Comments on: Python wanderings, part one http://www.tenshu.net/archives/2009/12/23/python-wanderings-part-one/ Pondering the mystery... Tue, 07 Sep 2010 21:09:35 +0000 http://wordpress.org/?v=2.9.2 hourly 1 By: cmsj http://www.tenshu.net/archives/2009/12/23/python-wanderings-part-one/comment-page-1/#comment-628 cmsj Fri, 25 Dec 2009 21:31:07 +0000 http://www.tenshu.net/?p=360#comment-628 Jamu: yeah I've cheated a bit with my testing for my Borg class, in that I haven't yet thought about how I'm going to fire off tests - right now I'm just abusing __main__ in .py files that otherwise don't get called directly to import doctest, do any setup and kick off the tests. I really appreciate you commenting with what you would do in this situation - that's exactly the kind of thing I want to come out of these posts. Professionals can come and tell anyone who reads this (me included) why I'm wrong :) Jamu: yeah I’ve cheated a bit with my testing for my Borg class, in that I haven’t yet thought about how I’m going to fire off tests – right now I’m just abusing __main__ in .py files that otherwise don’t get called directly to import doctest, do any setup and kick off the tests.

I really appreciate you commenting with what you would do in this situation – that’s exactly the kind of thing I want to come out of these posts. Professionals can come and tell anyone who reads this (me included) why I’m wrong :)

]]> By: Jamu Kakar http://www.tenshu.net/archives/2009/12/23/python-wanderings-part-one/comment-page-1/#comment-627 Jamu Kakar Fri, 25 Dec 2009 19:20:36 +0000 http://www.tenshu.net/?p=360#comment-627 The Borg pattern is cool, but it makes it very hard to create a fake instance of a Borg object for use in tests. Basically, it's a global variable in a jumpsuit. At first glance it seems to be *super cool*, but in the end you have basically the same set of problems as you do with global variables. In this kind of situation I would pass around object references (since you have only 4 core objects, the __init__ overhead isn't something worth working around) or I provide a function, such as get_terminator, that returns a singleton. It's not only in tests that this kind of thing matters. Consider a user that wants to replace Config with a version that backs onto SQLite, for example. This is quite hard with the Borg pattern, because it hard-codes dependencies. All that said, it is a cool pattern, and it does have a certain simplicity to it that is appealing. Thanks for the post, it's nice to read about the thought process you've gone through to get to a nice solution. The Borg pattern is cool, but it makes it very hard to create a fake instance of a Borg object for use in tests. Basically, it’s a global variable in a jumpsuit. At first glance it seems to be *super cool*, but in the end you have basically the same set of problems as you do with global variables. In this kind of situation I would pass around object references (since you have only 4 core objects, the __init__ overhead isn’t something worth working around) or I provide a function, such as get_terminator, that returns a singleton.

It’s not only in tests that this kind of thing matters. Consider a user that wants to replace Config with a version that backs onto SQLite, for example. This is quite hard with the Borg pattern, because it hard-codes dependencies. All that said, it is a cool pattern, and it does have a certain simplicity to it that is appealing.

Thanks for the post, it’s nice to read about the thought process you’ve gone through to get to a nice solution.

]]>