1"Scripts for generating, solving and sweeping programs" 

2from time import time 

3import numpy as np 

4from ad import adnumber 

5from ..nomials import parse_subs 

6from ..solution_array import SolutionArray 

7from ..keydict import KeyDict 

8from ..small_scripts import maybe_flatten 

9from ..exceptions import Infeasible 

10 

11 

12def evaluate_linked(constants, linked): 

13 "Evaluates the values and gradients of linked variables." 

14 kdc = KeyDict({k: adnumber(maybe_flatten(v), k) 

15 for k, v in constants.items()}) 

16 kdc_plain = None 

17 array_calulated = {} 

18 for key in constants: # remove gradients from constants 

19 if key.gradients: 

20 del key.descr["gradients"] 

21 for v, f in linked.items(): 

22 try: 

23 if v.veckey and v.veckey.original_fn: 

24 if v.veckey not in array_calulated: 

25 ofn = v.veckey.original_fn 

26 array_calulated[v.veckey] = np.array(ofn(kdc)) 

27 out = array_calulated[v.veckey][v.idx] 

28 else: 

29 out = f(kdc) 

30 constants[v] = out.x 

31 v.descr["gradients"] = {adn.tag: grad 

32 for adn, grad in out.d().items() 

33 if adn.tag} # else it's user-created 

34 except Exception as exception: # pylint: disable=broad-except 

35 from .. import settings 

36 if settings.get("ad_errors_raise", None): 

37 raise 

38 print("Warning: skipped auto-differentiation of linked variable" 

39 " %s because %s was raised. Set `gpkit.settings" 

40 "[\"ad_errors_raise\"] = True` to raise such Exceptions" 

41 " directly.\n" % (v, repr(exception))) 

42 if kdc_plain is None: 

43 kdc_plain = KeyDict(constants) 

44 constants[v] = f(kdc_plain) 

45 v.descr.pop("gradients", None) 

46 

47 

48def progify(program, return_attr=None): 

49 """Generates function that returns a program() and optionally an attribute. 

50 

51 Arguments 

52 --------- 

53 program: NomialData 

54 Class to return, e.g. GeometricProgram or SequentialGeometricProgram 

55 return_attr: string 

56 attribute to return in addition to the program 

57 """ 

58 def programfn(self, constants=None, **initargs): 

59 "Return program version of self" 

60 if not constants: 

61 constants, _, linked = parse_subs(self.varkeys, self.substitutions) 

62 if linked: 

63 evaluate_linked(constants, linked) 

64 prog = program(self.cost, self, constants, **initargs) 

65 prog.model = self # NOTE SIDE EFFECTS 

66 if return_attr: 

67 return prog, getattr(prog, return_attr) 

68 return prog 

69 return programfn 

70 

71 

72def solvify(genfunction): 

73 "Returns function for making/solving/sweeping a program." 

74 def solvefn(self, solver=None, *, verbosity=1, skipsweepfailures=False, 

75 **solveargs): 

76 """Forms a mathematical program and attempts to solve it. 

77 

78 Arguments 

79 --------- 

80 solver : string or function (default None) 

81 If None, uses the default solver found in installation. 

82 verbosity : int (default 1) 

83 If greater than 0 prints runtime messages. 

84 Is decremented by one and then passed to programs. 

85 skipsweepfailures : bool (default False) 

86 If True, when a solve errors during a sweep, skip it. 

87 **solveargs : Passed to solve() call 

88 

89 Returns 

90 ------- 

91 sol : SolutionArray 

92 See the SolutionArray documentation for details. 

93 

94 Raises 

95 ------ 

96 ValueError if the program is invalid. 

97 RuntimeWarning if an error occurs in solving or parsing the solution. 

98 """ 

99 constants, sweep, linked = parse_subs(self.varkeys, self.substitutions) 

100 solution = SolutionArray() 

101 solution.modelstr = str(self) 

102 

103 # NOTE SIDE EFFECTS: self.program and self.solution set below 

104 if sweep: 

105 run_sweep(genfunction, self, solution, skipsweepfailures, 

106 constants, sweep, linked, solver, verbosity, **solveargs) 

107 else: 

108 self.program, progsolve = genfunction(self) 

109 result = progsolve(solver, verbosity=verbosity, **solveargs) 

110 if solveargs.get("process_result", True): 

111 self.process_result(result) 

112 solution.append(result) 

113 solution.to_arrays() 

114 self.solution = solution 

115 return solution 

116 return solvefn 

117 

118 

119# pylint: disable=too-many-locals,too-many-arguments,too-many-branches 

120def run_sweep(genfunction, self, solution, skipsweepfailures, 

121 constants, sweep, linked, solver, verbosity, **solveargs): 

122 "Runs through a sweep." 

123 # sort sweeps by the eqstr of their varkey 

124 sweepvars, sweepvals = zip(*sorted(list(sweep.items()), 

125 key=lambda vkval: vkval[0].eqstr)) 

126 if len(sweep) == 1: 

127 sweep_grids = np.array(list(sweepvals)) 

128 else: 

129 sweep_grids = np.meshgrid(*list(sweepvals)) 

130 

131 N_passes = sweep_grids[0].size 

132 sweep_vects = {var: grid.reshape(N_passes) 

133 for (var, grid) in zip(sweepvars, sweep_grids)} 

134 

135 if verbosity > 0: 

136 print("Sweeping with %i solves:" % N_passes) 

137 tic = time() 

138 

139 self.program = [] 

140 last_error = None 

141 for i in range(N_passes): 

142 constants.update({var: sweep_vect[i] 

143 for (var, sweep_vect) in sweep_vects.items()}) 

144 if linked: 

145 evaluate_linked(constants, linked) 

146 program, solvefn = genfunction(self, constants) 

147 self.program.append(program) # NOTE: SIDE EFFECTS 

148 try: 

149 if verbosity > 1: 

150 print("\nSolve %i:" % i) 

151 result = solvefn(solver, verbosity=verbosity-1, **solveargs) 

152 if solveargs.get("process_result", True): 

153 self.process_result(result) 

154 solution.append(result) 

155 except Infeasible as e: 

156 last_error = e 

157 if not skipsweepfailures: 

158 raise RuntimeWarning( 

159 "Solve %i was infeasible; progress saved to m.program." 

160 " To continue sweeping after failures, solve with" 

161 " skipsweepfailures=True." % i) from e 

162 if verbosity > 0: 

163 print("Solve %i was %s." % (i, e.__class__.__name__)) 

164 if not solution: 

165 raise RuntimeWarning("All solves were infeasible.") from last_error 

166 

167 solution["sweepvariables"] = KeyDict() 

168 ksweep = KeyDict(sweep) 

169 for var, val in list(solution["constants"].items()): 

170 if var in ksweep: 

171 solution["sweepvariables"][var] = val 

172 del solution["constants"][var] 

173 elif linked: # if any variables are linked, we check all of them 

174 if hasattr(val[0], "shape"): 

175 differences = ((l != val[0]).any() for l in val[1:]) 

176 else: 

177 differences = (l != val[0] for l in val[1:]) 

178 if not any(differences): 

179 solution["constants"][var] = [val[0]] 

180 else: 

181 solution["constants"][var] = [val[0]] 

182 

183 if verbosity > 0: 

184 soltime = time() - tic 

185 print("Sweeping took %.3g seconds." % (soltime,))