minor bugfixes + improved error messages

src/baital.py

@@ -29,12 +29,14 @@
 import numpy as np
 import time
 import textwrap
+import utils
+import random
 
-def runPreprocess(inputfile, twise, outputdir, benchmarkName, strategy, isApprox, epsilon, delta):
+def runPreprocess(inputfile, twise, outputdir, benchmarkName, strategy, preprocessonly, isApprox, epsilon, delta):
 #    if strategy == 3:
 #        cnf_combinations.run(inputfile, twise, 2, outputdir, epsilon, delta)
 #        combinationsFile = os.path.join(outputdir, benchmarkName + '.count')
-    if (strategy == 1 or strategy == 3):
+    if (strategy == 1 or strategy == 3 or preprocessonly):
         if isApprox:
             cnf_combinations.run(inputfile, twise, 3, outputdir, epsilon, delta)
             combinationsFile = os.path.join(outputdir, benchmarkName + '_' + str(twise) + '.acomb')
@@ -72,7 +74,7 @@ def run(dimacscnf, strategy, twise, samples, descoverage, spr, rounds, outputdir
         ptime = -1
     else:
         pstart = time.time()
-        combinationsFile = runPreprocess(dimacscnf, twise, outputdir, benchmarkName, strategy, papprx, pepsilon, pdelta)
+        combinationsFile = runPreprocess(dimacscnf, twise, outputdir, benchmarkName, strategy, preprocessonly, papprx, pepsilon, pdelta)
         ptime = time.time() - pstart if combinationsFile != '' else -1
 
     if preprocessonly:
@@ -152,7 +154,7 @@ def check_rescombfile(combinations_file, twise):
         return len(line.split(",")) == twise
 
 epilog=textwrap.dedent('''\
-The tool performs 3 steps:
+The tool executes 3 steps:
     Step 1: preprocessing. Used for strategies 1, and 3. Has 2 options that are controlled by --strategy and --preprocess-approximate options
         (i) Lists the combinations of size <twise> allowed by cnf constraints (Could take hours for twise=2, infeasible for large models for twise=3). Results could be reused at step 3. 
             Precomputed file can be provided with --preprocess-file option, expected to have .comb extension.
@@ -199,7 +201,7 @@ def main():
 
     parser.add_argument("--weight-function", type=int, default=2, choices=range(1, 8), help="Function number between 1 and 7 for weight generation, used in strategies 1, 2, 3, and 4", dest='funcNumber')
     parser.add_argument("--no-sampling", action='store_true', help="skip step 2, computes only the coverage of a provided .sample file with --samples-file", dest='nosampling')
-    parser.add_argument("--samples-file", type=str, default='', help="file with samples to compute the coverage for --no-sampling option. Shall have .samples extension", dest="outersamplefile")
+    parser.add_argument("--samples-file", type=str, default='', help="file with samples to compute the coverage for --no-sampling option. Shall have .samples extension", dest="externalsamplefile")
 
     parser.add_argument("--maxcov-file", type=str, default='', help="file with pregenerated list of satisfiable feature combinations for the step 3. Shall have .comb extension", dest="rescombfile")
     parser.add_argument("--no-maxcov", action='store_true', help="Compute only number of combinatons in samples, instead of coverage", dest='nocomb')
@@ -235,10 +237,10 @@ def main():
         print("Delta and epsilon shall be in range (0,1)")
         sys.exit(1)
 
-    if args.nosampling and not args.outersamplefile:
-        print("--args.nosampling requires file with sample provided with --samples-file")
+    if args.nosampling and not args.externalsamplefile:
+        print("--args.nosampling must be used with --samples-file")
         sys.exit(1)
-        if args.outersamplefile[-8:] != '.samples':
+        if args.externalsamplefile[-8:] != '.samples':
             print("--samples-file is expected to have '.samples' extension")
             sys.exit(1)
         #TODO verify --samples-file format
@@ -254,7 +256,7 @@ def main():
         np.random.seed(args.seed)
         random.seed(args.seed)
 
-    run(args.DIMACSCNF, args.strategy, args.twise, args.samples, args.descoverage, args.spr, args.rounds, args.outputdir, args.outputfile, args.nocomb, args.combfile, args.papprx, args.pdelta, args.pepsilon, args.funcNumber, args.rescombfile, args.nosampling, args.outersamplefile, args.apprx, args.epsilon, args.delta, args.preprocess)
+    run(args.DIMACSCNF, args.strategy, args.twise, args.samples, args.descoverage, args.spr, args.rounds, args.outputdir, args.outputfile, args.nocomb, args.combfile, args.papprx, args.pdelta, args.pepsilon, args.funcNumber, args.rescombfile, args.nosampling, args.externalsamplefile, args.apprx, args.epsilon, args.delta, args.preprocess)
 
 if __name__== "__main__":
     main()

src/cnf_combinations.py

@@ -30,7 +30,7 @@
 from pathlib import Path
 import utils
 
-
+TMPCNFSUFFIX = '_tmp.cnf'
 
