Parse perfcheck results

refactoring arguments
process csv file only if newer commits
add columned output
calculate deltas & alert messages

Change-Id: Ib7ba87e9cb55b03ac8665db7a8dc302d2e8611a0
Reviewed-on: https://gerrit.libreoffice.org/12155Reviewed-by: Matúš Kukan <matus.kukan@collabora.com>
Tested-by: Matúš Kukan <matus.kukan@collabora.com>

bin/parse-perfcheck.py

@@ -7,7 +7,6 @@
 
 import sys
 import os
-import time
 
 parseTrigger = "desc: Trigger: Client Request: "
 parseTotal = "totals: "
@@ -15,12 +14,16 @@ parseTotal = "totals: "
 separator = os.path.sep
 
 lastCommitId = ""
-needsCsvHeader = True
+lastCommitDate = ""
+needsCsvHeader = True # needs header in csv file ? yes if new
+
+colsResult = {}
+allTests = []
 
 def processDirectory(rootDir):
 
   if needsCsvHeader:
-    intermediateResult = "lastCommit\ttest name\tfiledatetime\tdump comment\tcount\n"
+    intermediateResult = "lastCommit\tlastCommitDate\ttest filename\tdump comment\tcount\n"
   else:
    intermediateResult = ""
 
@@ -40,86 +43,251 @@ def parseFile(dirname, filename):
   callgrindFile = open(path,'r')
   lines = callgrindFile.readlines()
 
-  message = ""
-  total = ""
+  curTestComment = ""
+  total = "0"
 
   for line in lines:
     if line.startswith(parseTrigger):
-      message = line[len(parseTrigger):]
+      curTestComment = line[len(parseTrigger):].replace("\n","")
     elif line.startswith(parseTotal):
-      total = line[len(parseTotal):]
+      total = line[len(parseTotal):].replace("\n","")
 
   callgrindFile.close()
 
-  if message == "" and total == "0\n":
+  if curTestComment == "":
+    return ""
+
+  if total == "0": # should not occur, btw
     return ""
 
   dirs = dirname.split(separator)
   currentTest = dirs[-1:]
   testName = currentTest[0].replace(".test.core","")
 
-  message = message.replace("\n","")
+  if lastCommitId not in colsResult:
+    colsResult[lastCommitId] = {}
+    colsResult[lastCommitId]['date'] = lastCommitDate
+    colsResult[lastCommitId]['values'] = {}
 
-  fileDate = time.strftime('%m/%d/%Y %H:%M:%S', time.gmtime(os.path.getmtime(path)))
+  colsResult[lastCommitId]['values'][curTestComment] = total
 
-  result = lastCommitId + "\t" + testName + "\t" + fileDate + "\t" + message + "\t" + total
+  result = lastCommitId + "\t" + lastCommitDate + "\t" + testName + "\t" + curTestComment + "\t" + total + "\n"
 
   return result
 
-def getLastCommitId():
+def getLastCommitInfo():
 
-  stream = os.popen("git log")
+  stream = os.popen("git log --date=iso")
+  line = stream.readline()
+  commitId = line.replace("commit ","").replace("\n","")
+  line = stream.readline()
   line = stream.readline()
-  return line.replace("commit ","").replace("\n","")
+  commitDate = line.replace("Date: ","").replace("\n","").strip()
+
+  return commitId, commitDate
 
 def displayUsage():
 
-  print
-  print "Parses the callgrind results of make percheck"
-  print
-  print "Usage: bin/parse_perfcheck.py [targetFileName = perfcheckResult.csv] [sourceDirectory = ./workdir/CppunitTest]"
-  print "default assumes running from core root directory"
-  print
+  usage = """
+
+Parses the callgrind results of make percheck
+
+Arguments :
+
+  --csv-file\t\t the target CSV file - new or containing previous tests - default : perfcheckResult.csv
+  --source-directory\t directory containing make perfcheck output - default : ./workdir/CppunitTest
+  --alert-type\t\t mode for calculating alerts - valid values : previous first
+  --alert-value\t\t alert threshold in % - default = 10
+
+  --help\t\t this message
+
+Columned output is dumped into csv-file + ".col"
+
+Alerts, if any, are displayed in standard output
+
+"""
+  print(usage)
+
+def analyzeArgs(args):
+
+    isValid = True
+
+    targetFileName = "perfcheckResult.csv"
+    sourceDirectory = "./workdir/CppunitTest"
+    alertType = ""
+    alertValue = 10
+
+    if "--help" in args:
+      isValid = False
+
+    if isValid:
+
+      for arg in args[1:]:
+
+        found = False
+
+        if arg.startswith("--csv-file"):
+          spliter = arg.split("=")
+          if spliter[1] != "":
+            targetFileName = spliter[1]
+            found = True
+
+        elif arg.startswith("--source-directory"):
+          spliter = arg.split("=")
+          if spliter[1] != "":
+            sourceDirectory = spliter[1]
+            found = True
+
+        elif arg.startswith("--alert-type"):
+          spliter = arg.split("=")
+          if spliter[1] in ['previous','first']:
+            alertType = spliter[1]
+            found = True
+          else:
+            isValid = False
+
+        elif arg.startswith("--alert-value"):
+          spliter = arg.split("=")
+          if spliter[1] != "":
+            alertValue = float(spliter[1])
+            found = True
+
+        isValid = isValid and found
+
+    return isValid, targetFileName, sourceDirectory, alertType, alertValue
+
+def readCsvFile():
+
+    fileResult = open(targetFileName,'r')
+    lines = fileResult.readlines()
+    fileResult.close
+
+    lines = lines[1:] #skip header
+
+    for line in lines:
+
+      if line.strip() != "": # do not process empty lines
+
+        spliter = line.replace('\n','').split('\t')
+        curId = spliter[0]
+        curDate = spliter[1]
+        curTestName = spliter[2]
+        curTestComment = spliter[3]
+        curValue = spliter[4]
+
+        if curTestComment not in allTests:
+          allTests.append(curTestComment)
+
+        if curId not in colsResult:
+          colsResult[curId] = {}
+          colsResult[curId]['date'] = curDate
+          colsResult[curId]['values'] = {}
+
+        colsResult[curId]['values'][curTestComment] = curValue
 
