cutstock.py

Column generation is a solution process that begins with a small, manageable part of a problem (specifically, a few of the variables), solves that part, analyzes the partial solution to determine the next part of the problem (specifically, one or more variables) to add to the model, and then solves the new, enlarged model. Column generation repeats the process until a satisfactory solution to the whole problem is achieved.

  1
  2
  3
  4
  5
  6
  7
  8
  9
 10
 11
 12
 13
 14
 15
 16
 17
 18
 19
 20
 21
 22
 23
 24
 25
 26
 27
 28
 29
 30
 31
 32
 33
 34
 35
 36
 37
 38
 39
 40
 41
 42
 43
 44
 45
 46
 47
 48
 49
 50
 51
 52
 53
 54
 55
 56
 57
 58
 59
 60
 61
 62
 63
 64
 65
 66
 67
 68
 69
 70
 71
 72
 73
 74
 75
 76
 77
 78
 79
 80
 81
 82
 83
 84
 85
 86
 87
 88
 89
 90
 91
 92
 93
 94
 95
 96
 97
 98
 99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
265
266
267
268
269
270
271
272
273
274
275
276
# --------------------------------------------------------------------------
# Source file provided under Apache License, Version 2.0, January 2004,
# http://www.apache.org/licenses/
# (c) Copyright IBM Corp. 2015, 2022
# --------------------------------------------------------------------------

from collections import namedtuple
import json

from docplex.util.environment import get_environment
from docplex.mp.model import Model


# ----------------------------------------------------------------------------
# Initialize the problem data
# ----------------------------------------------------------------------------
DEFAULT_ROLL_WIDTH = 110
DEFAULT_ITEMS = [(1, 20, 48), (2, 45, 35), (3, 50, 24), (4, 55, 10), (5, 75, 8)]
DEFAULT_PATTERNS = [(i, 1) for i in range(1, 6)]  # (1, 1), (2, 1) etc
DEFAULT_PATTERN_ITEM_FILLED = [(p, p, 1) for p in range(1, 6)]  # pattern1 for item1, pattern2 for item2, etc.

FIRST_GENERATION_DUALS = [1, 1, 1, 1, 0]


# ----------------------------------------------------------------------------
# Build the model
# ----------------------------------------------------------------------------
class TItem(object):
    def __init__(self, item_id, item_size, demand):
        self.id = item_id
        self.size = item_size
        self.demand = demand
        self.dual_value = -1

    @classmethod
    def make(cls, args):
        arg_id = args[0]
        arg_size = args[1]
        arg_demand = args[2]
        return cls(arg_id, arg_size, arg_demand)

    def __str__(self):
        return 'item%d' % self.id


class TPattern(namedtuple("TPattern", ["id", "cost"])):
    def __str__(self):
        return 'pattern%d' % self.id

# ---


def make_cutstock_pattern_generation_model(items, roll_width, **kwargs):
    gen_model = Model(name='cutstock_generate_patterns', **kwargs)
    # store data
    gen_model.items = items
    gen_model.roll_width = roll_width
    # default values
    gen_model.duals = [1] * len(items)
    # 1. create variables: one per item
    gen_model.use_vars = gen_model.integer_var_list(keys=items, ub=999999, name='use')
    # 2 setup constraint:
    # --- sum of item usage times item sizes must be less than roll width
    gen_model.add(gen_model.dot(gen_model.use_vars, (it.size for it in items)) <= roll_width)

    # store dual expression for dynamic edition
    gen_model.use_dual_expr = 1 - gen_model.dot(gen_model.use_vars, gen_model.duals)
    # minimize
    gen_model.minimize(gen_model.use_dual_expr)

    return gen_model


def cutstock_update_duals(gmodel, new_duals):
    # update the duals array and the the duals expression...
    # edition is propagated to the objective of the model.
    gmodel.duals = new_duals
    use_vars = gmodel.use_vars
    assert len(new_duals) == len(use_vars)
    updated_used = [(use, -new_duals[u]) for u, use in enumerate(use_vars)]
    # this modification is notified to the objective.
    gmodel.use_dual_expr.set_coefficients(updated_used)
    return gmodel


def make_custstock_master_model(item_table, pattern_table, fill_table, roll_width, **kwargs):
    m = Model(name='custock_master', **kwargs)

