** UnhandledExceptionEventHandler :: OFFICIAL LC / CHATTER THREAD **
04-30-2013
, 05:07 PM
04-30-2013
, 05:18 PM
04-30-2013
, 05:51 PM
oh for sure. haven't you noticed a massive uptick in the pussy flowing your way?
05-01-2013
, 03:49 AM
Thanks for the tip though.
05-01-2013
, 11:17 PM
our code base at work has a lot of tests, but many of them are not very good. for example:
on its face, this looks kind of dumb. why are we testing urlopen? isn't that a built-in module? regardless, we're not testing it very well since all we're doing is asserting that calling it on a uri returns something truthy.
but of course, it gets dumber. for one thing, this is testing a wrapper around urlopen (util.urlopen.urlopen ldo), which looks like this:
as you can see, this wrapper adds a ton of value: it parses a few pieces out of the exception objet, then throws the rest away while helpfully coercing the original exception into an IOError.
the real magnum opus, though, is this function:
here is the current landing page for www.random.com:
coworker: so we're just generating revenue for some squatter?
me: hundreds of times per day since early 2011
this same guy also wrote this test, which was already infamous:
when one of the ops guys found about this, he promptly registered nobodywilleverregisterthisdomainbecauseitwouldberi diculoustodoso.com. so the author changed the test:
Code:
def test_correctly_loads_uri(self):
"""Function should load URI correctly"""
uri_reference = self.get_uri_reference()
assert urlopen(uri_reference) # should produce no errors
but of course, it gets dumber. for one thing, this is testing a wrapper around urlopen (util.urlopen.urlopen ldo), which looks like this:
Code:
import urllib2
def urlopen(uri, data=None, *args, **kwargs):
"""Return a more verbose error report and raises IOError"""
uri_component = 'URI: %s\nData: %s\n' % (uri, data)
error_string = None
try:
response = urllib2.urlopen(uri, data, *args, **kwargs)
return response
except urllib2.HTTPError, e:
error_string = 'Server failed to handle request.\nCode: %s\n' % e.code
message = error_string + uri_component
raise IOError(message)
except urllib2.URLError, e:
error_string = 'Unable to load from URI. \nReason: %s\n' % e.reason message = error_string + uri_component
raise IOError(message)
the real magnum opus, though, is this function:
Code:
def get_uri_reference(self):
uri_reference = 'http://www.random.com'
return uri_reference
coworker: so we're just generating revenue for some squatter?
me: hundreds of times per day since early 2011
this same guy also wrote this test, which was already infamous:
Code:
def test_URLError(self):
"""Function should reraise URLError """
URI = 'http://www.nobodywilleverregisterthisdomainbecauseitwouldberidiculoustodoso.com/'
assert_raises(IOError, urlopen, URI)
Code:
def test_URLError(self):
"""Function should reraise URLError """
URI = 'http://www.nobodywilleverregisterthisdomainbecauseitwouldberidiculoustodosounlesstheirnameis<name of ops guy>.com/'
assert_raises(IOError, urlopen, URI)
05-01-2013
, 11:33 PM
-- irreproachable way to spend $10
05-02-2013
, 10:53 PM
05-03-2013
, 12:35 AM
I haven't benchmarked hhvm yet, but I think on a cold read a 4-6x improvement is not unreasonable with an order of magnitude improvement being likely for most CPU-intensive apps.
If I improved my functional-style programming and multithreading code writing skills, I bet the gap would shrink and regular PHP+APC would not be as slow. Still I don't think it would ever be worse than 4x.
Quote:
Yeah.. problem solved on my end. Also, you stole my theme.
Just picked the least stupid clean one from WP themes.
