From 5c64035cc0543083a9803168eed48cbae2646326 Mon Sep 17 00:00:00 2001
From: Thomas Fetzer <thomas.fetzer@iws.uni-stuttgart.de>
Date: Thu, 26 Jan 2017 08:27:25 +0100
Subject: [PATCH] [bin/postprocessing] Add script to calculate l2error
E.g for two plot-over-lines.
---
bin/postprocessing/l2error.py | 73 +++++++++++++++++++++++++++++++++++
1 file changed, 73 insertions(+)
create mode 100644 bin/postprocessing/l2error.py
diff --git a/bin/postprocessing/l2error.py b/bin/postprocessing/l2error.py
new file mode 100644
index 0000000000..4f10fbd35b
--- /dev/null
+++ b/bin/postprocessing/l2error.py
@@ -0,0 +1,73 @@
+import argparse
+import csv
+import sys
+
+#Auxiliary function that provides a handy parser
+parser = argparse.ArgumentParser(prog='python ' + sys.argv[0], description='Calculate the l2 error of csv data files.')
+parser.add_argument('-f1', '--reference', type=str, required=True, help='Reference csv-file')
+parser.add_argument('-f2', '--newSolution', type=str, required=True, help='NewSolution csv-file')
+parser.add_argument('-xMin', '--xMin', type=float, required=False, default=-1e99, help='Restrict data to x>xMin')
+parser.add_argument('-xMax', '--xMax', type=float, required=False, default=1e99, help='Restrict data to x>xMax')
+parser.add_argument('-x1', '--xData1', type=int, required=True, help='Column index of x data in reference')
+parser.add_argument('-x2', '--xData2', type=int, required=True, help='Column index of x data in reference')
+parser.add_argument('-y1', '--yData1', type=int, required=True, help='Column index of y data in newSolution')
+parser.add_argument('-y2', '--yData2', type=int, required=True, help='Column index of y data in newSolution')
+parser.add_argument('-p', '--percent', action='store_true', help='Print errors in percent')
+parser.add_argument('-f', '--force', action='store_true', help='Ignore \'not-matching\' errors')
+parser.add_argument('-v', '--verbose', action='store_true', help='Verbosity of the script')
+args = vars(parser.parse_args())
+
+with open(args['reference'], 'rb') as referenceFile:
+ reader = csv.reader(referenceFile)
+ reference = list(reader)
+with open(args['newSolution'], 'rb') as newSolutionFile:
+ reader = csv.reader(newSolutionFile)
+ newSolution = list(reader)
+
+if (reference[0][args['xData1']] != newSolution[0][args['xData2']] and not args['force']):
+ print "X-Identifier not equal: ref=", reference[0][args['xData1']], ",new=", newSolution[0][args['xData2']], ". Aborting! (Use -f to continue anyway)"
+ exit (1)
+
+if (reference[0][args['yData1']] != newSolution[0][args['yData2']] and not args['force']):
+ print "Y-Identifier not equal. ref=", reference[0][args['yData1']], ",new=", newSolution[0][args['yData2']], ". Aborting! (Use -f to continue anyway)"
+ exit (2)
+
+if (len(reference) != len(newSolution)):
+ print "Length of reference and newSolution not equal: ref=", len(reference), ",new=", len(newSolution), ". Aborting!"
+ exit (3)
+
+distanceOld = 0.0
+sumError = 0.0
+sumReference = 0.0
+sumDistance = 0.0
+numPoints = 0
+for i in range(1,len(reference)):
+ coord_ref = float(reference[i][args['xData1']])
+ coord_newSolution = float(newSolution[i][args['xData2']])
+ if (coord_ref != coord_newSolution):
+ print "Coordinates not equal: ref=", coord_ref, ",new=", coord_newSolution, ". Aborting!"
+ exit (4)
+ if (coord_ref < float(args['xMin']) or coord_ref > float(args['xMax'])):
+ continue
+
+ if (i == 1):
+ distance = 0.5*(float(reference[2][args['xData1']]) - float(reference[1][args['xData1']]))
+ elif (i == len(reference)-1):
+ distance = 0.5*(float(reference[len(reference)-1][args['xData1']]) - float(reference[len(reference)-2][args['xData1']]))
+ else:
+ distance = 0.5*(float(reference[i+1][args['xData1']]) - float(reference[i-1][args['xData1']]))
+ sumError += ((float(reference[i][args['yData1']])-float(newSolution[i][args['yData2']]))**2)*distance
+ sumReference += ((float(reference[i][args['yData1']]))**2)*distance
+ sumDistance += distance
+ numPoints += 1
+
+if (numPoints < 999 and not args['force']):
+ print "Warning: numPoints=", numPoints, " is low, could result in bad the error approximation. (Use -f to suppress this warning)"
+
+l2normAbs = (sumError/sumDistance)**0.5 # numPoints is needed, resulting from the equidistant integration
+l2normRel = (sumError/sumReference)**0.5 # numPoints cancels out for equidistant integration
+
+if (args['percent']):
+ print "L2_Error_in_%: ", "{0:.5f}%".format(l2normAbs*100), "Rel_L2_Error_in_%: ", "{0:.5f}%".format(l2normRel*100)
+else:
+ print "L2_Error: ", "{0:.5e}".format(l2normAbs), " Rel_L2_Error: ", "{0:.5e}".format(l2normRel)
--
GitLab