Reset unknownTypeRefs after silent type checking

[rdulmina]

Purpose

$title

Fixes #43286

Approach

Compiler error at L#1749 was not reached because the field unknownTypeRefs was populated at the silent typechecking and does not reset after.

ballerina-lang/compiler/ballerina-lang/src/main/java/org/wso2/ballerinalang/compiler/semantics/analyzer/SymbolResolver.java

Line 1749 in a6e4f99

dlog.error(userDefinedTypeNode.pos, data.diagCode, typeName);

Introduced GlobalStateData inner class to make the solution scalable for such fields may occur in future

Samples

The following sample should give unknown type error of unknown types I and J

class H {};

H res = check new I();

int[] a = [1, 2, 3, 4, 5];

int[] b = from var i in a select <I> i;

float result = <J>1 + 2.0;

int result2 = true? <J>1 : 2;

var result3 = <H> new J();

Check List

• Read the Contributing Guide
• Updated Change Log
• Checked Tooling Support (#)
• Added necessary tests
  • Unit Tests
  • Spec Conformance Tests
  • Integration Tests
  • Ballerina By Example Tests
• Increased Test Coverage
• Added necessary documentation
  • API documentation
  • Module documentation in Module.md files
  • Ballerina By Examples

[MaryamZi]

Can't we do this via analyzer data instead of introducing a new data holder?

[rdulmina]

AnalyzerData act like set of global fields where we pass through the functions. GlobalStateData is used as a snapshot of several fields where we will be creating multiple of objects of it. However, we can still use the AnalyzerData for the same purpose. The issue is it will introduce unwanted object creations of the AnalyzerData which is not clean

[MaryamZi]

The addition is basically resetting unknown type refs, right? Can we just do it inline instead of introducing GlobalStateData? The name is ambiguous and can be confusing with analyzer data IMO.

[rdulmina]

Writing this inline is not a good idea in terms of scalability. We already have two fields that needs to be reset. There is a high chance of more fields require this in future. I have renamed the class name to GlobalStateSnapshot

[MaryamZi]

The only additional thing we've done is changing unknown type refs, right? Error count logic was already there? It makes sense to extract out repeated logic to a function, but IMO, here we are just doing it for part of it. I would rather refactor the whole thing properly or just reset type refs in line (similar to the others). With just what we have now, I also don't think GlobalStateSnapshot is an ideal name here, since it doesn't include all "global state" and can be confusing with analyzer data.

I would introduce the change in a way that makes it readable and clear at the moment, rather than anticipating future changes (when we can refactor everything properly).

[MaryamZi]

The addition is basically resetting unknown type refs, right? Can we just do it inline instead of introducing GlobalStateData? The name is ambiguous and can be confusing with analyzer data IMO.