 if __name__ == '__main__':
 
   #check args
+  isOk, targetFileName, sourceDirectory, alertType, alertValue = analyzeArgs(sys.argv)
 
-  if len(sys.argv) < 3:
-      if len(sys.argv) == 2:
-          if sys.argv[1] == "--help":
-            displayUsage()
-            sys.exit(1)
-          else:
-            targetFileName = sys.argv[1]
-            sourceDirectory = "./workdir/CppunitTest"
-      elif len(sys.argv) == 1:
-          targetFileName = "perfcheckResult.csv"
-          sourceDirectory = "./workdir/CppunitTest"
-      else:
-          displayUsage()
-          sys.exit(1)
-  else:
-      targetFileName = sys.argv[1]
-      sourceDirectory = sys.argv[2]
+  if not isOk:
+    displayUsage()
+    sys.exit(1)
 
   # check if sourceDirectorty exists
   if not os.path.isdir(sourceDirectory):
-    print "sourceDirectorty %s not found - Aborting" % (sourceDirectory)
+    print("sourceDirectory %s not found - Aborting" % (sourceDirectory))
     sys.exit(1)
 
+  # read the complete CSV file
+  if os.path.isfile(targetFileName):
+    readCsvFile()
+    needsCsvHeader = False
+
   # last commit Id
-  lastCommitId = getLastCommitId()
+  lastCommitId, lastCommitDate = getLastCommitInfo()
+
+  # walker through directory
+  if not lastCommitId in colsResult:
+
+    newResult = processDirectory(sourceDirectory)
+
+    print('\nNew results\n' + newResult)
+
+    # append raw result
+    with open(targetFileName,'a') as fileResult:
+      fileResult.write(newResult)
+
+    print("\nCSV file written at " + targetFileName + '\n')
+
+  else:
+    print("\nCSV file up to date " + targetFileName + '\n')
+
+
+  # build columned output
+
+  # header
+  mLine = '\t'.join(["commit", "date"] + allTests) + '\n'
+
+  alertTest = {}
+
+  for k in colsResult:
+
+    mLine += k + "\t" + colsResult[k]['date'] + "\t"
+
+    for t in allTests:
+
+      if t in colsResult[k]['values']:
+        mValue= colsResult[k]['values'][t]
+
+        if not t in alertTest:
+          alertTest[t] = {}
+        alertTest[t][colsResult[k]['date']] = mValue
+
+      else:
+        mValue = ""
+
+      mLine += mValue + "\t"
+
+    mLine += "\n"
+
+  # write columned result
+    with open(targetFileName + '.col','w') as fileResult:
+      fileResult.write(mLine)
+
+  print("Columned file written at " + targetFileName + '.col\n')
+
+  # check for Alerts
+
+  if alertType == "":
+    sys.exit(1)
+
+  alertResult = ""
+
+  for t in alertTest:
+
+    testDict = alertTest[t]
+
+    # sort
+    keylist = sorted(testDict.keys())
+    maxVal = float(testDict[keylist[-1]])
+    minVal = 0
+
+    if alertType == "previous":
+      if len(keylist) > 1:
+        minVal = float(testDict[keylist[-2]])
 
-  # needs header in csv file ?
-  needsCsvHeader = not os.path.isfile(targetFileName)
+    else:
+      minVal = float(testDict[keylist[0]])
 
-  # call walker
-  globalResult = processDirectory(sourceDirectory)
+    if minVal != 0:
+      delta = 100 * ((maxVal-minVal)/minVal)
+    else:
+      delta = 0
 
-  print globalResult
+    if delta > float(alertValue):
+      alertResult +=  t + "\t" + "{:.2f}".format(delta) + " %\n"
 
-  # write result
-  fileResult = open(targetFileName,'a')
-  fileResult.write(globalResult)
-  fileResult.close()
+  if alertResult != "":
+    print("!!!!!!!! ALERT !!!!!!!\n")
+    print(alertResult)