如何在Python中分析异常处理？

Question

Is there a way to find out how many exceptions are thrown (and caught) and maybe how much time exception handling took?

I'm using pyparsing (slightly modified) which relies heavily on exceptions and I want to find out (or at least estimate) if it would be worth the effort to rewrite it to work without exceptions.

Raising and catching of exceptions is widely scattered over the module so I'm searching a way which doesn't need to modify each try-except-block

Answer 1

Whenever you want to profile code, using a contextmanager is probably a good idea

import time
from contextlib import contextmanager

# global
NUMBER_OF_EXCEPTIONS = 0

# define the timer
class Timer(object):
    def __init__(self):
        self.t1 = time.time()
    def stop(self):
        self.t2 = time.time()
        self.elapsed = self.t2 - self.t1

# define the profiler
@contextmanager
def profiler():
    t = Timer()
    yield t
    t.stop()
    print("elapsed: ", t.elapsed)

# use the profiler!
with profiler():
    try:
        1/0
    except ZeroDivisionError:
        # handle exception
        NUMBER_OF_EXCEPTIONS += 1
        time.sleep(1.1)

# and use it again!
with profiler():
    try:
        1/2
    except ZeroDivisionError:
        # handle exception
        NUMBER_OF_EXCEPTIONS += 1
        time.sleep(1.1)

print("Total handled exceptions: ", NUMBER_OF_EXCEPTIONS)

The output should be something like this:

elapsed:  1.10120511055
elapsed:  4.05311584473e-06
Total handled exceptions:  1

Now if you want to get really fancy, you can write context managers for handling your exceptions too ( source ). This technique can lead to very clean code:

NUMBER_OF_EXCEPTIONS = 0

@contextmanager
def handle_zero_division_error():
    try:
        yield
    except ZeroDivisionError as err:
        global NUMBER_OF_EXCEPTIONS
        NUMBER_OF_EXCEPTIONS += 1
        time.sleep(1.1)

with profiler():
    with handle_zero_division_error():
        1/0

with profiler():
    with handle_zero_division_error():
        1/0

# you can even write it like this
with profiler(), handle_zero_division_error():
    1/2

print("Total handled exceptions: ", NUMBER_OF_EXCEPTIONS)

And the output:

elapsed:  1.10123705864
elapsed:  1.10085892677
elapsed:  1.90734863281e-05
Total handled exceptions:  2