Unfixable CodeSat, Jun 21, 2014 in Programming
Over the years I’ve come to believe something that I’m not sure others will agree with. I would like to hear your point of view on it.
I posit that some code can become literally unfixable. Programmers can paint themselves into a corner with the code and it becomes impossible to get out again.
The scenario arises when a specific set of conditions exists:
- the code is difficult to work with for some reason — inherent complexity, badly written code, whatever
- the code is not tested and is not written to be testable
- bugs are high-stakes, so changes are extremely risky
- the code needs to be changed for some reason (it’s buggy or doesn’t fulfill its requirements)
This situation is a deadlock. I can’t leave the code as it is, because it has bugs that are causing problems that need to be fixed. I can’t change the code, or I’ll cause bugs that may be truly serious in a variety of ways. I can’t test the code, because it’s untestable the way it’s written. I can’t refactor it to be testable for the same reasons I can’t just change it in the first place.
I have encountered such code many times in my career. I’ve also met people who say they can fix any code with enough work. I haven’t seen that happen. The only solution I’ve seen is a complete replacement, rewriting from scratch.
What do you think?
PS: I remember seeing claims that if more than a small amount (10% or so, as I recall?) of a program’s code needs to be modified, a rewrite from scratch will be a better outcome in less time. If you can find any sources for that claim, please leave them in the comments.
I'm Baron Schwartz, the founder and CEO of VividCortex. I am the author of High Performance MySQL and many open-source tools for performance analysis, monitoring, and system administration. I contribute to various database communities such as Oracle, PostgreSQL, Redis and MongoDB.