[GROOVY-10007] Breaking change in property resolution 2.4 -> 3.0

• Type: Bug
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Affects Version/s: 3.0.5, 3.0.7
• Component/s: Compiler
• Labels: breaking_change metaprogramming
• Environment: OpenJDK 11, IdeaC 2020.3.3
• Assignee: Unassigned
• Reporter: Jochen Eddelbuettel
• Created: 31/Mar/21 16:42
• Updated: 31/Mar/21 16:42

Description

I've created a library utilizing propertyMissing. By convention, it expects property names to be running against camel casing conventions (starting with a single upper case letter), to avoid conflicts with actual properties (bean style, with getters).

In Groovy 2.4 a call of obj.Owner would not be resolved to obj.getOwner(). Only obj.owner would be equivalent to obj.getOwner(). In Groovy 3.0 obj.Owner results in a call to obj.getOwner() as well. As a result, no call to propertyMissing for "Owner" occurs any more. Thus properties that used to be resolved by metaprogramming, can now result in actual method calls, which can potentially have a huge impact on the behavior of existing code.

I've looked back into the 3.0.0 release notes and there is no mention of this being a breaking change.

This script, who assertions were jolly fine in 2.4.15, should behave the same in 3.0.x:

import groovy.json.JsonSlurper
JsonSlurper.metaClass.propertyMissing = { String name -> name }
def slurp = new JsonSlurper()
slurp.maxSizeForInMemory = 500000
assert slurp.maxSizeForInMemory == 500000
assert slurp.MaxSizeForInMemory == "MaxSizeForInMemory"

Interestingly, this is not a problem for setters, which suggests to me that the change is unintended behaviour get getters and should be fixed. It is a problem with @CompileStatic too. obj.Owner statically compiles to obj.getOwner().