2 ms·
I agree that understanding the problem space is exponentially more powerful than just being able to make code do what you want (or figure out why it doesn't).
by kragen 5y ago
I agree that understanding the problem space is exponentially more powerful than just being able to make code do what you want (or figure out why it doesn't).
However, I don't find this kind of thing insulting.
Unfortunately "how do you work with garbage code written by shitty programmers" is a pretty important question, though not one I have to worry about quite as much when I'm working on my own projects where I've written all the code. You might think that at a startup the garbage code wouldn't've had time to pile up, but that isn't my experience. Often founders have a better grasp of the problem space than of the solution space (I mean, they're often shitty programmers), and even their grasp of the problem space is weak at first; this can lead to multiple rapid pivots in which code that was good for one thing is patched as quickly as possible to be barely suitable for something else, with the expectation that it will probably be thrown away anyway, and if not, they can hire someone else to rewrite it.
I'm surprised by your thoughts on the bug test. Usually when I fix a bug, any bug, my first step is to construct a test case that reliably demonstrates the bug, which is what it sounds like they asked you to do. But you say, "this diagnostic technique of arbitrarily working backwards ... is probably the worst method of programming." What do you do instead? Or have I misunderstood your description?
My experience with hiring programmers (18 years ago) was that a short emailed test with a few simple programming problems I could do in a few minutes was enough to easily rule out about 90% of applicants, including people whose credentials looked great on paper. Of course it tells you almost nothing about their understanding of the problem domain or their ability to design code to be maintainable in the large, but if they can't program then they can't do the job even if they're good at those things.
One deficiency in our process at the time was that we didn't test for the ability to read code, just write it, and we generally spend about five times as much of our time reading code as writing it. The debugging question sounds like a useful technique for giving some insight into that.