> "WARNING As a rule of thumb, enums are code smells and should be refactored to polymorphic classes. [8]" Seemann, Mark, Dependency Injection in .Net, 2011, p. 342
Gosh, what a load of baloney from a frog in a well.
Gosh, what a load of baloney from a frog in a well.
An enum is essentially an encoding over integers. I guess this depends on your programming language, so humour me.
If you have an integer whose meaning is a category (as opposed to a count or measure) - does this need refactoring into classes?
Probably not most of the time.
For example you import a CSV and one of the fields is a categorisation. Let’s say county code. Your code may occasionally care if you are in an EU country.
You could have a class per country and override isEU as needed but this is clumsy and ugly in my opinion.
You’d also need a switch statement, in a factory method to create these from the csv data.
A lot of the time it’s easier to follow code where the switch statement is in its intuitive place.
Highly OO abstracted code can be very hard to reason about too.
Definitely. Coming up with all the right design patterns and abstractions can feel quite clever and satisfying. But when someone else needs to debug that code, having to jump through layers and layers of object, interfaces, factories, etc to find out what a method call is actually doing is a pain.
I think in a lot of cases it’s definitely worth stopping and thinking, and making sure you are implementing the right thing if you find yourself using enums heavily, or using a class hierarchy as an enumerator replacement.
Of course, a catch-all would then somewhat reintroduce the problem.
Mark, for example, is now primarily a functional programmer and makes good use of F#’s type system:
https://blog.ploeh.dk/2015/08/10/type-driven-development/
So he might even agree with you these days (on the “baloney” part, but maybe not the “frog in a well”).
Then we end up with "corporate java code" that wants to throw out what was the basis of computing (if statements, switches) and turn all into some crazy type hierarchy.
Modern development has drifted toward lots of small functions (which causes a mess because modern IDEs dont have supporting features like this: https://youtu.be/baxtyeFVn3w?t=1786, re: gtoolkit). A switch requires you to look inside and reason about the internals of a function to understand how it will behave before exiting, as the compiler. This doesn't matter so much if you dont care about early exits (no branch prediction, etc). Either way, it's extra cognitive overhead and error prone for a small improvement in readbaility. It's the line where programming becomes tricky programming.
As a thought experiment, imagine there is ANOTHER keyword that acts like an if/else (or switch with mandatory breaks). If A. it's terse, B. it's structured, C. it's readable, would that be better?
We primarily use switch because it's easy to read (versus a ton of if{}/else{}), as it gives us almost-block scoping without having to define another indirection/function. Python has this switch equivalent in match/case, but you have to use Python to get it.
Switch use is a syntax problem. Lots of software is colored by bad syntax and language design^.
^ I think most people realize that type hierarchies are a brittle/limited form of mixin. Even Java has softened to this a bit (aspects, etc) but will likely never be fun to use.
The story itself seems to be ancient, but I couldn't find where it originates from.
I was able to find some references from ancient India, but nothing about it coming from there.
- There are many many languages without classes that make good use of enums
- Doing this refactoring is excessively verbose
- Someone making bold claims like this about a language feature that's never been considered error prone and been around since the beginning ought to provide some really good evidence to back things up.
There's cases where enums are good, and there's cases where polymorphic classes are good. Dismissing one of them by default is wrong, since both are useful at different times.
edit: and the citation there to Martin Fowler et al., Refactoring: Improving the Design of Existing Code doesn't even match the claim being made. That book mentions
- "The first part of this problem is that switch statement. It is a bad idea to do a switch based on an attribute of another object. If you must use a switch statement, it should be on your own data, not on someone else's." -- sure, I agree
- "Often you find the same switch statement scattered about a program in different places. If you add a new clause to the switch, you have to find all these switch, statements and change them. The object-oriented notion of polymorphism gives you an elegant way to deal with this problem." -- fair enough
This is a long way from "enums are a code smell," and honestly feels like padding the citation count. Another thing to note is that the second edition of Refactoring: Improving the Design of Existing Code says:
> Even in our more wild-eyed youth, we were never unconditionally opposed to the conditional. Indeed, the first edition of this book had a smell entitled “switch statements.” The smell was there because in the late 90’s we found polymorphism sadly underappreciated, and saw benefit in getting people to switch over.
> These days there is more polymorphism about, and it isn’t the simple red flag that it often was fifteen years ago.
Even that seems like it could easily be way overkill e.g. if you have multiple switches which, say, generate a label from an enum, the first step is probably to add a utility function / method, not to migrate the whole thing over to polymorphism.
Although there is one thing to be said about context:
* OP works in C#, whose enums are literally useless (they’re like C’s)
* apparently even in Java (which at least has type-safe enums even if not sum types), `switch` is unable to check for completeness
So when you have an enum-typed value, odds are good that it’s one of the named ones but there’s no mechanism anywhere preventing it to be any other integer of the underlying type.
Any caller can send any garbage (if you're publishing a package / API), likewise a dependency can return any garbage, etc... C#'s enums are entirely indicative.
Which, in my experience is a huge advantage rather than the supposed disadvantage Fowler is saying it is (although I could see that being the case in many other languages).
originally the meaning of code smell was not that code that had the smell was bad, but there was a chance it was and should be examined. For this reason of course it is useful to get rid of code smells so that people don't feel the need to investigate hey is this smelly code actually bad code.
But all that said not sure if an enum is a code smell.
on edit: this at any rate was the rationale behind the phrase code smell I was first introduced to.
Sometimes these polymorphic classes make sense, but my experience is that it’s a minority, and it’s better to err on the side of some simple enums.
Also note that Uncle Bob in Clean Code pretty much directly opposes in one of the chapters about logic and data separation (can’t find the chapter right now since I don’t have the book anymore)
In contrast, an alternative like polymorphism + a visitor is safe, but excruciatingly verbose and hard to follow and modify.
Either you end up with 'scary' enums, or with dynamic_cast and 'scary' null pointers.
Enums are essentially categorical variables, and there is nothing wrong with that: they are widely used in statistics and other real-world applications (arguably, even, booleans are a case of such.)
As a junior programmer, I was once assigned to work on an example of what could happen as a result of following this advice: an explosion of subclasses for the sake of avoiding a few conditional clauses. Algorithms had been broken up so that the common parts could be put in base classes, and you had to jump all around the code to see what was going on. Worse, objects could change their categorization during their lifetime, which required the substitution of a new object for the previous one. This complexity was contagious; once this one type was implemented in this manner, a lot of other things had to follow suit.
This was one of the experiences that turned me into a skeptic with regard to the proposition that there's no problem that cannot be simply solved by being more object-oriented.
See this pattern all the time in both technical and people side all the time.
“You should never estimate using time”
“Don’t start new work until all the teams sprint commitments are completed”
Etc.
In a word it is arrogance with a twist of ignorance.
Ignorance that complex systems of humans and code don’t follow perfect dogmatic rules.
It might be ok for junior devs to be a bit like this but always add a pinch of “most of the time, but when experienced you’ll know when to ignore the rule”
Ignoring some unwritten or written dogma is often what makes for a competitive advantage.
Let your competitor drown in EnumClassFactorys .
I think the best authoritative-sounding counter-quip is "Best practices are best not practiced."