        delimiter in the csv-file (use "tab" for tabs)
  -q QUOTECHAR, --quotechar=QUOTECHAR
                        quotecharacter in the csv-file
  -l LOWER_BOUND, --lower-bound=LOWER_BOUND
                        weights below this bound will not be considered
  -u UPPER_BOUND, --upper-bound=UPPER_BOUND
                        weights exceeding this bound will not be considered

Install

$ pip install binpacking

Examples

In the repository's directory

cd examples/
binpacking -f hamlet_word_count.csv -V 2000 -H -c count -l 10 -u 1000
binpacking -f hamlet_word_count.csv -N 4 -H -c count 

in Python

import binpacking

b = { 'a': 10, 'b': 10, 'c':11, 'd':1, 'e': 2,'f':7 }
bins = binpacking.to_constant_bin_number(b,4)
print("===== dict\n",b,"\n",bins)

b = list(b.values())
bins = binpacking.to_constant_volume(b,11)
print("===== list\n",b,"\n",bins)

Related packages

• prtpy by Erel Segal-Halevi.
• numberpartitioning by Søren Fuglede Jørgensen.