Output path set for debugging

Write them in the c++ extension for each process, combine them in
python. For now.

fasttrips/FastTrips.py

@@ -14,6 +14,7 @@
 """
 import os
 from operator import attrgetter
+import pandas
 import transitfeed
 
 from .Assignment import Assignment
@@ -36,6 +37,9 @@ class FastTrips:
     #: Debug log filename.  Detailed output goes here, including trace information.
     DEBUG_LOG = "ft_debug%s.log"
 
+    #: Pathset debug filename.  Writes pathset here.
+    PATHSET_LOG = "ft_pathset%s.txt"
+
     def __init__(self, input_network_dir, input_demand_dir, output_dir,
                  is_child_process=False, logname_append="", appendLog=False):
         """
@@ -93,6 +97,15 @@ def __init__(self, input_network_dir, input_demand_dir, output_dir,
                      os.path.join(self.output_dir, FastTrips.DEBUG_LOG % logname_append),
                      logToConsole=False if is_child_process else True, append=appendLog)
 
+        # clear pathset files if we're starting out -- reset them to just a header
+        # there will be one for the parent process, and one each for workers
+        pathset_filename = os.path.join(self.output_dir, FastTrips.PATHSET_LOG % logname_append)
+        if not appendLog:
+            FastTripsLogger.info("Writing %s" % pathset_filename)
+            pathset_file = open(pathset_filename, 'w')
+            pathset_file.write("iteration passenger_id_num trip_list_id_num path_cost path_probability path_board_stops path_trips path_alight_stops\n")
+            pathset_file.close()
+
         # Read the configuration
         Assignment.read_configuration(self.input_network_dir, self.input_demand_dir)
 
@@ -143,6 +156,50 @@ def read_input_files(self):
         else:
             self.passengers = None
 
+    def combine_pathset_files(self):
+        """
+        Since the pathset files are output by worker, let's combine them into a single file.
+        """
+        procnum      = 1
+        pathset_init = False
+        pathsets_df  = None
+        # if we don't have one, no worries
+
+        while True:
+            logname_append = "_worker%02d" % procnum
+            pathset_filename = os.path.join(self.output_dir, FastTrips.PATHSET_LOG % logname_append)
+
+            if not os.path.exists(pathset_filename):
+                break
+
+            # read the pathset
+            pathset_df = pandas.read_table(pathset_filename, sep="[ ]+")
+            FastTripsLogger.info("Read %d lines from %s" % (len(pathset_df), pathset_filename))
+            # remove it
+            os.remove(pathset_filename)
+
+            # append to ours
+            if not pathset_init:
+                pathsets_df = pathset_df
+                pathset_init = True
+            else:
+                pathsets_df = pandas.concat([pathsets_df, pathset_df], axis=0)
+
+            # see if we have more
+            procnum += 1
+
+        if pathset_init:
+            # sort it by iteration, trip_id_num
+            pathsets_df.sort(columns=['iteration','trip_list_id_num'], inplace=True)
+            # write it
+            pathset_filename = os.path.join(self.output_dir, FastTrips.PATHSET_LOG % "")
+            pathsets_df.to_csv(pathset_filename, sep=" ", index=False)
+            FastTripsLogger.info("Wrote %d lines to %s" % (len(pathsets_df), pathset_filename))
+
+
     def run_assignment(self, output_dir):
         # Do it!
-        Assignment.assign_paths(output_dir, self);
+        Assignment.assign_paths(output_dir, self)
+
+        self.combine_pathset_files()
+

src/pathfinder.cpp

@@ -1750,6 +1750,18 @@ namespace fasttrips {
             }
             if (logsum == 0) { return false; } // fail
 
+            // debug -- print pet set to file
+            std::ofstream pathset_file;
+            std::ostringstream ss;
+            ss << output_dir_ << kPathSeparator;
+            ss << "ft_pathset";
+            if (process_num_ > 0) {
+                ss << "_worker" << std::setfill('0') << std::setw(2) <<  process_num_;
+            }
+            ss << ".txt";
+            // append
+            pathset_file.open(ss.str().c_str(), (std::ios_base::out | std::ios_base::app));
+
             // for integerized probability*1000000
             int cum_prob    = 0;
             int cost_cutoff = 1;
@@ -1775,7 +1787,18 @@ namespace fasttrips {
                     printPathCompat(trace_file, path_spec, paths_iter->first);
                     trace_file << std::endl;
                 }
+                // print path to pathset file
+                pathset_file << path_spec.iteration_ << " ";  // Iteration
+                pathset_file << path_spec.passenger_id_ << " ";         // The passenger ID
+                pathset_file << path_spec.path_id_ << " ";              // The path ID - uniquely identifies a passenger+path
+                pathset_file << std::setw(8) << std::fixed << std::setprecision(2) << paths_iter->second.cost_ << " ";
+                pathset_file << std::setw(8) << std::fixed << std::setprecision(6) << paths_iter->second.probability_ << " ";
+                printPathCompat(pathset_file, path_spec, paths_iter->first);
+                pathset_file << std::endl;
             }
+
+            pathset_file.close();
+
             if (cum_prob == 0) { return false; } // fail
 
             // choose path