 ###combs is a list of precomputed combinations of smaller sizes: combs[x] contains combinatios of size x+1             
 def get_combinations(nVars, clauses, size, outputfile, combs):
@@ -199,25 +199,28 @@ def rmfile(filename):
 
 def runApproxmc(tmpCNF, epsilon, delta):
     approxOutput='out.pmc'
-    rmfile(approxOutput)
-
-    cmd = 'approxmc --seed ' + str(random.randint(1,10000)) + ' --epsilon ' + str(epsilon) + ' --delta ' + str(delta) + ' ' + tmpCNF + ' >' + approxOutput  
-    start = time.time()
-    os.system(cmd)
-    total_user_time = time.time() - start
-    result = -1
-    with open(approxOutput) as f:
-        for line in f:
-            if line.startswith('s mc'): #Note the version of ApproxMC
-                number= int(line.strip().split(' ')[2].strip())
-                result = number
-                break
-    return result
-
+    try:
+        cmd = 'approxmc --seed ' + str(random.randint(1,10000)) + ' --epsilon ' + str(epsilon) + ' --delta ' + str(delta) + ' ' + tmpCNF + ' >' + approxOutput  
+        start = time.time()
+        os.system(cmd)
+        total_user_time = time.time() - start
+        result = -1
+        with open(approxOutput) as f:
+            for line in f:
+                if line.startswith('s mc'): #Note the version of ApproxMC
+                    number= int(line.strip().split(' ')[2].strip())
+                    result = number
+                    break
+        os.remove(approxOutput)
+        return result
+    except:
+        print(f"Approxmc call failed.")
+        sys.exit(1)
 
 def approxComb(nVars, clauses, twise, tmpCNF, epsilon, delta):
     generateGFCNF(nVars, clauses, twise, tmpCNF)
     result = runApproxmc(tmpCNF, epsilon, delta)
+    os.remove(tmpCNF)
     return result
 
 def approxCombLiteral(nVars, clauses, twise, outputfile, tmpCNF, epsilon, delta):
@@ -237,6 +240,7 @@ def approxCombLiteral(nVars, clauses, twise, outputfile, tmpCNF, epsilon, delta)
         f.write(str(twise)+ '\n')
         for key in res.keys():
             f.write(str(key) + ' ' + str(res[key]) + '\n')
+    os.remove(tmpCNF)
     return res
 
 #-----------------------------------------------------------------------------------------------------
@@ -246,7 +250,7 @@ def run(dimacscnf, twise, mode, outputdir, epsilon, delta):
     benchmarkName = os.path.basename(dimacscnf).split('.')[0]
     if mode == 2:
         start_full = time.time()
-        res = approxComb(nVars, clauses, twise, benchmarkName + '_tmp.cnf', epsilon, delta)
+        res = approxComb(nVars, clauses, twise, benchmarkName + TMPCNFSUFFIX, epsilon, delta)
         print("Approximate number of combinations is " + str(res))
         print("Total time: " + str(time.time() - start_full))
 #    elif mode == 4:   #not used anymore
@@ -255,7 +259,7 @@ def run(dimacscnf, twise, mode, outputdir, epsilon, delta):
 #        print("Total time: " + str(time.time() - start_full))
     elif mode == 3:
         start_full = time.time()
-        res = approxCombLiteral(nVars, clauses, twise, os.path.join(outputdir,  benchmarkName + '_' + str(twise) + '.acomb'), benchmarkName + '_tmp.cnf', epsilon, delta)
+        res = approxCombLiteral(nVars, clauses, twise, os.path.join(outputdir,  benchmarkName + '_' + str(twise) + '.acomb'), benchmarkName + TMPCNFSUFFIX, epsilon, delta)
         print("Total time: " + str(time.time() - start_full))
     elif mode == 1:
         combs = []   
@@ -278,9 +282,9 @@ def run(dimacscnf, twise, mode, outputdir, epsilon, delta):
     parser.add_argument('DIMACSCNF', nargs='?', type=str, default="", help='input cnf file in dimacs format')
     mode = parser.add_mutually_exclusive_group()
     mode.add_argument('--combinations', action='store_true', help="counts number of combinations, this is the default mode", dest='m1')
-    mode.add_argument('--model-count', action='store_true', help="counts number of models for each literal", dest='m2')
-    mode.add_argument('--approximate', action='store_true', help="counts approximate number of combinations", dest='m3')
-    mode.add_argument('--approximate-literal', action='store_true', help="counts approximate number of combinations for each literal", dest='m4')
+    #mode.add_argument('--model-count', action='store_true', help="counts number of models for each literal", dest='m4')
+    mode.add_argument('--approximate', action='store_true', help="counts approximate number of combinations", dest='m2')
+    mode.add_argument('--approximate-literal', action='store_true', help="counts approximate number of combinations for each literal", dest='m3')
     parser.add_argument("--delta", type=float, default=0.05, help="Delta for approximate counting", dest='delta')
     parser.add_argument("--epsilon", type=float, default=0.05, help="Epsilon for approximate counting", dest='epsilon')
     parser.add_argument("--seed", type=int, help="Random seed. Unused if approximate is not selected", dest='seed')
@@ -297,8 +301,8 @@ def run(dimacscnf, twise, mode, outputdir, epsilon, delta):
         mode = 2
     elif args.m3:
         mode =3 
-    elif args.m4:
-        mode =4
+    #elif args.m4:
+    #    mode =4
 
