[GROOVY-11453] Spread-safe operator differences and efficiencies

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Fix Version/s: 5.0.0-alpha-11
• Component/s: Static compilation
• Labels: breaking

Language

groovy

• Assignee: Eric Milles
• Reporter: Eric Milles
• Created: 22/Aug/24 15:12
• Updated: 4 days ago 11:39
• Resolved: 29/Sep/24 16:47

Description

The spread-safe operator – aka a*.b or a*.b(c) – has some differences and inefficiencies.

1. When the receiver is null, the expression produces null for dynamic mode and an empty list for static compilation. Can this difference be safely resolved?
2. When the method return type is void, there is no need for a list to collect up nulls. This is producing "Useless object stored in variable var$1" warnings from SpotBugs. The only benefit of this is the ability to detect the number of elements.
3. The SC bytecode has a redundant null test in it. Since a for loop is also null-safe, guarding the loop bytecode by a null check is producing "Redundant check for null" warnings from SpotBugs. NOTE: If (1) above is changed in the SC classgen, the null check may not be redundant as it would guard list construction.

For reference, object*.method(arguments) is written by the static compiler like this:

def result = new ArrayList()
if (object != null) {
  for (Inferred element : object) { // also null-safe
    result.add(element?.method(arguments)
  }
}
result

And the dynamic impl: https://github.com/apache/groovy/blob/f46103d31b7574ce883ff3b6c01fa8a412f225be/src/main/java/org/codehaus/groovy/runtime/ScriptBytecodeAdapter.java#L189

GROOVY-7177