Hiro - time manipulation utilities for testing in python#

Often testing code that can be time dependent can become either fragile or slow. Hiro provides context managers and utilities to either freeze, accelerate or decelerate and jump between different points in time. Functions exposed by the standard library’s time, datetime and date modules are patched within the contexts exposed.

Hiro context manager and utilities#

Timeline context#

The Timeline context manager hijacks a few commonly used time functions to allow time manipulation within its context. Specifically sleep(), time(), time.time_ns(), monotonic(), time.monotonic_ns(), time.localtime(), gmtime(), datetime.datetime.now(), datetime.datetime.utcnow() and datetime.datetime.today() behave according the configuration of the context.

The context provides the following manipulation options:

  • rewind(): accepts seconds as an integer or an timedelta instance.

  • forward(): accepts seconds as an integer or an timedelta instance.

  • freeze(): accepts a floating point time since epoch or a datetime or date instance to freeze the time at.

  • unfreeze(): resumes time from the point it was frozen at.

  • scale(): accepts a floating point to accelerate/decelerate time by. > 1 = acceleration,  < 1 = deceleration

  • reset(): resets all time alterations.

import hiro
from datetime import timedelta, datetime
import time

datetime.now().isoformat()
# OUT: '2013-12-01T06:55:41.706060'
with hiro.Timeline() as timeline:

    # forward by an hour
    timeline.forward(60*60)
    datetime.now().isoformat()
    # OUT: '2013-12-01T07:55:41.707383'

    # jump forward by 10 minutes
    timeline.forward(timedelta(minutes=10))
    datetime.now().isoformat()
    # OUT: '2013-12-01T08:05:41.707425'

    # jump to yesterday and freeze
    timeline.freeze(datetime.now() - timedelta(hours=24))
    datetime.now().isoformat()
    # OUT: '2013-11-30T09:15:41'

    timeline.scale(5) # scale time by 5x
    time.sleep(5) # this will effectively only sleep for 1 second

    # since time is frozen the sleep has no effect
    datetime.now().isoformat()
    # OUT: '2013-11-30T09:15:41'

    timeline.rewind(timedelta(days=365))

    datetime.now().isoformat()
    # OUT: '2012-11-30T09:15:41'

To reduce the amount of statements inside the context, certain timeline setup tasks can be done via the constructor and/or by using the fluent interface.

import hiro
import time
from datetime import timedelta, datetime

start_point = datetime(2012,12,12,0,0,0)
my_timeline = hiro.Timeline(scale=5).forward(60*60).freeze()
with my_timeline as timeline:
    print datetime.now()
    # OUT: '2012-12-12 01:00:00.000315'
    time.sleep(5) # effectively 1 second
    # no effect as time is frozen
    datetime.now()
    # OUT: '2012-12-12 01:00:00.000315'
    timeline.unfreeze()
    # back to starting point
    datetime.now()
    # OUT: '2012-12-12 01:00:00.000317'
    time.sleep(5) # effectively 1 second
    # takes effect (+5 seconds)
    datetime.now()
    # OUT: '2012-12-12 01:00:05.003100'

Timeline can additionally be used as a decorator. If the decorated function expects a timeline argument, the Timeline will be passed to it.

import hiro
import time, datetime

@hiro.Timeline(scale=50000)
def sleeper():
    datetime.datetime.now()
    # OUT: '2013-11-30 14:27:43.409291'
    time.sleep(60*60) # effectively 72 ms
    datetime.datetime.now()
    # OUT: '2013-11-30 15:28:36.240675'

@hiro.Timeline()
def sleeper_aware(timeline):
    datetime.datetime.now()
    # OUT: '2013-11-30 14:27:43.409291'
    timeline.forward(60*60)
    datetime.datetime.now()
    # OUT: '2013-11-30 15:28:36.240675'

run_sync and run_async#

In order to execute certain callables within a Timeline context, two shortcut functions are provided.

Both functions return a ScaledRunner object which provides the following methods

  • get_execution_time(): The actual execution time of the callable

  • get_response() (will either return the actual return value of callable or raise the exception that was thrown)

run_async() returns a derived class of hiro.core.ScaledRunner that additionally provides the following methods

  • is_running(): True/False depending on whether the callable has completed execution

  • join(): blocks until the callable completes execution

import hiro
import time

def _slow_function(n):
    time.sleep(n)
    if n > 10:
        raise RuntimeError()
    return n

runner = hiro.run_sync(10, _slow_function, 10)
runner.get_response()
# OUT: 10

# due to the scale factor 10 it only took 1s to execute
runner.get_execution_time()
# OUT: 1.1052658557891846

runner = hiro.run_async(10, _slow_function, 11)
runner.is_running()
# OUT: True
runner.join()
runner.get_execution_time()
# OUT: 1.1052658557891846
runner.get_response()
# OUT: Traceback (most recent call last):
# ....
# OUT:   File "<input>", line 4, in _slow_function
# OUT: RuntimeError