     Path(args.outputdir).mkdir(parents=True, exist_ok=True)
     run(args.DIMACSCNF, args.size, mode, args.outputdir, args.epsilon, args.delta)

src/sampling.py

@@ -87,7 +87,7 @@ def loadApproxCount(combinationsFile):
     with open(combinationsFile) as f:
         lines = f.readlines()
     try:
-        for line in lines:
+        for line in lines[1:]:
             spl = line.strip().split(' ')
             res.update({int(spl[0]): int(spl[1])})
         return res
@@ -117,7 +117,7 @@ def loadMaxComb(nvars, twise, combinationsFile):
                 res[val] +=1
         return res
     except:
-        print(f"Wrong format of the file {combinationsFile}. Each line shall contain a comma separated combination of size {twise}")
+        print(f"Exception during the processing of {combinationsFile}. Each line shall contain a comma separated combination of size {twise}. Is {combinationsFile} corresponds to the input file?")
         sys.exit(1)
 
 # Attempts to get the total number of combinations or models from .count files.
@@ -207,30 +207,32 @@ def run(nSamples, rounds, dimacscnf, outputFile, twise, strategy, descoverage=No
             for i in range(len(newSamplesLines)):
                 if nSamples <0 or roundN != rounds-1 or roundN * samplesperround + i < nSamples: #ignore extra samples in case of non divisible by number of rounds
                     lineParts = newSamplesLines[i].strip().split(',')
-                    s = list(map(int, lineParts[1].strip().split(' ')))
-                    if strategy == 3 or strategy == 4:
-                        utils.updateBoxesCoverage(varsBoxes, twise, s)
-                    elif strategy == 5:
-                            for val in s:
-                                count[val] +=1
-                    elif strategy == 1 or strategy == 2:
-                        utils.getTuples_rec(s, twise, trie, count, [], True)
-                    else:
-                        print('Unknown strategy')
-                        sys.exit(1)
-                    # Copy samples to output file
+                    if roundN != rounds-1:
+                        s = list(map(int, lineParts[1].strip().split(' ')))
+                        if strategy == 3 or strategy == 4:
+                            utils.updateBoxesCoverage(varsBoxes, twise, s)
+                        elif strategy == 5:
+                                for val in s:
+                                    count[val] +=1
+                        elif strategy == 1 or strategy == 2:
+                            utils.getTuples_rec(s, twise, trie, count, [], True)
+                        else:
+                            print('Unknown strategy')
+                            sys.exit(1)
+                        # Copy samples to output file
                     sampleNumber = roundN * samplesperround + int(lineParts[0].strip())
                     output.write(str(sampleNumber) + ',' + lineParts[1] + '\n')
 
         os.remove(TMPSAMPLEFILE)
 
         # Update counter for strategies 3 and 4 - done once, not after each new sample
-        if strategy == 3 or strategy == 4:
-            count = boxesCoveragePerLiteral(varsBoxes, nBoxesLit, nvars, twise)
-            print("Current approximation of sampled combinations " + str(sum(count.values()) / twise) )
-
-        # Generate weights for next round
-        weightFunctions.generateWeights(count, nvars, strategy, maxComb, WEIGHTFILEPREF, roundN+2, funcNumber)
+        if roundN != rounds-1:
+            if strategy == 3 or strategy == 4:
+                count = boxesCoveragePerLiteral(varsBoxes, nBoxesLit, nvars, twise)
+                print("Current approximation of sampled combinations " + str(sum(count.values()) / twise) )
+
+            # Generate weights for next round
+            weightFunctions.generateWeights(count, nvars, strategy, maxComb, WEIGHTFILEPREF, roundN+2, funcNumber)
         print("The time taken by round " + str(roundN+1) + " :" +  str(time.time()-round_start))
         print("Round " + str(roundN+1) + ' finished...')
 
@@ -248,12 +250,14 @@ def run(nSamples, rounds, dimacscnf, outputFile, twise, strategy, descoverage=No
     output.close()
 
     print("The time taken by sampling:", time.time()-start)
-
     #cleanup
     os.remove(PICKLEFILE)
-    for roundN in range(rounds+1):
-        os.remove(WEIGHTFILEPREF + str(roundN+1)  + '.txt')
+    for rn in range(roundN+1):
+        os.remove(WEIGHTFILEPREF + str(rn+1)  + '.txt')
+    if descoverage and roundN < rounds-1:
+        os.remove(WEIGHTFILEPREF + str(roundN+2)  + '.txt')
 
+
 
 epilog=textwrap.dedent('''\
 Strategies information: strategies define what parameter is used to select weights for the next round. The parameter is computed for each literal. The following lists the parameter for each strategy.