Loop Time Measure

Introduction

loop_time_measure is a very simple Python thread safe library to measure the time spent in segments of code in single or repeated executions.

How it works

The meters are identified by an string and can be nested and/or interleaved. The time amounts spent between the start and the stop methods are added to the respective id. The getTime method can be used the get the current amount of time (or the list of times in "stats" mode) on a given id. The report method prints the amount of time of all ids.

Compatibility

• Python 3

Dependencies

• time
• threading
• sys
• numpy (or math)
• functools

Usage example:

As functions:

importloop_time_measureltm=loop_time_measure.loopTimeMeasureClass()
ltm.start('Total time')
foriinrange(200):
ltm.start('code A')
forjinrange(100):
print('Time consumming code A')
ltm.stop('code A')
ltm.start('code B')
forkinrange(50):
print('Time consumming code B')
ltm.stop('code B')
ltm.stop('Total time')
ltm.report()

Output:

Time consumming code A
Time consumming code A
(...)
Time consumming code A
(...)
Time consumming code B
Time consumming code B
(...)
Time consumming code B
-----------------
Time statistics:
-----------------
all:0.232278s
codeA:0.153208s
codeB:0.077236s

As decorator:

fromloop_time_measureimportmeasureFunctionTime, reportFunctionsTimes@measureFunctionTime(mode='stats') #the first decorator rules the parametersdefmyFunction():
print('Time consumming code')
@measureFunctionTime()defmyOtherFunction(text):
text=text[::-1] #bogus processingprint(text)
returntextforiinrange(25):
myFunction()
foriinrange(50):
res=myOtherFunction('Other text')
print("res =", res)
print("-----------------")
print("Time statistics:")
print("-----------------")
reportFunctionsTimes()

Output:

Time consumming code
Time consumming code
(...)
Time consumming code
(...)
txet rehtO
res = txet rehtO
txet rehtO
res = txet rehtO
(...)
txet rehtO
res = txet rehtO
-----------------
Time statistics:
-----------------
id, total_time, average_time, min_time, max_time, std_time, num_samples
myOtherFunction, 0.000592, 0.000012, 0.000011, 0.000014, 0.000001, 50
myFunction, 0.000308, 0.000012, 0.000010, 0.000041, 0.000006, 25

Documentation

Class

• loopTimeMeasureClass(timeFunction='process_time', mode = 'timeSum')

  Class for the time measurements. You need only one object for all meters.

  The timeFunction parameter selects the function used to get the time.

  • The default timeFunction is "process_time" ( time.process_time() ) and it measures CPU time.

  • The alternative is "time" ( time.time() ), for real time measurement.

  The mode parameter selects the time storing mode.

  • The default timeFunction is "timeSum". It keeps only the sum of times.

  • The alternative is "stats". It keeps a list with all time measurement and, on the report, the output is a CSV with several statistical data.

---

Methods

• start(id)

  Starts a meter.

---

• stop(id)

  Stop (or pause) a meter

---

• pause(id)

  Alias to stop

---

• reset(id)

  Reset (set to zero the time amount) a meter

---

• report([sort])

  Print a report of all meters. The time unit is seconds. The parameter sort indicates the sort criteria to be used to determine the meters print order. In "timeSum" mode, the default is "time", to sort by descending total time values. The alternative is "id", to sort by the id strings. In "stats" mode, the default is "time", to sort by descending total time values. The alternatives are:

  • "id" - Meter Id (or function name, in decorator mode);
  • "average" - Average of the registered times (between starts and stops);
  • "min" - Minimum registered time;
  • "max" - Maximum registered time;
  • "stdev" - Standard deviation of the registered times;
  • "num_samples" - Number of registered times.

Decorator

• @measureFunctionTime(timeFunction='process_time', mode = 'timeSum')

Wraps a function in a time measurement

The first decorator parameters are replicated to all the other decorators.

Functions

• reportFunctionsTimes(sort = 'time')

Report the measured times on all decorated functions

Author

• Sergio Schmiegelow
• sergio.schmiegelow@gmail.com