My Pittsburgh Python talk: "Testing with isolation: concepts and examples using the Python standard library `mock`"
A week ago, I saw a message on the Pittsburgh Python mailing list:
We are looking for a presenter to talk about mock, pymox or other Python testing topics for our next meeting. If you have a different topic that you would like to present on, please let us know about that too, though preference will be given to testing-related talks this month.
I fairly quickly volunteered to talk about the Python standard library
mock, despite never having used it!
Here’s a brief report of the Pittsburgh Python meeting in Google Pittsburgh (along with links to slides).
Slides are here.
Why did I volunteer?
When I volunteered to give my last talk, a talk at Pittsburgh Ruby, the reason was that I had a specific topic of my choosing on which I already had an informed opinion, and my goal was to explain it.
In stark contrast to that situation, I volunteered to give a talk on testing with
mock in Python while not having ever used it, and in fact not ever having used mock objects at all in my own unit testing, although I knew the basic concepts behind mocks.
So I was volunteering in this case in order to force myself to learn something in more depth.
Preparing for this talk was quite challenging for me. (Hmm, I have a feeling I’ll always feel this way for the rest of my talks in my life!?)
Deciding to suddenly prepare a talk threw a mini-bomb into my schedule for the following week, because I had many other things to work on. However, I thought it was a great opportunity to investigate something in depth that I had already wanted to for some time, so I put some other things aside. Sometimes one has to be flexible in the face of a good opportunity. I care a huge deal about the topic of software testing, so I felt I could not pass up the opportunity to share some ideas with an audience.
A huge number of people (over forty) quickly signed up to attend the Pittsburgh Python meeting, during which two talks were scheduled: one on Selenium WebDriver by Steve Gross, and one by me. Wow. I don’t know what’s causing this explosion in the size of the Pittsburgh Python community, but that’s great!
I made a very quick decision not to focus on all the features of the
mock library. That’s not a talk that I would myself go to. I put myself in the shoes of the forty participants. Obviously there was going to be a wide variety of background knowledge and experience about testing in general as well as about mocks.
In fact, Abby was going to come along with me to attend the Pittsburgh Python meeting. She just started learning computer programming a couple of weeks ago, using Python as her first programming language, and Coursera as the primary vehicle of instruction. I hoped to give a talk that she would get something usefully high-level and conceptual out of.
I figured on speaking for no more than half an hour.
So I wanted to focus on a clean, correct review (or introduction, as the case was surely going to be for some of the participants) of unit testing. I deliberately took a test-driven development (TDD) perspective as I designed some easy-to-understand examples of the entire process of writing a test, making it fail, then writing the implementation, then running the test again, etc. I felt that TDD was the ideal way to discuss the notion of testing with isolation, because if collaborating code for a system under test (SUT) does not yet exist at all, it becomes quite clear where the possible isolation boundaries are.
I also wanted to use this example as a running example in order to introduce and explore questions of stubbing, or more generally, test doubles. In particular, I brought up the ongoing controversy over the use of mocks in testing at all, framed around the pretty fair analysis by Martin Fowler. In order to be able to explain the controversy, I developed my running example to show a way of refactoring code in order to test without mocking, as well as showing how one could do it with mocking. And so I only explained enough features of Python
mock to enable illustrating the issues. I gave some examples of pros and cons to using mocks.
Of course, I had to read a bunch of articles and book excerpts, and writing running code, as part of my own process of learning and understanding the controversy; these were resources I had bookmarked a year ago when I had first heard about it. Hooray for talk-driven learning!
Since I was not going to have time, I deliberately had no plan to discuss what I think is quite interesting, which is the philosophy that drives a strictly mock-driven approach to testing. This philosophy involves rigorously taking the outside-in approach as a matter of design. I have to confess that I do not currently take this extreme approach, but it intrigues me, and now that I know better what the techniques and controversies are, I may give it a try in new projects in order to see how well it works.
I felt more relaxed when giving this talk than I did in my previous one (for Pittsburgh Ruby). Just this sheer fact probably helped some.
I continue to use landslide to generate an HTML slide show.
Steve Gross gave a little demo of Selenium WebDriver first, and took questions, before I gave my half-hour talk.
I tried to ask useful questions of the audience and answer people’s questions during the talk. This is a kind of skill that has to come from experience, because one can’t anticipate everything ahead of time, and one has to gauge where people’s minds might be. I know that when I’m at someone’s talk, and I’m confused or lost or bored, I appreciate when the speaker can tell, and can adjust accordingly. This is clearly key. With forty people around, however, man, it’s hard.
I enjoyed speaking at the Pittsburgh Python group meeting, just my second time since my first lightning talk for the group during which I didn’t even have any slides prepared, but just drew some pictures on the whiteboard and scrolled through some code in an Emacs buffer.
Supporting Web site
I have a supporting Web site for the talk with code.
I put a lot of work into the presentation I gave at Pittsburgh Python, and am grateful to Nick Sloan for bringing up the theme and calling for volunteers. I like the idea of a “theme” for a meeting, such as the theme of “testing” that was our theme this last time.
And I have to thank Abby for supporting me as my week got frantic as a result of suddenly volunteering to give my talk, and for attending a talk by me for the first time ever!comments powered by Disqus