+ got rid of some legacy code
[qpalma.git] / scripts / qpalma_pipeline.py
index 17d64e0..ac7194e 100644 (file)
 # This file contains the main interface to the QPalma pipeline.
 #
 
-from optparse import OptionParser
+import os
+import os.path
+import pdb
+import sys
 
 from qpalma.gridtools import ApproximationTask,PreprocessingTask
 from qpalma.gridtools import AlignmentTask,PostprocessingTask
 
-def create_option_parser():
-   parser = OptionParser()
+from qpalma.DatasetUtils import generatePredictionDataset,generateTrainingDataset
 
-   #  
-   parser.add_option("-ci", "--check_and_init", help="check configuration and initialize directories")
+from qpalma.SettingsParser import parseSettings
 
-   #
-   parser.add_option("-r", "--run", help="write report to FILE", metavar="FILE")
+from qpalma.utils import logwrite
 
-   #
-   parser.add_option("-xx", "--clear", action="store_false", dest="verbose", help="cleanup directories delete all created data")
-
-   return parser
-
-global_settings = {\
-'experiment_dir':'/fml/ag-raetsch/...',\
-'read_ascii_data_fn':'/fml/ag-raetsch/...',\
-'num_splits':50
-'global_log_fn':'~/qpalma.log'
-}
 
+Errormsg = """Usage is: python qpalma_pipeline.py predict|train <config filename>"""
 
 
 class System:
@@ -53,11 +43,15 @@ class System:
 
    """
 
-   def __init__(self):
+   def __init__(self,filename):
       """
+      Inititalize the system by loading and parsing the settings file to obtain
+      all parameters.
       """
-      parser = create_option_parser()
-      (options, args) = parser.parse_args()
+
+      self.settings = parseSettings(filename)
+      logwrite('Parsed settings system set up.',self.settings)
+
 
    def training(self):
       """
@@ -65,13 +59,29 @@ class System:
       converts the data to the right format needed by QPalma for the training
       algorithm.
       """
+      logwrite('Begin of training.\n',self.settings)
+
+      print '#'*80
+      print '\t\t\tStarting approximation...\n'
+      print '#'*80
 
-      pre_task = TrainingPreprocessingTask(global_settings,run_specific_settings)
+      # 
+      pre_task = TrainingPreprocessingTask(self.settings)
       pre_task.createJobs()
       pre_task.submit() 
-      while pre_task.checkIfTaskFinished() == False:
-         sleep(20)
-      
+      pre_task.checkIfTaskFinished()
+
+      # Collect the data and create a pickled training set
+      generateTrainingDataset(self.settings)
+
+      # Now that we have a dataset we can perform training
+      train_task = TrainingTask(self.settings)
+      train_task.CreateJobs()
+      train_task.Submit()
+      train_task.CheckIfTaskFinished()
+
+      logwrite('End of training.\n',self.settings)
+
 
    def prediction(self):
       """
@@ -80,44 +90,65 @@ class System:
       generate several output files containing the spliced alignments
       """
 
+      logwrite('Begin of prediction.\n',self.settings)
+
+      print '#'*80
+      print '\t\t\tStarting approximation...\n'
+      print '#'*80
+
       # Before creating a candidate spliced read dataset we have to first filter
       # the matches from the first seed finding run.
 
-      approx_task = ApproximationTask(config_obj)
-      approx_task.createJobs()
-      approx_task.submit()
-      approx_task.checkIfTaskFinished()
+      #approx_task = ApproximationTask(self.settings)
+      #approx_task.CreateJobs()
+      #approx_task.Submit()
+      #approx_task.CheckIfTaskFinished()
       
       # After filtering combine the filtered matches from the first run and the
       # found matches from the second run to a full dataset
 
-      pre_task = PreprocessingTask(...)
-      pre_task.createJobs()
-      pre_task.submit() 
-      while pre_task.checkIfTaskFinished() == False:
-         sleep(20)
+      print '#'*80
+      print '\t\t\tStarting dataset generation...\n'
+      print '#'*80
+
+      generatePredictionDataset(self.settings)
 
-      # Now that we have a dataset we can perform the accurate alignments for this
-      # data
+      print '#'*80
+      print '\t\t\tStarting alignments...\n'
+      print '#'*80
 
-      align_task = AlignmentTask(...)
-      align_task.createJobs()
-      align_task.submit()
-      while align_task.checkIfTaskFinished() == False:
-         sleep(20)
+      # Now that we have a dataset we can perform accurate alignments
+      align_task = AlignmentTask(self.settings)
+      align_task.CreateJobs()
+      align_task.Submit()
+      align_task.CheckIfTaskFinished()
+
+      print '#'*80
+      print '\t\t\tPostprocessing...\n'
+      print '#'*80
 
       # The results of the above alignment step can be converted to a data format
       # needed for further postprocessing.
+      post_task = PostprocessingTask(self.settings)
+      post_task.CreateJobs()
+      post_task.Submit()
+      post_task.CheckIfTaskFinished()
 
-      post_task = PostprocessingTask(...)
-      post_task.createJobs()
-      post_task.submit()
-      while post_task.checkIfTaskFinished() == False:
-         sleep(20)
-
-      print "Success!"
+      logwrite('End of prediction.\n',self.settings)
    
 
 if __name__ == '__main__':
-   system_obj = System() 
-   system_obj.run()
+   mode     = sys.argv[1]
+   assert mode in ['predict','train'], Errormsg
+   filename = sys.argv[2]
+   assert os.path.exists(filename), Errormsg
+
+   # creating system object
+   system_obj = System(filename)
+
+   if mode == 'predict':
+      system_obj.prediction()
+   elif mode == 'train':
+      system_obj.training()
+   else:
+      assert False