    # store data as properties
    m.items = [TItem.make(it_row) for it_row in item_table]
    m.items_by_id = {it.id: it for it in m.items}
    m.patterns = [TPattern(*pattern_row) for pattern_row in pattern_table]
    m.patterns_by_id = {pat.id: pat for pat in m.patterns}
    m.max_pattern_id = max(pt.id for pt in m.patterns)

    # build a dictionary storing how much each pattern fills each item.
    m.pattern_item_filled = {(m.patterns_by_id[p], m.items_by_id[i]): f for (p, i, f) in fill_table}
    m.roll_width = roll_width

    # --- variables
    # one cut var per pattern...
    m.MAX_CUT = 9999
    m.cut_vars = m.continuous_var_dict(m.patterns, lb=0, ub=m.MAX_CUT, name="cut")

    # --- add fill constraints
    #
    all_patterns = m.patterns
    all_items = m.items
    m.item_fill_cts = []
    for item in all_items:
        item_fill_ct = m.sum(
            m.cut_vars[p] * m.pattern_item_filled.get((p, item), 0) for p in all_patterns) >= item.demand
        item_fill_ct.name = 'ct_fill_{0!s}'.format(item)
        m.item_fill_cts.append(item_fill_ct)
    m.add_constraints(m.item_fill_cts)

    # --- minimize total cut stock
    m.total_cutting_cost = m.sum(m.cut_vars[p] * p.cost for p in all_patterns)
    m.minimize(m.total_cutting_cost)

    return m


def add_pattern_to_master_model(master_model, item_usages):
    """ Adds a new pattern to the master model.

    This function performs the following:

    1. build a new pattern instance from item usages (taken from sub-model)
    2. add it to the master model
    3. update decision objects with this new pattern.
    """
    new_pattern_id = max(pt.id for pt in master_model.patterns) + 1
    new_pattern = TPattern(new_pattern_id, 1)
    master_model.patterns.append(new_pattern)
    for used, item in zip(item_usages, master_model.items):
        master_model.pattern_item_filled[new_pattern, item] = used

    # --- add one decision variable, linked to the new pattern.
    new_pattern_cut_var = master_model.continuous_var(lb=0, ub=master_model.MAX_CUT,
                                                      name='cut_{0}'.format(new_pattern_id))
    master_model.cut_vars[new_pattern] = new_pattern_cut_var

    # update constraints
    for item, ct in zip(master_model.items, master_model.item_fill_cts):
        # update fill constraint by changing lhs
        ctlhs = ct.lhs
        filled = master_model.pattern_item_filled[new_pattern, item]
        if filled:
            ctlhs += new_pattern_cut_var * filled

    # update objective:
    #   side-effect on  the total cutting cost expr propagates to the objective.
    cost_expr = master_model.total_cutting_cost
    cost_expr += new_pattern_cut_var * new_pattern.cost  # this performw a side effect!

    return master_model


def cutstock_print_solution(cutstock_model):
    patterns = cutstock_model.patterns
    cut_var_values = {p: cutstock_model.cut_vars[p].solution_value for p in patterns}
    pattern_item_filled = cutstock_model.pattern_item_filled
    print("| Nb of cuts | Pattern   | Pattern's detail (# of item1,item2,...) |")
    print("| {} |".format("-" * 70))
    for p in patterns:
        if cut_var_values[p] >= 1e-3:
            pattern_detail = {b.id: pattern_item_filled[a, b] for a, b in pattern_item_filled if
                              a == p}
            print(
                "| {:<10g} | {!s:9} | {!s:45} |".format(cut_var_values[p],
                                                        p,
                                                        pattern_detail))
    print("| {} |".format("-" * 70))


