Encapsulates the logic of a retry loop using a generator function.
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 | # Simple version:
import sys, time
class RetryError(Exception):
pass
def retryloop(attempts, timeout):
starttime = time.time()
success = set()
for i in range(attempts):
success.add(True)
yield success.clear
if success:
return
if starttime + timeout < time.time():
break
raise RetryError
"""
Usage:
for retry in retryloop(10, timeout=30):
try:
something
except SomeException:
retry()
for retry in retryloop(10, timeout=30):
something
if somecondition:
retry()
"""
# Fancy version:
def fancyretryloop(attempts, timeout=None, delay=0, backoff=1):
starttime = time.time()
success = set()
for i in range(attempts):
success.add(True)
yield success.clear
if success:
return
duration = time.time() - starttime
if timeout is not None and duration > timeout:
break
if delay:
time.sleep(delay)
delay = delay * backoff
e = sys.exc_info()[1]
# No pending exception? Make one
if e is None:
try: raise RetryError
except RetryError as e: pass
# Decorate exception with retry information:
e.args = e.args + ("on attempt {0} of {1} after {2:.3f} seconds".format(i, attempts + 1, duration),)
raise
|
The retry loop is common in any code that deals with resources that might fail intermittently such as calls to remote servers. This generator function takes away the noise of retry loop implementation and encapsulates it into a structure that almost looks like a language feature and should be easy to understand for a reader even without looking at the implementation of the retryloop() generator.
The fancy version adds delay and exponential backoff. It also preserves the original exception which terminated the retry loop, decorating it with useful information about the retry loop.
Caveat: Calling retry() does not terminate the current attempt. You should use "retry() ; continue" if you have multiple try/catch statements in order to prevent it from trying the next step of an attempt even if one step has failed.