Community
Participate
Working Groups
The quick assist needs to be updated to use String#equals() instead of == I will attach a patch.
Good catch. Please make sure you generate the checks as "if (key.equals(value))". That will also preserve the NPE if key == null.
(In reply to comment #1) > Good catch. Please make sure you generate the checks as > "if (key.equals(value))". That will also preserve the NPE if key == null. hmm... I had implemented this as "if (value.equals(key))". I know that in general quick assists should preserve behavior, however I wonder if in this case we should generate the 'right' code i.e. one that does not result in an NPE. I am assuming most users will find it desirable to avoid the NPE.
We should not silently change behavior. An alternative would be to add an explicit if (key == null) throw new NullPointerException(); in front that can easily be deleted. But we would look bad if the compiler immediately flags the null check as unnecessary... So maybe it's better to just add " (removing null check on condition)" to the quick assist label in this case.
Created attachment 197671 [details] fix + tests (In reply to comment #3) > So maybe it's better to just add " (removing null check on condition)" to the > quick assist label in this case. I have changed the label to "Convert 'switch' to 'if-else' (removing null check on switch expression)".
Fixed in BETA_JAVA7
Fixed NPE in BETA_JAVA7 of AdvancedQuickAssistProcessor when switch expression type could not be resolved (e.g. an unresolved variable).
verified
The label of that quick fix is too long and not clear. I've removed it (HEAD and R3_7_maintenance) and filed bug 353759 to have a generic way for marking quick fixes/assists that change the code semantics.