def cutstock_save_as_json(model, json_file):
    patterns = model.patterns
    cut_var_values = {p: model.cut_vars[p].solution_value for p in patterns}
    solution = []
    for p in patterns:
        if cut_var_values[p] >= 1e-3:
            pattern_detail = {b.id: model.pattern_item_filled[(a, b)] for (a, b) in model.pattern_item_filled if
                              a == p}
            n = {'pattern': str(p),
                 'cuts': "%g" % cut_var_values[p],
                 'details': pattern_detail}
            solution.append(n)
    json_file.write(json.dumps(solution, indent=3).encode('utf-8'))


def cutstock_solve(item_table, pattern_table, fill_table, roll_width, **kwargs):
    verbose = kwargs.pop('verbose', True)
    master_model = make_custstock_master_model(item_table, pattern_table, fill_table, roll_width, **kwargs)

    # these two fields contain named tuples
    items = master_model.items
    patterns = master_model.patterns
    gen_model = make_cutstock_pattern_generation_model(items, roll_width, **kwargs)

    rc_eps = 1e-6
    obj_eps = 1e-4
    loop_count = 0
    best = 0
    curr = 1e+20
    ms = None

    while loop_count < 100 and abs(best - curr) >= obj_eps:
        ms = master_model.solve(**kwargs)
        loop_count += 1
        best = curr
        if not ms:
            print('{}> master model fails, stop'.format(loop_count))
            break
        else:
            assert ms
            curr = master_model.objective_value
            if verbose:
                print('{}> new column generation iteration, #patterns={}, best={:g}, curr={:g}'
                      .format(loop_count, len(patterns), best, curr))
            duals = master_model.dual_values(master_model.item_fill_cts)
            if verbose:
                print('{0}> moving duals from master to sub model: {1}'
                      .format(loop_count, list(map(lambda x: float('%0.2f' % x), duals))))
            cutstock_update_duals(gen_model, duals)
            gs = gen_model.solve(**kwargs)
            if not gs:
                print('{}> slave model fails, stop'.format(loop_count))
                break

            rc_cost = gen_model.objective_value
            if rc_cost <= -rc_eps:
                if verbose:
                    print('{}> slave model runs with obj={:g}'.format(loop_count, rc_cost))
            else:
                if verbose:
                    print('{}> pattern-generator model stops, obj={:g}'.format(loop_count, rc_cost))
                break

            use_values = gen_model.solution.get_values(gen_model.use_vars)
            if verbose:
                print('{}> add new pattern to master data: {}'.format(loop_count, str(use_values)))
            # make a new pattern with use values
            if not (loop_count < 100 and abs(best - curr) >= obj_eps):
                print('* terminating: best-curr={:g}'.format(abs(best - curr)))
                break
            add_pattern_to_master_model(master_model, use_values)

    ret = None
    if ms:
        if verbose:
            print('\n* Cutting-stock column generation terminates, best={:g}, #loops={}'.format(curr, loop_count))
            cutstock_print_solution(master_model)
        ret = ms
    else:
        print("!!!!  Cutting-stock column generation fails  !!!!")
        ret = None
    gen_model.end()

    return (master_model, ret)

def cutstock_solve_default(**kwargs):
    return cutstock_solve(DEFAULT_ITEMS, DEFAULT_PATTERNS, DEFAULT_PATTERN_ITEM_FILLED, DEFAULT_ROLL_WIDTH,
                          **kwargs)


# -----------------------------------------------------------------------------
# Solve the model and display the result
# -----------------------------------------------------------------------------
if __name__ == '__main__':
    m,s = cutstock_solve_default()
    assert abs(s.objective_value - 46.25) <= 0.1
    # Save the solution as "solution.json" program output.
    with get_environment().get_output_stream("solution.json") as fp:
        cutstock_save_as_json(m, fp)
    m.end()