Merge lp://staging/~mascha/beeseek/fixesLP505596 into lp://staging/beeseek/1.0

Fixes lp #505596

Hi and thanks again for your contribution.

Your branch will work and your patch does its job correctly, but I think that using a regular expression is too much complicated for this bug. I think that the better way to fix it is catching ValueError. The reason why I think this is that catching an exception is faster than compiling a matching a regular expression.

Here's what `timeit` says on my computer:

>>> timeit("""
... try:
... index = int(value)
... except ValueError:
... index = 0
... """, "value = '123'")
0.83609890937805176
>>> timeit("""
... if re.match('\d+$', value):
... index = int(value)
... else:
... index = 0
... """, "import re; value = '123'")
3.7652208805084229

In this case, where the index given by the user is a valid number, the difference is big (~2.929 seconds).

>>> timeit("""
... try:
... index = int(value)
... except ValueError:
... index = 0
... """, "value = 'abc'")
3.6588859558105469
>>> timeit("""
... if re.match('\d+$', value):
... index = int(value)
... else:
... index = 0
... """, "import re; value = 'abc'")
2.3073050975799561

Here the index is not valid, and the faster code is the one that uses the regex (~1.351 seconds). However: 1. the delta is smaller than the first case and 2. this case is more rare (if a non-digit is added to the URL, it's because the user did it intentionally).

Of course, this is my opinion and you are free to reply if you thing I'm wrong or if I missed something.

Fix performance concerns by pre compiling the regular expression.

Pre-compiling the regular expression gives a great advantage:

>>> timeit("""
... if NUMBER_MATCH.match(value):
... index = int(value)
... else:
... index = 0
... """, "import re; NUMBER_MATCH = re.compile(r'^\d+$'); value = '123'")
1.5664610862731934
>>> timeit("""
... if NUMBER_MATCH.match(value):
... index = int(value)
... else:
... index = 0
... """, "import re; NUMBER_MATCH = re.compile(r'^\d+$'); value = 'abc'")
0.54535198211669922

It's still slower than the try..except block, but the difference is not as big as before.
Thank you for your fix!

Thank you for accepting. The reason why I prefer the solution with the regular expression is, that now the time for the worst case is much lower.