While coding today, I had a realization: refactoring can introduce subtle bugs.
I’m not just talking about variable name changes that break logic where the variable names are referenced external to the code (ex. request parameters in a web application).
I’ll give a quick example. Here’s a code snippet before inline refactoring:
boolean saidHi = false;
for (int i = 0; i < 2; i++) {
String message = saidHi ? "Hello again" : "Hello";
saidHi = true;
System.out.println(message);
}This will output:
Hello
Hello again
If you then apply inline to the message variable because, let’s say, you like it a little more succinct, then you get this:
boolean saidHi = false;
for (int i = 0; i < 2; i++) {
saidHi = true;
System.out.println(saidHi ? "Hello again" : "Hello");
}This will output:
Hello again
Hello again
A full class with the function both ways:
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
class BadRefactor {
public static void main(String[] args) {
beforeRefactor();
System.out.println("--------"); // print separator
afterRefactor();
}
public static void beforeRefactor() {
boolean saidHi = false;
for (int i = 0; i < 2; i++) {
String message = saidHi ? "Hello again" : "Hello";
saidHi = true;
System.out.println(message);
}
}
public static void afterRefactor() {
boolean saidHi = false;
for (int i = 0; i < 2; i++) {
saidHi = true;
System.out.println(saidHi ? "Hello again" : "Hello");
}
}
}
This was using the inline refactoring in Eclipse Juno SR1.
So be careful when using refactoring; if using Eclipse, use the preview feature to make sure there are no unwanted side effects before applying.
Have you thought of, or run into any cases where refactoring messed things up in a subtle way? Feel free to comment.