ttimer

ttimer

ttimer is a simple timer that keeps track of the call hierarchy. ttimer is intended to fulfill a use case somewhere between a timer and a profiler: like a timer, it measures only where you explicitly write it, while like a profiler, it handles the call hierarchy and measures own time.

Install

> pip install ttimer

or simply copy whole ttimer.py into your project. (the license text is also included in ttimer.py)

Usage

>>> from time import sleep
>>> from ttimer import get_timer

>>> timer = get_timer(timer_name="my timer")

>>> with timer("a"):
>>>     sleep(0.1)
>>>     for _ in range(10):
>>>         with timer("b"):
>>>             sleep(0.01)
>>> with timer("b"):
>>>     sleep(0.1)

>>> print(timer.render())

path      count      time    own time    cpu time    own cpu time
------  -------  --------  ----------  ----------  --------------
a             1  0.203831    0.100531    0.001314        0.000784
└── b        10  0.103299    0.103299    0.00053         0.00053
b             1  0.103603    0.103603    8.2e-05         8.2e-05

>>> print(timer.render(flat=True))

path      count      time    own time    cpu time    own cpu time
------  -------  --------  ----------  ----------  --------------
a             1  0.203831    0.100531    0.001314        0.000784
b            11  0.206903    0.206903    0.000612        0.000612

ttimer records the following metrics in the with-statement:

• count: Call count.
• time: Elapsed time measured by perf_counter. It includes time elapsed during sleep and is system-wide.
• own time: Time, excluding the total time of its children.
• cpu time: CPU time measured by process_time.
• own cpu time: CPU time, excluding the total time of its children.

If the name is not passed in the with-statement, the name will be automatically resolved from the file and function names.

>>> from ttimer import get_timer

>>> t = get_timer(timer_name="foo")
>>> with t:
>>>     pass

>>> print(t.render())
path                                  count         time     own time    cpu time    own cpu time
----------------------------------  -------  -----------  -----------  ----------  --------------
test_get_timers(test_timer.py:144)        1  0.000347945  0.000347945    0.000228        0.000228

You can also use decorators instead of with-statement:

from ttimer import timer

@timer(timer_name="my timer")
def foo(a: int):
    pass

In both get_timer and @timer, timers with the same timer_name share the same elapsed time. This is useful when you want to measure times across modules.

All named timers are stored as a thread-local variable, and you can use get_timers to enumerate the stored timers.

>>> from ttimer import get_timer, get_timers

>>> with get_timer("foo"):
>>>    pass
>>> with get_timer("bar"):
>>>     pass

>>> all_timers = get_timers()
{'foo': <ttimer.timer.Timer object at 0x7fc9a334fc50>, 'bar': <ttimer.timer.Timer object at 0x7fc9a334df98>}

Local timers

If you do not prefer global (thread-local) variables, you can simply create a local Timer instance. In this style, if you use a decorator, you should pass the timer you created as an additional timer keyword argument.

from ttimer import Timer, timer

t = Timer()  # local timer

@timer()
def foo(a: int):
    pass

with t("a"):
    foo(a=1, timer=t)  # additional "timer" keyword argument are used to specify the context

Properties

By accessing timer[key], you can get the accumulated result as an instance of Record dataclass. You can of course also get a list of records by .records.

from dataclasses import asdict
from ttimer import get_timer

timer = get_timer("my timer")

with timer("a"):
    pass

print("result of {}:".format(timer["a"].name))
print("time:         {}".format(timer["a"].time))
print("cpu time:     {}".format(timer["a"].cpu_time))
print("own time:     {}".format(timer["a"].own_time))
print("own cpu time: {}".format(timer["a"].own_cpu_time))
print("count:        {}".format(timer["a"].count))

print(asdict(timer["a"]))  # result is dataclass

timer.records  # list of records

The results you can get with above are equivalent to flat=True: i.e., the measurements with the same name are accumulated. If you want to get the measurements for each call stack separately, you can use .nodes.

.nodes returns all the nodes, but if you want to get only the root node, use .trees. Both return an instance of Node, and you can access the child nodes with .children, or access the node's records with .record.

from ttimer import get_timer, Record

timer = get_timer("my timer")

with timer("a"):
    with timer("b"):
        pass
    
with timer("b"):
    pass

assert len(timer.records) == 2
assert len(timer.nodes) == 3
assert timer.nodes[1].stack == ("a", "b")
assert isinstance(timer.nodes[1].record, Record)
assert len(timer.trees) == 2